Fault prediction method and apparatus, electronic device and storage medium

Abstract

The application provides a fault prediction method and apparatus, an electronic device and a storage medium. The method comprises: obtaining a vibration signal collected by a sensor installed on a target component of a target device, and cleaning and segmenting the vibration signal to obtain a target vibration signal; inputting the target vibration signal into a target anomaly judgment model to obtain an output result, and according to the output result, determining whether the target vibration signal is abnormal to obtain an anomaly judgment result; when the abnormal judgment result shows that the target vibration signal is abnormal, inputting the target vibration signal into a target fault prediction model to obtain a fault prediction result, wherein the target fault prediction model is obtained by combining an encoder of the target anomaly judgment model and a newly added full connection layer. The application solves the problem in the prior art that a faulty device cannot be timely maintained because the fault can only be monitored when the device fault is quite obvious.

Description

Fault prediction method, device, electronic device and storage medium

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on November 9, 2023, with application number 202311494002.6 and invention name “Fault monitoring method, device, storage medium and electronic device”, the entire contents of which are incorporated by reference into this application.

[0002] This application claims priority to the Chinese patent application filed with the China Patent Office on November 9, 2023, with application number 202311495365.1 and invention name “Fault Prediction Method, Device, Storage Medium and Electronic Device”, the entire contents of which are incorporated by reference into this application.

[0003] This application claims priority to the Chinese patent application filed with the China Patent Office on November 24, 2023, with application number 202311585660.6 and invention name “A sensor fault identification method, device, electronic device and storage medium”, the entire contents of which are incorporated by reference into this application.

[0004] This application claims priority to the Chinese patent application filed with the China Patent Office on November 24, 2023, with application number 202311587076.4 and invention name “Moving equipment fault monitoring method, device, electronic device and storage medium”, the entire contents of which are incorporated by reference into this application. Technical Field

[0005] The present application relates to the field of equipment fault diagnosis, and more specifically, to a fault prediction method, apparatus, electronic device, and storage medium. Background Art

[0006] Reciprocating equipment plays a vital role in various industrial sectors, particularly in the energy, chemical, and manufacturing sectors. For example, reciprocating compressors are widely used in oil and natural gas refining, chemical production, and air and other gas compression. To ensure continuous and efficient production processes, the efficiency and reliability of reciprocating equipment are crucial.

[0007] However, various components of reciprocating equipment, such as pistons, cylinders, valves, packings, and valve bodies, will malfunction due to long-term operation and wear. For example, if the piston rings of a reciprocating compressor are excessively worn, it may lead to reduced compression efficiency and even cause the compressor to malfunction.

[0008] Therefore, it is essential to monitor the health and integrity of reciprocating equipment. Regular maintenance and inspections can help identify and correct equipment failures, thus preventing damage and interruptions to production processes. This is crucial for ensuring the efficient operation of the energy and other industries.

[0009] Currently, when monitoring the operating status of reciprocating equipment, equipment failure can usually only be detected when the failure occurs and has a significant impact on the operation of the equipment. However, in this case, the equipment has been running in the faulty state for a long time. Therefore, this monitoring method has poor timeliness and cannot detect equipment failures and perform maintenance on the equipment in a timely manner.

[0010] Currently, no effective solution has been proposed to the problem that in related technologies, a fault can only be detected when the equipment fault is relatively obvious, resulting in an inability to maintain the faulty equipment in a timely manner. Summary of the Invention

[0011] The present application provides a fault prediction method, device, electronic device and storage medium to improve the speed and efficiency of fault monitoring and timely maintain the target equipment according to the cause of the fault.

[0012] In a first aspect, the present application provides a fault prediction method, comprising:

[0013] Acquire a vibration signal collected by a sensor installed on a target component of a target device, and clean and segment the vibration signal to obtain a target vibration signal;

[0014] Input the target vibration signal into the target abnormality judgment model to obtain an output result, and determine whether the target vibration signal is abnormal based on the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is trained by the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder;

[0015] When the abnormality judgment result indicates that the target vibration signal is abnormal, the target vibration signal is input into the target fault prediction model to obtain a fault prediction result. The target fault prediction model is obtained by combining the encoder of the target abnormality judgment model and the newly added fully connected layer. The target fault prediction model is trained using multiple fault types and the vibration signals of each fault type as samples.

[0016] In combination with the first aspect, in a first possible implementation method of the first aspect, cleaning and segmenting the vibration signal to obtain a target vibration signal includes: segmenting the vibration signal according to the time dimension to obtain multiple first vibration signals, wherein each first vibration signal includes vibration signals of the same duration, and the duration includes multiple continuous vibration cycles; judging in turn whether there is an error signal in each first vibration signal, wherein the error signal represents abnormal operation of the sensor; in the case of a first vibration signal with an error signal, deleting the first vibration signal with the error signal from the multiple first vibration signals to obtain multiple second vibration signals; and inputting each second vibration signal into a preset filter in turn to obtain multiple target vibration signals.

[0017] In combination with the first aspect, in a second possible implementation method of the first aspect, the target abnormality judgment model is trained in the following manner: obtaining a sample vibration signal set of the target component, wherein the sample vibration signal set includes multiple sample vibration signals generated by the target component under normal operation; inputting the sample vibration signals in the sample vibration signal set into the initial abnormality judgment model, and processing to obtain an output signal corresponding to each sample vibration signal, wherein the initial abnormality judgment model is composed of an encoder and a decoder, the encoder is used to extract the characteristics of the sample vibration signal and reduce the data dimension, and the decoder is used to restore the data dimension and restore the data characteristics; determining the difference value between each output signal and the corresponding sample vibration signal to obtain multiple difference values, and judging whether there is a difference value greater than a preset threshold among the multiple difference values; when there is a difference value greater than the preset threshold among the multiple difference values, changing the neuron connection weights in the initial abnormality judgment model, and retraining the changed initial abnormality judgment model until there is no difference value greater than the preset threshold among the multiple difference values; when there is no difference value greater than the preset threshold among the multiple difference values, obtaining the target abnormality judgment model.

[0018] In combination with the first aspect, in a third possible implementation method of the first aspect, determining whether the target vibration signal has an abnormality based on the output result, and obtaining the abnormality judgment result includes: determining the difference value between the output result and the target vibration signal, obtaining the target difference value, and judging whether the target difference value is greater than a preset threshold; when the target difference value is greater than the preset threshold, determining that the target vibration signal has an abnormality; when the target difference value is less than or equal to the preset threshold, determining that the target vibration signal does not have an abnormality.

[0019] In combination with the first aspect, in a fourth possible implementation method of the first aspect, the target fault prediction model is trained by: obtaining multiple fault types and historical vibration signals under each fault type to obtain multiple groups of historical vibration signals; adding a label to each historical vibration signal in each group of historical vibration signals according to the fault type to obtain multiple groups of updated historical vibration signals; using the multiple groups of updated historical vibration signals as samples to train the initial fault prediction model to obtain the target fault prediction model.

[0020] In combination with the first aspect, in the fifth possible implementation of the first aspect, after obtaining the fault prediction result, the method also includes: obtaining a historical abnormal vibration signal collected by the sensor when an abnormality occurs in the target component; determining the characteristic information of the vibration signal of the target component based on the historical abnormal vibration signal to obtain multiple characteristic information, and obtaining the characteristic value corresponding to each characteristic information; determining the judgment standard of each characteristic information based on the characteristic value corresponding to each characteristic information to obtain multiple judgment standards; obtaining the vibration data value corresponding to each characteristic information in the target vibration signal, and using the same judgment standard of the characteristic information to perform abnormal judgment on the vibration data value to obtain a judgment result; when the judgment result indicates that there is no abnormality in the target vibration signal, determining that there is an abnormality in the target abnormality judgment model, and issuing an alarm message, wherein the alarm information indicates that there is an abnormality in the target abnormality judgment model.

[0021] In combination with the first aspect, in the sixth possible implementation method of the first aspect, the vibration data value is judged as abnormal using the judgment criteria with the same characteristic information, and the judgment result obtained includes: obtaining the vibration data value under each characteristic information and the judgment criteria under the same characteristic information, and determining whether the vibration data value meets the corresponding judgment criteria, and obtaining multiple sub-judgment results, wherein the sub-judgment results are used to characterize whether the vibration data value meets the corresponding judgment criteria; when there are sub-judgment results that do not meet the judgment criteria among the multiple sub-judgment results, the judgment result is determined to be that there is an abnormality in the target vibration signal; when there are no sub-judgment results that do not meet the judgment criteria among the multiple sub-judgment results, the judgment result is determined to be that there is no abnormality in the target vibration signal.

[0022] In a second aspect, the present application further provides a fault prediction device, which may include a unit module for executing the method steps in various implementations of the first aspect.

[0023] In a third aspect, the present application further provides a fault monitoring method, which is characterized by comprising:

[0024] Acquiring an initial vibration signal collected by a preset sensor, wherein the preset sensor is disposed at a target position of a target device, and the preset sensor is used to collect the vibration signal at the target position;

[0025] Filtering the initial vibration signal by a zero-phase filtering method to obtain an initial high-frequency signal;

[0026] Obtaining a reference key phase signal of a reference device, and determining a mapping relationship between a time in the reference key phase signal and a rotation angle of the reference device according to a rotation speed of the reference device, and converting the abscissa of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device;

[0027] Obtain standard data of the target device, detect the target angular vibration signal using the standard data to obtain a detection result, and determine the fault state of the target device based on the detection result, wherein the standard data is data in the vibration signal of the target device during normal operation.

[0028] In conjunction with the third aspect, in a first possible implementation of the third aspect, filtering the initial vibration signal by a zero-phase filtering method to obtain the initial high-frequency signal includes:

[0029] Obtaining a preset high-pass filter, and inputting the initial vibration signal into the preset high-pass filter to obtain a first time domain signal;

[0030] Flipping the first time domain signal around the center line of the X-axis, and inputting the flipped first time domain signal into the preset high-pass filter to obtain a second time domain signal;

[0031] The second time domain signal is flipped around the center line of the X-axis to obtain the initial high-frequency signal.

[0032] In combination with the third aspect, in a second possible implementation manner of the third aspect, after obtaining a reference key phase signal of a reference device, the method further includes:

[0033] Determining a rotation period of the reference device according to the rotation speed of the reference device, and intercepting a signal within a preset time period from the reference key phase signal to obtain an updated reference key phase signal, wherein the preset time period includes a plurality of continuous rotation periods;

[0034] intercepting a signal within the preset time period from the initial high-frequency signal to obtain an updated initial high-frequency signal;

[0035] Determining a mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and converting the abscissa of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship, to obtain a target angular domain vibration signal includes:

[0036] The mapping relationship between the time in the updated reference key phase signal and the rotation angle of the reference device is determined according to the rotation speed of the reference device, and the updated initial high-frequency signal is converted according to the mapping relationship to obtain a target angular domain vibration signal.

[0037] In conjunction with the third aspect, in a third possible implementation of the third aspect, the reference device is provided with a code disk, the rotation speed of the reference device is the same as the rotation speed of the code disk, and determining the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device includes:

[0038] Determine the number of revolutions of the code disk within the preset time period according to the rotation speed to obtain a preset number of revolutions;

[0039] The rotation angle corresponding to the preset number of revolutions is determined, and the duration required for each rotation of 1 degree is determined according to the rotation angle and the preset time period to obtain the mapping relationship.

[0040] In combination with the third aspect, in a fourth possible implementation of the third aspect, converting the abscissa of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain the target angular domain vibration signal includes:

[0041] Converting the time value on the time axis of the initial high-frequency signal into a preset angle value according to the mapping relationship to obtain an initial angular domain vibration signal;

[0042] Determine whether each preset angle value has a corresponding amplitude value in the initial angular domain vibration signal, and if the target preset angle value does not have a corresponding amplitude value, obtain multiple preset angle values ​​adjacent to the target preset angle value and the amplitude value of each preset angle value, use interpolation to calculate the amplitude value of the target preset angle value based on the amplitude value of each preset angle value, and add the amplitude to the corresponding position in the initial angular domain vibration signal to obtain the target angular domain vibration signal.

[0043] In combination with the third aspect, in a fifth possible implementation of the third aspect, detecting the target angular vibration signal using the standard data to obtain a detection result, and determining the fault state of the target device based on the detection result includes:

[0044] Acquire a calibration angle and a standard amplitude at the calibration angle from the standard data;

[0045] Determining the amplitude at the calibration angle in the target angle domain vibration signal to obtain a target amplitude, and determining whether the target amplitude is greater than the standard amplitude;

[0046] When the target amplitude is greater than the standard amplitude, determining that a fault exists in the target device;

[0047] When the target amplitude is less than or equal to the standard amplitude, it is determined that the target device does not have a fault.

[0048] With reference to the third aspect, in a sixth possible implementation of the third aspect, determining the amplitude at the calibration angle in the target angular domain vibration signal to obtain the target amplitude includes:

[0049] Obtaining an angular range in the target angular domain vibration signal, and obtaining the number of rotation periods included in the angular range to obtain the number of periods, wherein the angular range is greater than 360 degrees;

[0050] Calculating the angle value corresponding to the calibration angle in each cycle to obtain a plurality of preset angles;

[0051] A target angle corresponding to each preset angle in the target angle domain vibration signal is determined to obtain a plurality of target angles, and an amplitude at each target angle is acquired to obtain a plurality of target amplitudes.

[0052] In a fourth aspect, the present application further provides a fault monitoring device, which may include a unit module for executing the method steps in various implementations of the third aspect.

[0053] In a fifth aspect, the present application further provides a method for monitoring a fault of a moving device, which is applied to a moving device fault monitoring system, and the method comprises:

[0054] Determine multiple test points of the target device and the parameter matching rules for each test point;

[0055] For each test point, matching the target parameter data of the test point according to the corresponding parameter matching rule;

[0056] determining a fault symptom quantity of the test point according to the target parameter data;

[0057] Determining the operating status of each test point based on the fault symptom quantity corresponding to each test point;

[0058] When it is determined that the target moving device has a fault through the operating status of each test point, the fault point and fault type of the target moving device are determined according to the operating status of each test point.

[0059] In conjunction with the fifth aspect, in a first possible implementation of the fifth aspect, matching target parameter data of each test point according to a corresponding parameter matching rule includes:

[0060] For each test point, target parameter data corresponding to the test point is obtained from a preset database according to a parameter matching rule corresponding to the test point, wherein the database pre-stores the following parameter data of the target mobile device: static attribute data of the target mobile device, maintenance and warranty data of the target mobile device, operating data of the target mobile device, and sensor signal data of the target mobile device collected by a sensor;

[0061] The method further comprises:

[0062] Get the preset sensor collection rules;

[0063] According to the sensor acquisition rule, the corresponding sensor is controlled to acquire the corresponding sensor signal data.

[0064] In conjunction with the fifth aspect, in a second possible implementation of the fifth aspect, determining a parameter matching rule for each test point includes:

[0065] Obtaining the position information of each test point on the target moving device;

[0066] Determining a parameter matching rule for each of the test points according to the location information;

[0067] Determining the fault symptom quantity of the test point according to the target parameter data includes:

[0068] Determining, based on the location information of the test point, a characteristic value calculation rule and a fault symptom quantity calculation rule corresponding to the test point;

[0069] Calculating the target parameter data according to the characteristic value calculation rule to obtain at least one characteristic value corresponding to the test point;

[0070] The target parameter data and / or the characteristic value are combined according to the fault symptom calculation rule to generate the fault symptom of the test point.

[0071] In conjunction with the fifth aspect, in a third possible implementation manner of the fifth aspect, determining the operating status of each test point based on the fault symptom quantity corresponding to each test point includes:

[0072] Obtaining the maximum speed and the minimum speed of the target moving equipment during multiple operation cycles;

[0073] determining a speed difference between the maximum speed and the minimum speed, and determining whether the speed difference is less than a preset difference threshold;

[0074] When it is determined that the rotational speed difference is less than the difference threshold, the operating state of each test point is determined based on a preset dynamic warning model and a fault symptom quantity corresponding to each test point.

[0075] In conjunction with the fifth aspect, in a fourth possible implementation of the fifth aspect, determining the operating status of each test point based on a preset dynamic early warning model and a fault symptom quantity corresponding to each test point includes:

[0076] Obtaining historical operating data of the target dynamic equipment in a preset historical time period and configuration parameters of a preset dynamic early warning model;

[0077] Inputting the historical operation data, the fault symptom quantity corresponding to each test point, and the configuration parameters into the dynamic early warning model, and obtaining the early warning result corresponding to each fault symptom quantity output by the dynamic early warning model;

[0078] Based on the early warning result corresponding to each fault symptom, the operating state of each test point is determined, wherein the operating state includes a fault state and a non-fault state.

[0079] In conjunction with the fifth aspect, in a fifth possible implementation of the fifth aspect, the fault sign quantity includes a sensor abnormality sign quantity, and determining the operating status of each test point based on the early warning result corresponding to each fault sign quantity includes:

[0080] Determining whether the warning result corresponding to the abnormal sign quantity of the sensor is a fault warning;

[0081] If it is determined that the warning result corresponding to the sensor abnormality sign quantity is a fault warning, an alarm message of sensor abnormality is output and the process ends;

[0082] If it is determined that the warning result corresponding to the abnormal sign of the sensor is a non-fault warning, obtaining sensor signal data collected by the sensor;

[0083] Inputting the sensor signal data into a preset sensor fault identification model to obtain a sensor abnormality identification result output by the sensor fault identification model;

[0084] If the sensor abnormality identification result indicates that the sensor is abnormal, determining that the operating state of each test point is a non-fault state;

[0085] If the sensor abnormality identification result indicates that the sensor is normal, the warning result level of each fault symptom corresponding to each test point is determined, and the fault state corresponding to the warning result with the highest warning result level is determined as the operating state of the test point.

[0086] In conjunction with the fifth aspect, in a sixth possible implementation manner of the fifth aspect, determining whether the target moving device has a fault through the operating status of each test point includes:

[0087] If it is determined that the operating state of any test point is a fault state, determining that the target moving device has a fault;

[0088] Determining the fault point and fault type of the target moving device according to the operating status of each test point includes:

[0089] Determine whether the operating status of each test point is a fault state;

[0090] Determine the test point whose operating state is a fault state as the initial fault point;

[0091] For each of the initial fault points, the early warning result corresponding to the initial fault point is input into a preset fault detection model to obtain the fault point and fault type of the target moving equipment output by the fault detection model.

[0092] In conjunction with the fifth aspect, in a seventh possible implementation of the fifth aspect, the fault detection model includes a typical fault classification model and a fault analysis model, the typical fault classification model is used to determine whether the fault type of the target moving device is a typical fault, and the fault analysis model is used to analyze the fault point and basic fault type of the target moving device. Inputting the early warning result corresponding to the initial fault point into the preset fault detection model to obtain the fault point and fault type of the target moving device output by the fault detection model includes:

[0093] Inputting the early warning result corresponding to the initial fault point into the typical fault classification model and the fault analysis model respectively, to obtain the typical fault classification result output by the typical fault classification model and the fault analysis result output by the fault analysis model;

[0094] A weighted sum is performed on the typical fault classification result and the fault analysis result to obtain the fault point and fault type corresponding to the target moving equipment.

[0095] In conjunction with the fifth aspect, in an eighth possible implementation manner of the fifth aspect, the method further includes:

[0096] Determining whether the fault type is a typical fault type;

[0097] When it is determined that the fault type is a typical fault type, the fault type is matched with a preset typical fault case library to obtain a target solution corresponding to the fault type, wherein the typical fault case library is used to store typical faults and solutions corresponding to each typical fault;

[0098] The target solution is output through a visual interface.

[0099] In conjunction with the fifth aspect, in a ninth possible implementation of the fifth aspect, the method further includes:

[0100] According to preset signal processing rules, signal processing and feature transformation are performed on the parameter data of the target moving device stored in the database to obtain a variety of atlas data;

[0101] The atlas data is output through a visual interface to analyze the target dynamic device according to the atlas data.

[0102] In conjunction with the fifth aspect, in a tenth possible implementation manner of the fifth aspect, after determining the fault point and fault type of the target moving device, the method further includes:

[0103] generating a three-dimensional image of the target moving device;

[0104] Marking the fault point and fault type of the target moving equipment in the three-dimensional image to obtain a target three-dimensional image;

[0105] Output the three-dimensional image of the target and issue an early warning through a preset early warning method.

[0106] In a sixth aspect, the present application further provides a device for monitoring faults of moving equipment, which may include a unit module for executing the method steps in various implementations of the fifth aspect.

[0107] In a seventh aspect, the present application further provides a sensor fault identification method, comprising:

[0108] Obtain the historical sensor data set corresponding to the target sensor to be identified;

[0109] Determining a fault identification model corresponding to the target sensor based on the historical sensor data set;

[0110] When actual sensor data of the target sensor is acquired, at least one preset mechanism model corresponding to the target sensor is acquired, wherein the preset mechanism model is used to identify whether the target sensor itself has a fault and, if the target sensor itself has a fault, the corresponding fault type of the target sensor itself;

[0111] Fault identification is performed on the target sensor according to the fault identification model, at least one of the preset mechanism models and the actual sensor data to obtain a target fault identification result corresponding to the target sensor.

[0112] In conjunction with the seventh aspect, in a first possible implementation manner of the seventh aspect, the fault identification model is used to identify whether the target sensor has a fault caused by environmental interference and, when the target sensor has a fault caused by environmental interference, the type of environmental fault corresponding to the target sensor;

[0113] The performing fault identification on the target sensor according to the fault identification model, at least one of the preset mechanism models, and the actual sensor data to obtain a target fault identification result corresponding to the target sensor includes:

[0114] For each of the at least one preset mechanism model, performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor;

[0115] When all the first fault identification results indicate that the target sensor itself does not have a fault, the actual sensor data is input into the fault identification model, so that the fault identification model outputs a target fault identification result corresponding to the target sensor.

[0116] In conjunction with the seventh aspect, in a second possible implementation manner of the seventh aspect, the fault identification model is used to identify whether the target sensor itself has a fault, a corresponding fault type of the target sensor itself when the target sensor itself has a fault, whether the target sensor has a fault caused by environmental interference, and a corresponding environmental fault type of the target sensor when the target sensor has a fault caused by environmental interference;

[0117] The performing fault identification on the target sensor according to the fault identification model, at least one of the preset mechanism models, and the actual sensor data to obtain a target fault identification result corresponding to the target sensor includes:

[0118] For each of the at least one preset mechanism model, performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor;

[0119] Inputting the actual sensor data into the fault identification model so that the fault identification model outputs a second fault identification result corresponding to the target sensor;

[0120] A target fault identification result corresponding to the target sensor is determined according to all the first fault identification results and the second fault identification results.

[0121] In conjunction with the seventh aspect, in a third possible implementation manner of the seventh aspect, determining the target fault identification result corresponding to the target sensor based on all the first fault identification results and the second fault identification results includes:

[0122] determining whether there is a fault identification result in the second fault identification results that is consistent with all the first fault identification results;

[0123] When there is a fault identification result in the second fault identification results that is consistent with all the first fault identification results, determining the second fault identification result as the target fault identification result corresponding to the target sensor;

[0124] When there is a fault identification result in the second fault identification result that is inconsistent with at least one of the first fault identification results, generating an alarm prompt information according to the inconsistent at least one first fault identification result;

[0125] The alarm prompt information is pushed to the target terminal corresponding to the target sensor.

[0126] In conjunction with the seventh aspect, in a fourth possible implementation manner of the seventh aspect, determining the fault identification model corresponding to the target sensor based on the historical sensor data set includes:

[0127] For each historical sensor data in the historical sensor data set, performing fault classification and labeling on the historical sensor data to obtain training sample data corresponding to the historical sensor data;

[0128] The preset classification model is trained according to the training sample data corresponding to all the historical sensor data in the historical sensor data set to obtain a fault recognition model corresponding to the target sensor.

[0129] In conjunction with the seventh aspect, in a fifth possible implementation manner of the seventh aspect, at least one of the preset mechanism models includes a bias voltage model, an output signal model, and a spectrum model; the actual sensor data includes an actual bias voltage, an actual output signal value, and an actual spectrum; each of the preset mechanism models stores a preset threshold value and a correspondence between a comparison result and a fault identification result; and the comparison result is a comparison result between the actual sensor data and the preset threshold value;

[0130] The performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor includes:

[0131] When the preset mechanism model is the bias voltage model, comparing the actual bias voltage with a preset threshold in the bias voltage model to obtain a first comparison result; and determining a first fault identification result corresponding to the first comparison result based on a correspondence between the comparison result in the bias voltage model and the fault identification result;

[0132] When the preset mechanism model is the output signal model, comparing the actual output signal value with a preset threshold in the output signal model to obtain a second comparison result; and determining a first fault identification result corresponding to the second comparison result based on a correspondence between the comparison result in the output signal model and the fault identification result;

[0133] When the preset mechanism model is the spectrum model, a ski slope factor is determined based on the actual spectrum; the ski slope factor is compared with a preset threshold in the spectrum model to obtain a third comparison result; and based on a correspondence between the comparison result in the spectrum model and the fault identification result, a first fault identification result corresponding to the third comparison result is determined.

[0134] In conjunction with the seventh aspect, in a sixth possible implementation of the seventh aspect, determining the ski slope factor according to the actual spectrum includes:

[0135] Acquire multiple preset frequency intervals corresponding to the target sensor;

[0136] For each of the preset frequency intervals, determining a passing frequency within the preset frequency interval according to the actual frequency spectrum;

[0137] From all said pass frequencies, a ski slope factor is determined.

[0138] In an eighth aspect, the present application further provides a sensor fault identification device, which may include a unit module for executing the method steps in various implementations of the seventh aspect.

[0139] In the ninth aspect, the present application provides an electronic device comprising: a processor and a memory, the processor being connected to the memory, and the processor being used to execute a fault prediction program, a fault monitoring program, a moving equipment fault monitoring program, or a sensor fault identification program stored in the memory, so as to implement the fault prediction method described in the first aspect, the fault monitoring method described in the third aspect, the moving equipment fault monitoring method described in the fifth aspect, or the sensor fault identification method described in the seventh aspect.

[0140] In the tenth aspect, the present application also provides a storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the fault prediction method described in the first aspect, the fault monitoring method described in the third aspect, the dynamic equipment fault monitoring method described in the fifth aspect, or the sensor fault identification method described in the seventh aspect.

[0141] The fault prediction method, device, electronic device, and storage medium provided in the embodiments of the present application solve the problem in related technologies whereby a fault can only be detected when the equipment fault is relatively obvious, resulting in an inability to promptly maintain the faulty equipment. After preprocessing the vibration signal of the target component, a target anomaly judgment model is used to determine whether the collected vibration signal has an anomaly. If an anomaly exists, the target fault prediction model is used to determine the cause of the abnormality in the vibration signal. These two models can then be used to quickly determine whether the target component has a fault and the cause of the fault, thereby improving the speed and efficiency of fault monitoring and promptly maintaining the target equipment based on the cause of the fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0142] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0143] FIG1 is a flow chart of a fault prediction method according to an embodiment of the present application;

[0144] FIG2 is a schematic diagram of a model structure of an optional abnormality judgment model provided according to an embodiment of the present application;

[0145] FIG3 is a flow chart of training an abnormality judgment model according to an embodiment of the present application;

[0146] FIG4 is a training flowchart of a fault prediction model provided according to an embodiment of the present application;

[0147] FIG5 is a schematic diagram of a fault prediction device according to an embodiment of the present application;

[0148] FIG6 is a flowchart of a fault monitoring method provided according to an embodiment of the present application;

[0149] FIG7 is a schematic diagram of an optional reference key phase signal provided according to an embodiment of the present application;

[0150] FIG8 is a schematic diagram of an optional initial high-frequency signal provided according to an embodiment of the present application;

[0151] FIG9 is a schematic diagram of an optional target angle domain vibration signal provided according to an embodiment of the present application;

[0152] FIG10 is a flow chart of a zero-phase filtering method according to an embodiment of the present application;

[0153] FIG11 is a schematic diagram of a fault monitoring device provided according to an embodiment of the present application;

[0154] FIG12 is a schematic structural diagram of a dynamic equipment fault monitoring system provided in an embodiment of the present application;

[0155] FIG13 is a flow chart of an embodiment of a method for monitoring a fault of a moving device provided in an embodiment of the present application;

[0156] FIG14 is a flow chart of another embodiment of a method for monitoring faults of moving equipment provided in an embodiment of the present application;

[0157] FIG15 is a flow chart of another embodiment of a method for monitoring a fault of a moving device provided in an embodiment of the present application;

[0158] FIG16 is a schematic diagram of a threshold configuration data function interface provided in an embodiment of the present application;

[0159] FIG17 is a schematic diagram of a historical alarm log data function interface provided by an embodiment of the present application;

[0160] FIG18 is a flow chart of another embodiment of a method for monitoring a fault of a moving device according to an embodiment of the present application;

[0161] FIG19 is a schematic structural diagram of another dynamic equipment fault monitoring system provided in an embodiment of the present application;

[0162] FIG20 is a block diagram of an embodiment of a moving equipment fault monitoring device provided in an embodiment of the present application;

[0163] FIG21 is a flow chart of a sensor fault identification method provided in an embodiment of the present application;

[0164] FIG22 is a flow chart of another sensor fault identification method provided in an embodiment of the present application;

[0165] FIG23 is a flow chart of another sensor fault identification method provided in an embodiment of the present application;

[0166] FIG24 is a schematic diagram of a sensor short-circuit bias voltage according to an embodiment of the present application;

[0167] FIG25 is a schematic diagram of a short-circuit bias voltage of a sensor provided in an embodiment of the present application;

[0168] FIG26 is a schematic diagram of a ski slope provided in an embodiment of the present application;

[0169] FIG27 is a schematic diagram of the structure of a sensor fault identification device provided in an embodiment of the present application;

[0170] FIG28 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0171] In the embodiments of the present application, in order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of this application.

[0172] The disclosure below provides many different embodiments or examples for implementing different configurations of the present invention. To simplify the disclosure of the present invention, the components and configurations of specific examples are described below. Of course, these are merely examples and are not intended to limit the present invention. In addition, the present invention may repeat reference numerals and / or letters in different examples. Such repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or configurations discussed.

[0173] It should be noted that the fault prediction method determined in the present disclosure can be used in the field of equipment fault diagnosis, and can also be used in any field other than the field of equipment fault diagnosis. The application field of the fault prediction method, device, storage medium and electronic device determined in the present disclosure is not limited.

[0174] For ease of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0175] CNN autoencoder: It is a model that combines CNN (Convolutional Neural Network) with autoencoder, which can be used for tasks such as image denoising, image de-noising, and feature extraction.

[0176] According to an embodiment of the present application, a fault prediction method is provided.

[0177] FIG1 is a flow chart of a fault prediction method according to an embodiment of the present application. As shown in FIG1 , the method includes the following steps:

[0178] Step S102: obtaining a vibration signal collected by a sensor installed on a target component of a target device, and cleaning and segmenting the vibration signal to obtain a target vibration signal.

[0179] Specifically, the target device can be a reciprocating device. When the target device fails, the failed component is relatively fixed, such as a piston, cylinder, etc. At the same time, the vibration signals of some components during operation can provide feedback on whether there is any abnormality in the operation of the device. Therefore, a sensor can be installed on the target component that often fails in the target device, so as to determine whether the target component has a fault based on the vibration signal of the target component collected by the sensor when the target device is running.

[0180] Furthermore, after receiving the vibration signal collected by the sensor on the target component, the sensor may be abnormal, or the operating condition of the target equipment may be unstable, or signal interference may occur, resulting in noise or abnormal signals in the vibration signal. Therefore, the vibration signal needs to be cleaned to make the signal availability and accuracy higher. At the same time, since the vibration signal will be continuously collected as the equipment runs, when analyzing the vibration signal, the collected vibration signal can be segmented and the vibration signal for analysis can be selected, thereby improving the accuracy of fault identification.

[0181] It should be noted that the signal detected by the sensor is a time domain signal. The sensor is installed at the target position close to the monitored target component, including contact installation (such as vibration sensor, etc.) and non-contact installation (such as ultrasonic sensor, etc.). The original measurement data of the corresponding channel is collected through unidirectional or multi-directional sensors, and the relevant channel acquisition parameters are configured through data acquisition software. Time domain measurement data is collected according to the specified acquisition frequency. At the same time, some signal processing methods (such as wavelet transform, empirical mode decomposition, filtering, etc.) are used to remove irrelevant interference signals, so that the relevant information of the monitored parts can be displayed, so that subsequent prediction operations can be performed based on the processed vibration signals.

[0182] For example, to predict a valve assembly failure at the hydraulic end of a five-cylinder plunger pump in fracturing equipment, a vibration sensor is installed at an appropriate location on the hydraulic end to collect time-domain vibration measurement data that can reflect the operating mechanism of the valve assembly. The corresponding position sensor and the collected raw vibration data are labeled AI1-5, and the sensor sampling frequency is 25,600 Hz.

[0183] In step S104, the target vibration signal is input into the target abnormality judgment model to obtain an output result, and whether the target vibration signal is abnormal is determined based on the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is obtained by training the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder.

[0184] Specifically, after processing the vibration signal to obtain the target vibration signal, the target vibration signal needs to be input into the target abnormality judgment model to obtain an output result, wherein the target abnormality judgment model can be a CNN automatic encoder, which can identify the characteristic information of the input vibration signal through the encoder in the target abnormality judgment model, and reconstruct the vibration signal through the decoder, so as to determine whether the target vibration signal is abnormal based on the difference value (also called loss value) between the reconstructed vibration signal and the target vibration signal.

[0185] It should be noted that, since only the vibration signals of the target components of the target equipment under normal operating conditions are used for training when training the target abnormality judgment model, when the vibration signal collected under abnormal conditions is input into the target abnormality judgment model, the target abnormality judgment model cannot accurately reconstruct the vibration signal, resulting in a large difference between the generated reconstructed vibration signal and the vibration signal collected under abnormal conditions. Therefore, the difference value can be used to determine whether the target vibration signal is abnormal.

[0186] Step S106: When the abnormality judgment result indicates that the target vibration signal is abnormal, the target vibration signal is input into the target fault prediction model to obtain a fault prediction result, wherein the target fault prediction model is obtained by combining the encoder of the target abnormality judgment model and the newly added fully connected layer, and the target fault prediction model is obtained by training multiple fault types and the vibration signals of each fault type as samples.

[0187] Specifically, when it is determined that the target vibration signal has an abnormality, due to the large number of abnormalities, the cause of the abnormality of the target equipment cannot be known. Therefore, it is necessary to input the target vibration signal into the target fault prediction model and determine the cause of the failure of the target equipment through the target fault prediction model. Based on the determination that the target equipment has a fault, the cause of the failure can be accurately determined, thereby providing assistance to the operation and maintenance personnel for equipment maintenance.

[0188] It should be noted that since the target anomaly judgment model has been trained, and the target fault prediction model also needs to train the feature extraction part, and since the number of fault signals is limited, it is difficult to train an effective fault classification model based on the fault signals alone. Therefore, based on the existing target anomaly judgment model, a fault classification and recognition model based on transfer learning can be constructed, and the encoder part of the target anomaly judgment model can be used as the feature extraction part of the target fault prediction model, thereby achieving the effect of reducing dependence on new data and saving training time.

[0189] The fault prediction method provided by the embodiment of the present application obtains the vibration signal collected by the sensor installed on the target component of the target device, cleans and segments the vibration signal, and obtains the target vibration signal; inputs the target vibration signal into the target abnormality judgment model to obtain an output result, and determines whether the target vibration signal is abnormal based on the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is obtained by training the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder; when the abnormality judgment result indicates that the target vibration signal is abnormal, the target vibration signal is input into the target fault prediction model to obtain a fault prediction result, wherein the target fault prediction model is obtained by combining the encoder of the target abnormality judgment model and the newly added fully connected layer, and the target fault prediction model is obtained by training multiple fault types and the vibration signal of each fault type as samples. This solves the problem in the related art that the occurrence of a fault can only be monitored when the equipment fault is more obvious, resulting in the inability to maintain the faulty equipment in a timely manner. After preprocessing the vibration signal of the target component, the target abnormality judgment model is used to determine whether the collected vibration signal has any abnormality. If there is an abnormality, the target fault prediction model is used to determine the cause of the abnormal vibration signal. Then, the two models can be used to quickly determine whether the target component has a fault and the cause of the fault, thereby achieving the effect of improving the speed and efficiency of fault monitoring and timely maintaining the target equipment according to the cause of the fault.

[0190] In order to ensure high availability of the signal, optionally, in the fault prediction method provided in the embodiment of the present application, the vibration signal is cleaned and segmented to obtain a target vibration signal, including: segmenting the vibration signal according to the time dimension to obtain multiple first vibration signals, wherein each first vibration signal includes vibration signals of the same duration, and the duration includes multiple continuous vibration cycles; judging in turn whether there is an error signal in each first vibration signal, wherein the error signal characterizes abnormal operation of the sensor; in the case of a first vibration signal with an error signal, deleting the first vibration signal with the error signal from the multiple first vibration signals to obtain multiple second vibration signals; and inputting each second vibration signal into a preset filter in turn to obtain multiple target vibration signals.

[0191] Specifically, after obtaining the vibration signal, the vibration signal is segmented according to the time granularity (each signal segment after segmentation according to the time granularity contains at least one full cycle of working information of the target component, so at least two working cycles of the target component need to be intercepted as the time granularity), thereby obtaining a set of multiple vibration signals after segmentation, wherein the set includes multiple segments of vibration signals.

[0192] Furthermore, it is necessary to determine whether there is an error signal in each vibration signal, where the error signal may be an abnormality in the sensor, or an unstable operating condition of the target device, or signal interference, resulting in noise or abnormal signals in the vibration signal, and if an error signal exists, the error signal should be removed.

[0193] Furthermore, after removing the error signal, it is necessary to infer the frequency range that needs to be focused on and the interference frequency components that need to be filtered out based on the mechanism characteristics of the monitored target component. Each vibration signal in the vibration signal set after removing the error signal uses a signal processing algorithm to remove interference components (including noise and interference such as information transmitted by other components) to obtain a processed signal, so that the characteristic information of the target component is revealed, thereby laying the foundation for subsequent feature extraction operations and improving the accuracy of feature extraction.

[0194] Optionally, in the fault prediction method provided in an embodiment of the present application, the target abnormality judgment model is trained in the following manner: obtaining a sample vibration signal set of the target component, wherein the sample vibration signal set includes multiple sample vibration signals generated by the target component under normal operation; inputting the sample vibration signals in the sample vibration signal set into the initial abnormality judgment model, and processing to obtain an output signal corresponding to each sample vibration signal, wherein the initial abnormality judgment model is composed of an encoder and a decoder, the encoder is used to extract the characteristics of the sample vibration signal and reduce the data dimension, and the decoder is used to restore the data dimension and restore the data characteristics; determining the difference value between each output signal and the corresponding sample vibration signal to obtain multiple difference values, and judging whether there is a difference value greater than a preset threshold among the multiple difference values; when there is a difference value greater than the preset threshold among the multiple difference values, changing the neuron connection weights in the initial abnormality judgment model, and retraining the changed initial abnormality judgment model until there is no difference value greater than the preset threshold among the multiple difference values; when there is no difference value greater than the preset threshold among the multiple difference values, obtaining the target abnormality judgment model.

[0195] It should be noted that when training the abnormality judgment model, the sample vibration signals used in training also need to be labeled, cleaned and segmented after acquisition, and the signals need to be filtered to obtain multiple sample vibration signals, so as to ensure the training effect and accuracy of the model.

[0196] Furthermore, since the target abnormality judgment model needs to determine whether the target vibration signal is abnormal based on the difference between the output result and the input signal, the sample vibration signal used when training the model needs to be the vibration signal generated by the target component under normal operation, and the normal vibration signal is learned through the abnormal judgment model so that the difference between the result obtained after processing the normal vibration signal and the input signal is less than the preset threshold. At this time, after the abnormal vibration signal is input into the target abnormality judgment model, since the model was not trained with the abnormal vibration signal during training, after outputting the result according to the abnormal vibration signal, the difference between the output signal and the input abnormal vibration signal will be much larger than the preset threshold. At this time, whether the input vibration signal is an abnormal vibration signal can be determined based on the quantitative relationship between the difference value and the preset threshold.

[0197] It should be noted that the initial abnormality judgment model is composed of an encoder and a decoder. The encoder is used to extract the characteristics of the sample vibration signal and reduce the data dimension. The decoder is used to restore the data dimension and restore the data characteristics. Figure 2 is a schematic diagram of the model structure of the optional abnormality judgment model provided according to the embodiment of the present application. As shown in Figure 2, the encoder part includes 4 layers of one-dimensional convolution layers and 4 layers of pooling layers. The function of the one-dimensional convolution layer is to extract the local characteristics of the input data, and the pooling layer is used to reduce the dimension of the data and reduce the amount of calculation; the decoder part includes 4 layers of one-dimensional deconvolution layers and 4 layers of depooling layers. The function of the one-dimensional deconvolution layer is to restore the data after the dimensionality reduction of the encoder part to the original dimension, and the depooling layer is used to restore the local characteristics of the data. The ReLu activation function is used between the neurons of this model. The loss function uses the MSE function to calculate the final network loss by inputting data and predicting and reconstructing data. The Adam function is used to optimize the neuron connection weights during the model training process. When the difference between the output result and the input signal is greater than the preset threshold during the training model, the neuron connection weights are adjusted to complete the optimization of the model and then complete the training of the model.

[0198] Figure 3 is a training flowchart of the abnormality judgment model provided according to an embodiment of the present application. As shown in Figure 3, first, a sample vibration signal of the target component under normal operation is obtained, and the sample vibration signal is labeled (the labels are all normal operating values), cleaned and segmented, and interference components are removed to obtain a processed sample vibration signal, and the sample vibration signal is divided into a training set, a verification set, and a test set.

[0199] The CNN autoencoder model is trained using a training set and a validation set, and the sample vibration signal is processed according to the trained model to obtain multiple loss values. It is then determined whether the loss values ​​decrease and are all less than a preset threshold. If they are not less than the preset threshold, the parameters or weights in the model are changed and retrained so that the loss values ​​are on a downward trend and are all less than the preset threshold. At this point, the model training is completed, and the model is tested using a test set. After the test is correct, the trained target anomaly judgment model is obtained.

[0200] The following is a training example of an anomaly judgment model training:

[0201] The experimental data consisted of data from the hydraulic end of a five-cylinder plunger pump in a fracturing system, along with data from AI1-5 sensors at appropriate locations and orientations. Data was collected from multiple key components of each cylinder at a sampling frequency of 25,600 Hz. AI1-5 sensor data was collected from eight units over the past nine days of normal operation. The vibration data collected by each sensor was cleaned and segmented, and interference removed. Because the mechanisms of multiple components within each cylinder are similar, the signals collected by the five sensors were combined to train a single fault prediction model. A total of 27,681 normal data sets were generated, and the samples were split in a 7:2:1 ratio, resulting in 19,376 normal training data sets, 5,537 normal test data sets, and 2,768 normal verification data sets.

[0202] Model training and model testing are performed based on the above training data and test data. The relevant parameters are as follows:

[0203] After training, the network loss of all training sets was used to define the fault alarm threshold, where is the upper quartile of all training set loss values ​​and is the lower quartile of all training set loss values. Ultimately, 99% of the loss values ​​for all training, test, and validation sets fell below . Furthermore, verification using data from multiple equipment failures revealed that network loss showed a trend before a failure occurred, and could be used as a fault signature factor to predict hydraulic end failures in advance.

[0204] Optionally, determining whether the target vibration signal has an abnormality based on the output result, and obtaining the abnormality judgment result includes: determining the difference value between the output result and the target vibration signal, obtaining the target difference value, and judging whether the target difference value is greater than a preset threshold; when the target difference value is greater than the preset threshold, determining that the target vibration signal has an abnormality; when the target difference value is less than or equal to the preset threshold, determining that the target vibration signal does not have an abnormality.

[0205] Specifically, after completing the training of the target abnormality judgment model, when determining whether the target vibration signal is an abnormal signal through the target abnormality judgment model, since multiple signals of the same length are used when training the target abnormality judgment model, it is first necessary to input each target vibration signal into the target abnormality judgment model, generate the output result of each target vibration signal through the target abnormality judgment model, and determine the difference value between each output result and the corresponding input vibration signal. When there is one or several difference values ​​greater than the preset threshold, it indicates that there is an abnormality in this vibration signal, and a subsequent abnormality determination step is required. When the difference values ​​of each vibration signal are less than the preset threshold, it indicates that there is no abnormality in the target vibration signal.

[0206] Optionally, in the fault prediction method provided in an embodiment of the present application, the target fault prediction model is trained in the following manner: obtaining multiple fault types and historical vibration signals under each fault type to obtain multiple groups of historical vibration signals; adding a label to each historical vibration signal in each group of historical vibration signals according to the fault type to obtain multiple groups of updated historical vibration signals; using the multiple groups of updated historical vibration signals as samples to train the initial fault prediction model to obtain the target fault prediction model.

[0207] Specifically, when training the target fault prediction model, since the cause of the fault needs to be determined through the model, when obtaining sample data, the sample data needs to include historical vibration signals under each fault type. These historical vibration signals are all vibration signals generated by the target component when the fault occurs. Similarly, the sample data also requires vibration signals of the target component during normal operation for model training.

[0208] When training the model, it is necessary to add a label to each historical vibration signal. The label is used to determine the fault type corresponding to each vibration signal. The initial fault prediction model is trained using the labeled vibration signal to obtain the target fault prediction model.

[0209] It should be noted that due to the small number of failures, some types of failures may only occur once. Therefore, due to the small number of vibration signals under the fault, there are fewer samples when training the model, and the model cannot be trained accurately. At this time, the feature recognition part of the abnormality judgment model that has been trained can be used as the feature recognition part of the fault prediction model, and the fault prediction model can be trained on this basis to improve the recognition accuracy of the feature recognition part, thereby improving the accuracy of the fault prediction model.

[0210] Figure 4 is a training flowchart of the fault prediction model provided according to an embodiment of the present application. As shown in Figure 4, first, the historical vibration signals of the target component under normal operation and abnormal operation are obtained, and the historical vibration signals are labeled, cleaned and segmented, and the interference components are removed to obtain the processed historical vibration signals, and the historical vibration signals are divided into a training set, a verification set and a test set.

[0211] The encoder part of the target anomaly judgment model is obtained as the feature extraction part of the fault prediction model. On this basis, two one-dimensional convolution layers and two pooling layers are added to continue dimensionality reduction. A fully connected layer is added at the end of the network to obtain the fault prediction model. The fault prediction model is trained using the training set and the validation set. After the training is completed, the fault prediction model is tested using the test set to complete the training of the fault prediction model.

[0212] The following is a training case for a fault prediction model:

[0213] For example, the encoder part of the trained anomaly judgment model has learned the useful features of the input data. Therefore, the network structure and parameters of the encoder part of the anomaly judgment model are retained. On this basis, two layers of one-dimensional convolution layers and two layers of pooling layers are added to continue dimensionality reduction, and a fully connected layer is added at the end of the network to realize fault classification.

[0214] The ReLu activation function is used between neurons in this model, and the loss function uses the categorical_crossentropy function to input data labels and predict data labels to calculate the final classification accuracy. The Adam function is also used to optimize the neuron connection weights during the model training process.

[0215] It's important to note that when training a new network, the migrated encoding layers have pre-trained network parameters, while the newly added network structure has initialized parameters. Furthermore, due to the addition of fault data from different components, the encoder portion of the original network structure is not yet capable of extracting features from the fault data. Therefore, distributed training is required when training the new network. First, all migrated network layers are frozen, and the parameters of all newly added layers are trained to adapt to the new classification task. Then, the parameters of each migrated layer are unfrozen layer by layer, and fine-tuned to enhance the migrated network's ability to extract features from the fault data. Training concludes when the validation set recognition accuracy no longer significantly improves.

[0216] In order to determine whether the result obtained by the target abnormality judgment model is accurate, optionally, in the fault prediction method provided in the embodiment of the present application, after obtaining the fault prediction result, the method also includes: obtaining a historical abnormal vibration signal collected by the sensor when an abnormality occurs in the target component; determining the characteristic information of the vibration signal of the target component based on the historical abnormal vibration signal to obtain multiple characteristic information, and obtaining the characteristic value corresponding to each characteristic information; determining the judgment standard of each characteristic information based on the characteristic value corresponding to each characteristic information to obtain multiple judgment standards; obtaining the vibration data value corresponding to each characteristic information in the target vibration signal, and using the same judgment standard of the characteristic information to perform abnormality judgment on the vibration data value to obtain a judgment result; when the judgment result indicates that there is no abnormality in the target vibration signal, determining that there is an abnormality in the target abnormality judgment model, and issuing an alarm message, wherein the alarm information indicates that there is an abnormality in the target abnormality judgment model.

[0217] Specifically, after obtaining the fault prediction result, there may be a problem of inaccurate model prediction, that is, the target component does not have abnormal operation, but the result of the model output determines that the target component has a fault. At this time, it is necessary to first determine the characteristic information in the historical vibration signal, where the characteristic information may include: time domain characteristics (such as kurtosis, peak value, effective value, etc.), frequency domain characteristics (center of gravity frequency, mean square frequency, etc.) and other related characteristics (such as permutation entropy, spread entropy, etc.), and determine the characteristic value of each characteristic information and the judgment standard. For example, the characteristic value of the peak value of the characteristic information can be 100, and the judgment standard is that the peak value is less than 100 for normal and the peak value is greater than 100 for abnormal. Then, it can be determined whether the target vibration signal is abnormal based on the above judgment standard.

[0218] Furthermore, after judging the target vibration signal through each of the above-mentioned judgment criteria, it is possible to determine whether the target vibration signal is abnormal based on the judgment result. If the judgment result determines that the target vibration model is not abnormal, it indicates that there is an abnormality in the target abnormality judgment model, and the target abnormality judgment model needs to be adjusted to achieve the effect of monitoring the model.

[0219] In order to determine whether the target vibration signal has an abnormality, optionally, in the fault prediction method provided in the embodiment of the present application, the vibration data value is judged to be abnormal using the same judgment criteria of the characteristic information, and the judgment result obtained includes: obtaining the vibration data value under each characteristic information and the judgment criteria under the same characteristic information, and determining whether the vibration data value meets the corresponding judgment criteria, and obtaining multiple sub-judgment results, wherein the sub-judgment results are used to characterize whether the vibration data value meets the corresponding judgment criteria; when there is a sub-judgment result that does not meet the judgment criteria among the multiple sub-judgment results, it is determined that the judgment result is that the target vibration signal has an abnormality; when there is no sub-judgment result that does not meet the judgment criteria among the multiple sub-judgment results, it is determined that the judgment result is that the target vibration signal does not have an abnormality.

[0220] Specifically, when generating a judgment result, it is necessary to obtain the sub-judgment result of each judgment criterion. When all the sub-judgment results represent that the vibration data value meets the corresponding judgment criterion, it is determined that there is no abnormality in the target vibration signal. When any sub-judgment result represents that the vibration data value does not meet the corresponding judgment criterion, it is determined that there is an abnormality in the target vibration signal, thereby achieving the effect of accurately determining whether there is an abnormality in the target vibration signal.

[0221] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0222] The present application also provides a fault prediction device. It should be noted that the fault prediction device of the present application can be used to execute the fault prediction method provided in the present application. The fault prediction device provided in the present application is introduced below.

[0223] FIG5 is a schematic diagram of a fault prediction device according to an embodiment of the present application. As shown in FIG5 , the device includes: a first acquisition unit 51 , a first determination unit 52 , and a prediction unit 53 .

[0224] The first acquisition unit 51 is configured to acquire a vibration signal collected by a sensor installed on a target component of a target device, and clean and segment the vibration signal to obtain a target vibration signal.

[0225] The first determination unit 52 is used to input the target vibration signal into the target abnormality judgment model to obtain an output result, and determine whether the target vibration signal is abnormal based on the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is obtained by training the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder.

[0226] The prediction unit 53 is used to input the target vibration signal into the target fault prediction model to obtain a fault prediction result when the abnormality judgment result indicates that the target vibration signal is abnormal. The target fault prediction model is obtained by combining the encoder of the target abnormality judgment model and the newly added fully connected layer. The target fault prediction model is obtained by training multiple fault types and the vibration signals of each fault type as samples.

[0227] The fault prediction device provided in the embodiment of the present application obtains the vibration signal collected by the sensor installed on the target component of the target equipment through the first acquisition unit 51, and cleans and segments the vibration signal to obtain the target vibration signal; the first determination unit 52 inputs the target vibration signal into the target abnormality judgment model to obtain an output result, and determines whether the target vibration signal is abnormal based on the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is trained by the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder; the prediction unit 53 inputs the target vibration signal into the target fault prediction model when the abnormality judgment result indicates that the target vibration signal is abnormal, and obtains a fault prediction result, wherein the target fault prediction model is obtained by combining the encoder of the target abnormality judgment model and the newly added fully connected layer, and the target fault prediction model is trained by using multiple fault types and the vibration signal of each fault type as samples. This solves the problem in the related art that the occurrence of a fault can only be detected when the equipment fault is relatively obvious, resulting in the inability to maintain the faulty equipment in a timely manner. After preprocessing the vibration signal of the target component, the target abnormality judgment model is used to determine whether the collected vibration signal has any abnormality. If there is an abnormality, the target fault prediction model is used to determine the cause of the abnormal vibration signal. Then, the two models can be used to quickly determine whether the target component has a fault and the cause of the fault, thereby achieving the effect of improving the speed and efficiency of fault monitoring and timely maintaining the target equipment according to the cause of the fault.

[0228] Optionally, in the fault prediction device provided in the embodiment of the present application, the first acquisition unit 51 includes: a segmentation module, which is used to segment the vibration signal according to the time dimension to obtain multiple first vibration signals, wherein each first vibration signal includes vibration signals of the same duration, and the duration includes multiple continuous vibration cycles; a judgment module, which is used to judge in turn whether there is an error signal in each first vibration signal, wherein the error signal represents abnormal operation of the sensor; a deletion module, which is used to delete the first vibration signal with an error signal from the multiple first vibration signals in the case of a first vibration signal with an error signal, to obtain multiple second vibration signals; an input module, which is used to input each second vibration signal into a preset filter in turn to obtain multiple target vibration signals.

[0229] Optionally, in the fault prediction device provided in an embodiment of the present application, the target abnormality judgment model is trained in the following manner: a second acquisition unit is used to acquire a sample vibration signal set of the target component, wherein the sample vibration signal set includes multiple sample vibration signals generated by the target component under normal operation; an input unit is used to input the sample vibration signals in the sample vibration signal set into the initial abnormality judgment model, and process them to obtain an output signal corresponding to each sample vibration signal, wherein the initial abnormality judgment model is composed of an encoder and a decoder, the encoder is used to extract features of the sample vibration signals and reduce data dimensions, and the decoder is used to restore the data dimensions and restore the data features; a second determination unit is used to determine the difference value between each output signal and the corresponding sample vibration signal, obtain multiple difference values, and determine whether there is a difference value greater than a preset threshold in the multiple difference values; a change unit is used to change the neuron connection weights in the initial abnormality judgment model when there is a difference value greater than the preset threshold in the multiple difference values, and re-train the changed initial abnormality judgment model until there is no difference value greater than the preset threshold in the multiple difference values; a third acquisition unit is used to obtain the target abnormality judgment model when there is no difference value greater than the preset threshold in the multiple difference values.

[0230] Optionally, in the fault prediction device provided in the embodiment of the present application, the first determination unit 52 includes: a first determination module, used to determine the difference value between the output result and the target vibration signal, obtain the target difference value, and judge whether the target difference value is greater than a preset threshold; a second determination module, used to determine that there is an abnormality in the target vibration signal when the target difference value is greater than the preset threshold; a third determination module, used to determine that there is no abnormality in the target vibration signal when the target difference value is less than or equal to the preset threshold.

[0231] Optionally, in the fault prediction device provided in the embodiment of the present application, the target fault prediction model is trained in the following manner: a fourth acquisition unit is used to acquire multiple fault types and historical vibration signals under each fault type to obtain multiple groups of historical vibration signals; an adding unit is used to add a label to each historical vibration signal in each group of historical vibration signals according to the fault type to obtain multiple groups of updated historical vibration signals; a training unit is used to train the initial fault prediction model using the multiple groups of updated historical vibration signals as samples to obtain the target fault prediction model.

[0232] Optionally, in the fault prediction device provided in the embodiment of the present application, the device also includes: a fifth acquisition unit, used to acquire historical abnormal vibration signals collected by the sensor when an abnormality occurs in the target component; a third determination unit, used to determine characteristic information of the vibration signal of the target component based on the historical abnormal vibration signal, obtain multiple characteristic information, and obtain characteristic values ​​corresponding to each characteristic information; a fourth determination unit, used to determine the judgment standard of each characteristic information based on the characteristic value corresponding to each characteristic information, and obtain multiple judgment standards; a judgment unit, used to obtain the vibration data value corresponding to each characteristic information in the target vibration signal, and use the same judgment standard of the characteristic information to perform abnormal judgment on the vibration data value to obtain a judgment result; an alarm unit, used to determine that there is an abnormality in the target abnormality judgment model when the judgment result indicates that there is no abnormality in the target vibration signal, and to issue an alarm message, wherein the alarm information indicates that there is an abnormality in the target abnormality judgment model.

[0233] Optionally, in the fault prediction device provided in the embodiment of the present application, the judgment unit includes: an acquisition module, used to obtain the vibration data value under each characteristic information and the judgment criteria under the same characteristic information, and determine whether the vibration data value meets the corresponding judgment criteria to obtain multiple sub-judgment results, wherein the sub-judgment results are used to characterize whether the vibration data value meets the corresponding judgment criteria; a fourth determination module, used to determine that the judgment result is that there is an abnormality in the target vibration signal when there is a sub-judgment result that does not meet the judgment criteria among the multiple sub-judgment results; a fifth determination module, used to determine that the judgment result is that there is no abnormality in the target vibration signal when there is no sub-judgment result that does not meet the judgment criteria among the multiple sub-judgment results.

[0234] The above-mentioned fault prediction device includes a processor and a memory. The above-mentioned first acquisition unit 51, first determination unit 52, prediction unit 53, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0235] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured. By adjusting kernel parameters, this solves the problem in related technologies where a device failure can only be detected after it is obvious, resulting in a delay in timely maintenance of the faulty device.

[0236] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0237] Reciprocating equipment plays a vital role in various industrial sectors, particularly in the energy, chemical, and manufacturing sectors. For example, reciprocating compressors are widely used in oil and natural gas refining, chemical production, and air and other gas compression. To ensure continuous and efficient production processes, the efficiency and reliability of reciprocating equipment are crucial.

[0238] It should be noted that various components of reciprocating equipment, such as pistons, cylinders, valves, packings, and valve bodies, are subject to failure due to long-term operation and wear. For example, if the piston rings of a reciprocating compressor are excessively worn, it may lead to reduced compression efficiency and even cause the compressor to fail.

[0239] Therefore, it is essential to monitor the health and integrity of reciprocating equipment. Regular maintenance and inspections can help identify and correct equipment failures, thus preventing damage and interruptions to production processes. This is crucial for ensuring the efficient operation of the energy and other industries.

[0240] Currently, monitoring the operating status of reciprocating equipment typically requires detecting equipment failures only after they occur and significantly impact equipment operation. Experienced technicians can then determine whether a fault has occurred by monitoring the equipment's sound and appearance at the construction site, and collaborate with maintenance workers to conduct regular inspections during operations to ensure the equipment's proper functioning. However, the accuracy of manual identification of reciprocating equipment relies heavily on the technician's experience, which can easily lead to misjudgments and missed detections. Furthermore, faults can only be detected after they have a significant impact, making it difficult to detect and address them promptly and accurately.

[0241] Currently, no effective solution has been proposed to address the issues of low timeliness and accuracy in manually determining whether a device is faulty in related technologies.

[0242] In this regard, the present application also provides a fault monitoring method to solve the problem of low timeliness and accuracy in manually determining whether a device has a fault in the related art.

[0243] For ease of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0244] Key phase signal: A groove or convex key is set on the measured shaft to form a key phase mark. When this mark rotates to the probe position, it is equivalent to a sudden change in the distance between the probe and the measured surface. The sensor will generate a pulse signal. The signal generated by the shaft rotating multiple circles is the key phase signal.

[0245] According to an embodiment of the present application, a fault monitoring method is provided.

[0246] FIG6 is a flow chart of a fault monitoring method according to an embodiment of the present application. As shown in FIG6 , the method includes the following steps:

[0247] Step S602 : acquiring an initial vibration signal collected by a preset sensor, wherein the preset sensor is disposed at a target position of a target device, and the preset sensor is used to collect the vibration signal at the target position.

[0248] Specifically, the target device can be a reciprocating device. Since the location of the target device is relatively fixed when it fails, such as a piston, cylinder, etc., and the vibration signals of some components during operation can provide feedback on whether there is any abnormality in the operation of the device, a sensor can be installed at the target position of the target device to determine whether the target device has a fault based on the vibration signal of the target position of the target device collected by the sensor during operation.

[0249] Step S604: Filter the initial vibration signal by zero-phase filtering to obtain an initial high-frequency signal.

[0250] Specifically, after receiving the vibration signal collected by the sensor at the target position, since the received vibration signal may be mixed with noise, which affects the analysis of the vibration signal, it is necessary to eliminate or weaken the noise in the vibration signal to make the signal availability and accuracy higher.

[0251] It should be noted that when filtering vibration signals, an infinite impulse response filter is usually used, and a filter with corresponding functions is selected according to the filtering requirements, such as a low-pass filter, a high-pass filter, a band-pass filter, and a band-stop filter. However, after the infinite impulse response filter processes the signal, the obtained signal will have a delay and there will be a slight change in the signal waveform. Therefore, when filtering, it is necessary to use a zero-phase filtering method to filter the initial vibration signal to obtain an initial high-frequency signal, thereby ensuring that the obtained high-frequency signal will not have delays and waveform changes.

[0252] Step S606, obtain the reference key phase signal of the reference device, and determine the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and change the horizontal coordinate of the initial high-frequency signal from the time value to the corresponding angle value according to the mapping relationship to obtain the target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device.

[0253] Specifically, after obtaining the initial high-frequency signal, since the initial high-frequency signal is a relationship curve between the acquisition time and the amplitude, when analyzing the operating status of the target device, it is necessary to convert it into a relationship curve between the rotation angle and the amplitude. Therefore, it is necessary to determine the mapping relationship between the rotation angle of the target device and the acquisition time, so as to change the horizontal coordinate of the initial high-frequency signal from the time value to the corresponding angle value according to the mapping relationship to obtain the target angular domain vibration signal.

[0254] Furthermore, when determining the mapping relationship between the target device's rotation angle and the acquisition time, since the relationship between the target device's number of rotations and the rotation time is unknown, the relationship between time and rotation angle can be determined by obtaining a reference key phase signal from a reference device and, based on the reference device's vibration signal during operation. Since the reference device and the target device have the same rotational speed, the time required to rotate the same angle is the same. After obtaining the relationship between the reference device's number of rotations and the rotation time, this relationship can be determined as the mapping relationship between the target device's number of rotations and the rotation time, thereby completing the conversion of the horizontal coordinate of the initial angular domain vibration signal.

[0255] Figure 7 is a schematic diagram of an optional reference key phase signal provided according to an embodiment of the present application. As shown in Figure 7, the collection time of each amplitude collection point can be seen from the figure, and since the key phase signal is an amplitude collection when the reference component rotates to a certain position, the rotation angle between adjacent amplitudes is 360 degrees. Therefore, the collection time corresponding to each degree can be determined according to Figure 2, thereby obtaining a mapping relationship between time and rotation angle.

[0256] Figure 8 is a schematic diagram of an optional initial high-frequency signal provided according to an embodiment of the present application, and Figure 9 is a schematic diagram of an optional target angular domain vibration signal provided according to an embodiment of the present application. As shown in Figures 8 and 9, Figure 8 is a schematic diagram of the initial high-frequency signal. At this time, the signal is a time domain signal, that is, a relationship curve between amplitude and time. After conversion according to the mapping relationship between time and rotation angle, the X-axis is converted from the acquisition time to the rotation angle, thereby obtaining the target angular domain vibration signal in Figure 9.

[0257] Step S608, obtain standard data of the target device, detect the target angular vibration signal through the standard data, obtain a detection result, and determine the fault state of the target device based on the detection result, wherein the standard data is the data in the vibration signal of the target device during normal operation.

[0258] Specifically, after obtaining the target angular domain vibration signal, the standard data of the target device can be obtained, wherein the standard data can be the amplitude value corresponding to the important angle in the target device. For example, the standard data can be: the amplitude at the 180-degree position cannot exceed 50. Then, the standard data can be compared with the amplitude of 180 degrees in the target angular domain vibration signal to determine whether the amplitude in the target angular domain vibration signal is a normal value. Then, based on whether there is an abnormal signal in the target angular domain vibration signal, and in the case of an abnormal signal, it can be determined that the operating condition of the target device is abnormal and maintenance is required.

[0259] The fault monitoring method provided by the embodiment of the present application obtains an initial vibration signal collected by a preset sensor, wherein the preset sensor is set at a target position of a target device and is used to collect the vibration signal of the target position; the initial vibration signal is filtered by a zero-phase filtering method to obtain an initial high-frequency signal; a reference key phase signal of a reference device is obtained, and a mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device is determined according to the rotation speed of the reference device, and the horizontal coordinate of the initial high-frequency signal is changed from a time value to a corresponding angle value according to the mapping relationship to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device; standard data of the target device is obtained, the target angular domain vibration signal is detected by the standard data to obtain a detection result, and the fault state of the target device is determined according to the detection result, wherein the standard data is the data in the vibration signal of the target device during normal operation. The method solves the problem of low timeliness and accuracy of manual judgment of whether a device has a fault in the related art. The collected signal is filtered by a zero-phase filtering method, and the time value of the vibration signal is changed to an angle value according to the mapping relationship between angle and time, so that the vibration signal can be determined whether it is abnormal based on the vibration data corresponding to the rotation angle, thereby achieving the effect of improving the speed and accuracy of fault monitoring.

[0260] Optionally, in the fault monitoring method provided in the embodiment of the present application, the initial vibration signal is filtered by zero-phase filtering to obtain an initial high-frequency signal, including: obtaining a preset high-pass filter, and inputting the initial vibration signal into the preset high-pass filter to obtain a first time domain signal; flipping the first time domain signal with the center line of the X-axis as the center, and inputting the flipped first time domain signal into the preset high-pass filter to obtain a second time domain signal; flipping the second time domain signal with the center line of the X-axis as the center to obtain the initial high-frequency signal.

[0261] Specifically, in order to remove noise from the vibration signal while ensuring the accuracy of the processed signal, we can first use the bilinear transformation method to design infinite impulse response filters, such as low-pass filters, high-pass filters, band-pass filters and band-stop filters, based on the vibration signal and noise characteristics of the reciprocating equipment.

[0262] For example: assuming that the useful frequency band of reciprocating equipment is mostly concentrated above 1000Hz, and the noise signal is mostly generated by the gear meshing frequency (<1000Hz) and its harmonics, and low-frequency (<500Hz) noise interference, the bilinear transformation method is used to design an infinite impulse response high-pass filter to filter out low-pass interference.

[0263] Furthermore, since the infinite impulse response high-pass filter will produce a delayed signal after processing the signal, and there will be slight changes in the signal waveform, it is necessary to use zero-phase filtering to filter the initial vibration signal during filtering to obtain the initial high-frequency signal, thereby ensuring that the obtained high-frequency signal will not be delayed or have waveform changes.

[0264] FIG10 is a flow chart of a zero-phase filtering method provided according to an embodiment of the present application. As shown in FIG10 , the initial vibration signal x(z) is first input into an infinite impulse response high-pass filter H(z), and then the obtained time domain signal is flipped to obtain a signal xH(z); the obtained signal xH(z) is passed through the filter H(z) again to obtain a time domain signal xHH(z); and finally, xHH(z) is flipped to obtain a zero-phase filtered signal Y(z), i.e., the initial high-frequency signal. This ensures the original waveform of the signal and reduces phase distortion.

[0265] It should be noted that when flipping the signal, the flipping method is: obtain the center position of the signal on the X-axis, draw a perpendicular line to the X-axis, obtain the center line, and perform a mirror flip along the center line to complete the flipping operation of the signal.

[0266] Optionally, in the fault monitoring method provided in the embodiment of the present application, after obtaining the reference key phase signal of the reference device, the method further includes: determining the rotation period of the reference device according to the rotation speed of the reference device, and intercepting the signal within a preset time period in the reference key phase signal to obtain an updated reference key phase signal, wherein the preset time period includes multiple continuous rotation periods; intercepting the signal within a preset time period in the initial high-frequency signal to obtain an updated initial high-frequency signal; determining the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and changing the horizontal coordinate of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain the target angular domain vibration signal, including: determining the mapping relationship between the time in the updated reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and converting the updated initial high-frequency signal according to the mapping relationship to obtain the target angular domain vibration signal.

[0267] Specifically, when performing signal analysis, in order to increase the speed of signal analysis, the rotation period of the reference device can be determined based on the rotation speed of the reference device. For example, if the rotation speed is one rotation per second, the rotation period is 1s, and the signal within a preset time period is intercepted in the reference key phase signal based on 1s, where the preset time period can be multiple rotation cycles, such as intercepting a signal within 1 minute.

[0268] Furthermore, to modify the X-axis value of the initial high-frequency signal based on the relationship between speed and time, it is also necessary to intercept the high-frequency vibration signal within a preset time period to obtain an updated high-frequency vibration signal. It should be noted that the intercepted high-frequency vibration signal and the intercepted reference key phase signal must be captured over the same time period, with the same start and end times. After obtaining the intercepted high-frequency vibration signal, the angle corresponding to each acquisition moment can be determined based on the mapping relationship, thereby converting the time axis of the high-frequency vibration signal into an angle axis to obtain the target angular domain vibration signal.

[0269] Optionally, in the fault monitoring method provided in the embodiment of the present application, the reference device is provided with a code disk, and the rotation speed of the reference device is the same as the rotation speed of the code disk. Determining the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device includes: determining the number of rotations of the code disk within a preset time period according to the rotation speed to obtain the preset number of rotations; determining the rotation angle corresponding to the preset number of rotations, and determining the time required for each rotation of 1 degree according to the rotation angle and the preset time period to obtain a mapping relationship.

[0270] Specifically, when determining the mapping relationship between time and rotation angle, the code disk can be used to determine the length of time required for the reference device to rotate one circle, thereby determining the length of time required for the reference device to rotate 1 degree, and thus the degree of rotation of the reference device corresponding to each acquisition moment can be determined based on the time required to rotate 1 degree.

[0271] For example, if the target device rotates 1 degree per second, it will make one rotation in 360 seconds. When the acquisition frequency is 2 seconds, the X-axis of the reference device changes from 1s, 2s, ... to 2°, 4°, ...

[0272] Optionally, in the fault monitoring method provided in the embodiment of the present application, the horizontal coordinate of the initial high-frequency signal is changed from a time value to a corresponding angle value according to a mapping relationship, and the target angular domain vibration signal is obtained, including: converting the moment value in the time axis of the initial high-frequency signal into a preset angle value according to the mapping relationship to obtain the initial angular domain vibration signal; judging whether each preset angle value has a corresponding amplitude value in the initial angular domain vibration signal, and when the target preset angle value does not have a corresponding amplitude value, obtaining multiple preset angle values ​​adjacent to the target preset angle value and the amplitude value of each preset angle value, using the interpolation method to calculate the amplitude value of the target preset angle value according to the amplitude value of each preset angle value, and adding the amplitude to the corresponding position in the initial angular domain vibration signal to obtain the target angular domain vibration signal.

[0273] It should be noted that when analyzing the target angular vibration signal, it is necessary to obtain the amplitude values ​​at certain specific angles. However, due to the setting of the acquisition frequency, the amplitude value may not be collected at the moment corresponding to a certain angle, resulting in the inability to obtain the amplitude value at the specific angle, and further making it impossible to determine whether the target device has an abnormality based on the amplitude value.

[0274] Specifically, after converting the time value into an angle value according to the mapping relationship, if it is determined that a certain specific angle does not have a corresponding amplitude value, the amplitude values ​​of multiple preset angle values ​​adjacent to the specific angle can be used, and the amplitude value at the specific angle can be calculated according to the amplitude values ​​of the adjacent multiple preset angle values ​​according to the interpolation algorithm, and the amplitude value can be added to the initial angular domain vibration signal to obtain a target angular domain vibration signal with a complete amplitude value.

[0275] For example, take the high-frequency data Hi{hi0, hi1, hi2, ..., hin} and the key phase signal Ki{ki1, ki2, ki3, ..., kin} within the same time period [t1, t2] (two cycles can be taken). First, it is necessary to interpolate the key phase signal to obtain the angular domain data within two cycles, that is, convert the horizontal coordinate of the key phase signal from the original time to the angle, and calculate the time corresponding to each angle. For example, if the angle interval after conversion is 1°, it is necessary to obtain the time TAi{ta1, ta2, ta3, ..., tan} and the vertical coordinate Kai{ka1, ka2, ka3, ..., kan} corresponding to 1°, 2°, 3°...720°. Both TAi and Kai need to use the interpolation function to calculate the corresponding values. The high-frequency data Hi uses the time information Tai calculated previously and is converted into the angular domain signal Ph{ph1, ph2, ph3, ..., phn} through the same interpolation calculation steps, thereby obtaining the target angular domain vibration signal.

[0276] Optionally, in the fault monitoring method provided in the embodiment of the present application, the target angular domain vibration signal is detected through standard data to obtain a detection result, and the fault state of the target device is determined based on the detection result, including: obtaining a calibration angle and a standard amplitude at the calibration angle from the standard data; determining the amplitude at the calibration angle in the target angular domain vibration signal to obtain a target amplitude, and determining whether the target amplitude is greater than the standard amplitude; when the target amplitude is greater than the standard amplitude, determining that there is a fault in the target device; when the target amplitude is less than or equal to the standard amplitude, determining that there is no fault in the target device.

[0277] Specifically, when detecting the vibration signal in the target angular domain, it is first necessary to obtain standard data, where the standard data can be the amplitude value corresponding to the important angle of the target device. For example, the standard data can be: the amplitude at the 180-degree position cannot exceed 50. After obtaining the standard data, multiple verification angles can be obtained from the standard data, where the amplitude value at each verification angle needs to meet the corresponding verification requirements.

[0278] Furthermore, after obtaining multiple calibration angles and the standard amplitude at each calibration angle, the amplitude at each calibration angle can be obtained from the target angle domain vibration signal to obtain multiple target amplitudes, and the target amplitude and the standard amplitude at the same calibration angle can be compared. When the target amplitude is greater than the standard amplitude, it can be indicated that there is a fault in the target device, and the fault condition of the target device can be determined based on the amplitude.

[0279] Optionally, in the fault monitoring method provided in the embodiment of the present application, the amplitude at the verification angle is determined in the target angular domain vibration signal, and obtaining the target amplitude includes: obtaining the angle range in the target angular domain vibration signal, and obtaining the number of rotation cycles contained in the angle range to obtain the number of cycles, wherein the angle range is greater than 360 degrees; calculating the angle value corresponding to the verification angle in each cycle to obtain multiple preset angles; determining the target angle corresponding to each preset angle in the target angular domain vibration signal to obtain multiple target angles, and obtaining the amplitude at each target angle to obtain multiple target amplitudes.

[0280] Specifically, since the calibration angle in the standard data is within the range of 0-360 degrees, as shown in FIG4 , the angle range of the target angular domain vibration signal obtained will be greater than 360 degrees. At this time, it is necessary to determine the rotation period corresponding to the angle range. For example, if the angle range of the target angular domain vibration signal is 0 degrees-720 degrees, the number of rotation periods to be calculated can be 720 / 360=2, i.e., two rotation periods. Based on the number of rotation periods, the angle value corresponding to the calibration angle in each period is calculated. For example, the calibration angle is 180° in the first period and 540° in the second period. Thus, the amplitude value at 540° can be obtained in the target angular domain vibration signal, and the amplitude value at 540° is compared with the standard amplitude at the calibration angle to obtain a comparison result. Thus, the amplitude value at each target angle in the entire target angular domain vibration signal is compared to determine whether the vibration signal is abnormal, and then determine whether the target device is abnormal based on the abnormality determination result.

[0281] For example, when the calibration angle is 100 degrees, the standard amplitude is 50, and the target amplitude is 100. At this time, the target amplitude is greater than the standard amplitude, and the calibration angle corresponds to a lower body. This indicates that there is an abnormality in the lower body in the target device.

[0282] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0283] The present application also provides a fault monitoring device. It should be noted that the fault monitoring device of the present application can be used to execute the fault monitoring method provided in the present application. The following describes the fault monitoring device provided in the present application.

[0284] FIG11 is a schematic diagram of a fault monitoring device according to an embodiment of the present application. As shown in FIG11 , the device includes: an acquisition unit 111 , a filtering unit 112 , a changing unit 113 , and a first determining unit 114 .

[0285] The acquisition unit 111 is configured to acquire an initial vibration signal collected by a preset sensor, wherein the preset sensor is disposed at a target position of a target device, and the preset sensor is configured to collect the vibration signal at the target position.

[0286] The filtering unit 112 is configured to filter the initial vibration signal by using a zero-phase filtering method to obtain an initial high-frequency signal.

[0287] The changing unit 113 is used to obtain a reference key phase signal of a reference device, and determine a mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and change the horizontal coordinate of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device.

[0288] The first determination unit 114 is used to obtain standard data of the target device, detect the target angular vibration signal through the standard data, obtain a detection result, and determine the fault state of the target device according to the detection result, wherein the standard data is data in the vibration signal of the target device during normal operation.

[0289] The fault monitoring device provided in the embodiment of the present application is configured to obtain an initial vibration signal collected by a preset sensor through an acquisition unit 111, wherein the preset sensor is set at a target position of a target device, and the preset sensor is used to collect the vibration signal at the target position; a filtering unit 112 is configured to filter the initial vibration signal by a zero-phase filtering method to obtain an initial high-frequency signal; a changing unit 113 is configured to obtain a reference key phase signal of a reference device, and determine a mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and change the horizontal coordinate of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is proportional to the rotation angle of the target device. The rotation speed of the target device is the same; the first determination unit 114 is used to obtain standard data of the target device, detect the target angular domain vibration signal through the standard data, obtain a detection result, and determine the fault state of the target device according to the detection result, wherein the standard data is the data in the vibration signal of the target device during normal operation, which solves the problem of low timeliness and accuracy of manual judgment of whether there is a fault in the equipment in the related art, and filters the collected signal by zero-phase filtering, and changes the time value of the vibration signal to the angle value according to the mapping relationship between angle and time, so that it can be determined whether the vibration signal is abnormal based on the vibration data corresponding to the rotation angle, thereby achieving the effect of improving the fault monitoring speed and accuracy.

[0290] Optionally, in the fault monitoring device provided in the embodiment of the present application, the filtering unit 112 includes: a first acquisition module, used to acquire a preset high-pass filter, and input the initial vibration signal into the preset high-pass filter to obtain a first time domain signal; a first flipping module, used to flip the first time domain signal with the center line of the X-axis as the center, and input the flipped first time domain signal into the preset high-pass filter to obtain a second time domain signal; a second flipping module, used to flip the second time domain signal with the center line of the X-axis as the center to obtain an initial high-frequency signal.

[0291] Optionally, in the fault monitoring device provided in the embodiment of the present application, the device also includes: a second determination unit, used to determine the rotation period of the reference device according to the rotation speed of the reference device, and intercept the signal within a preset time period in the reference key phase signal to obtain an updated reference key phase signal, wherein the preset time period includes multiple continuous rotation cycles; an interception unit, used to intercept the signal within a preset time period in the initial high-frequency signal to obtain an updated initial high-frequency signal; the change unit 113 includes: a conversion module, used to determine the mapping relationship between the time in the updated reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and convert the updated initial high-frequency signal according to the mapping relationship to obtain a target angular domain vibration signal.

[0292] Optionally, in the fault monitoring device provided in the embodiment of the present application, the reference device is provided with a code disk, and the rotational speed of the reference device is the same as the rotational speed of the code disk. The changing unit 113 includes: a first determination module, used to determine the number of revolutions of the code disk within a preset time period according to the rotational speed, and obtain the preset number of revolutions; a second determination module, used to determine the rotation angle corresponding to the preset number of revolutions, and determine the time required for each rotation of 1 degree according to the rotation angle and the preset time period, and obtain a mapping relationship.

[0293] Optionally, in the fault monitoring device provided in the embodiment of the present application, the change unit 113 includes: a second conversion module, used to convert the moment value in the time axis of the initial high-frequency signal into a preset angle value according to a mapping relationship to obtain an initial angular domain vibration signal; an interpolation module, used to determine whether each preset angle value has a corresponding amplitude value in the initial angular domain vibration signal, and when there is no corresponding amplitude value for the target preset angle value, obtain multiple preset angle values ​​adjacent to the target preset angle value and the amplitude value of each preset angle value, use the interpolation method to calculate the amplitude value of the target preset angle value according to the amplitude value of each preset angle value, and add the amplitude to the corresponding position in the initial angular domain vibration signal to obtain the target angular domain vibration signal.

[0294] Optionally, in the fault monitoring device provided in the embodiment of the present application, the first determination unit 114 includes: a second acquisition module, used to obtain the verification angle and the standard amplitude at the verification angle from the standard data; a third determination module, used to determine the amplitude at the verification angle in the target angular domain vibration signal, obtain the target amplitude, and determine whether the target amplitude is greater than the standard amplitude; a fourth determination module, used to determine that there is a fault in the target device when the target amplitude is greater than the standard amplitude; and a fifth determination module, used to determine that there is no fault in the target device when the target amplitude is less than or equal to the standard amplitude.

[0295] Optionally, in the fault monitoring device provided in the embodiment of the present application, the third determination module includes: an acquisition submodule, used to obtain the angle range in the target angular domain vibration signal, and obtain the number of rotation cycles contained in the angle range to obtain the number of cycles, wherein the angle range is greater than 360 degrees; a calculation submodule, used to calculate the angle value corresponding to the verification angle in each cycle to obtain multiple preset angles; a determination submodule, used to determine the target angle corresponding to each preset angle in the target angular domain vibration signal to obtain multiple target angles, and obtain the amplitude at each target angle to obtain multiple target amplitudes.

[0296] The above-mentioned fault monitoring device includes a processor and a memory. The above-mentioned acquisition unit 111, filtering unit 112, change unit 113, first determination unit 114, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0297] The processor includes a core, which retrieves the corresponding program unit from memory. One or more cores can be configured, and adjusting the core parameters solves the problem of low timeliness and accuracy in manually determining whether a device is faulty in related technologies.

[0298] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0299] In the relevant technologies, dynamic equipment, mainly reciprocating equipment and rotating equipment, plays a vital role in industrial applications. Among them, the main body of the dynamic equipment generally includes: power unit, transmission unit, and execution unit. The power unit includes motors, piston engines, steam turbines, etc.; the transmission unit includes couplings, gearboxes, etc.; the execution unit includes centrifugal compressors, centrifugal pumps, reciprocating compressors, reciprocating pumps, etc. Due to the complex structure of the dynamic equipment and the fact that it generally performs high-intensity operations in harsh environments, it is prone to component fatigue failure or sudden failure. If it cannot be discovered and handled in advance, it will seriously affect production efficiency and increase production and operation and maintenance costs.

[0300] Existing technologies typically collect data from various sensors, generate signal time-frequency maps, and then identify faults based on expert experience or comparative analysis. However, this approach is not always effective in providing early warnings, relies on expert experience for fault identification, and results in high diagnostic and repair costs.

[0301] In this regard, the present application provides a method for monitoring faults of dynamic equipment to solve the technical problems in the existing technology of collecting various types of sensor data, generating signal time-frequency graphs, and discovering faults based on expert experience or comparative analysis, that is, the overall fault warning timeliness is not early enough, fault identification relies on expert experience, and the diagnostic and maintenance costs are high.

[0302] In order to solve the technical problems in the prior art of collecting various types of sensor data, generating signal time-frequency graphs, and discovering faults based on expert experience or comparative analysis, such as the overall fault warning timeliness is not enough in advance, fault identification relies on expert experience, and the diagnosis and maintenance costs are high, the present application provides a dynamic equipment fault monitoring method, device, electronic equipment and storage medium, which can realize real-time fault monitoring of dynamic equipment, and determine the fault point and fault type of the dynamic equipment when a dynamic equipment operation fault is detected, so as to realize timely discovery of dynamic equipment operation faults and diagnosis of the faults, reduce diagnosis and maintenance costs, and improve user experience.

[0303] To facilitate understanding of the dynamic equipment fault monitoring method provided in the embodiment of the present application, the dynamic equipment fault monitoring system involved in the embodiment of the present application is first illustrated below by way of example:

[0304] 12 is a schematic diagram of the structure of a mobile equipment fault monitoring system provided in an embodiment of the present application. As shown in FIG12 , the mobile equipment fault monitoring system 120 may include: a basic layer 121 , a functional layer 122 , and an application layer 123 .

[0305] Among them, the above-mentioned basic layer 121 may include but is not limited to a data source module 1211 and a data acquisition module 1212. The above-mentioned data source module 1211 may be a database for storing the basic operating parameter data of the target moving device, and the above-mentioned data acquisition module 1212 may be used to collect the above-mentioned basic operating parameter data of the target moving device, and send the collected above-mentioned basic operating parameter data to the data source module 1211 for storage.

[0306] Optionally, the basic operating parameter data of the target moving equipment may include but is not limited to: static attribute data of the target moving equipment, maintenance policy data of the target moving equipment, operating data of the target moving equipment, and sensor signal data of the target moving equipment collected by sensors.

[0307] The above-mentioned functional layer 122 may include but is not limited to a dynamic equipment monitoring module 1221. The dynamic equipment monitoring module 1221 can obtain the basic operating parameter data of the target dynamic equipment stored in the data source module 1211, and use the dynamic equipment fault monitoring method provided in the embodiment of the present application to perform fault monitoring on the target dynamic equipment based on the obtained basic operating parameter data.

[0308] The application layer 123 may include but is not limited to a terminal display module 1231, which may be used to display the fault monitoring status of the target mobile device, and when a fault is detected in the target mobile device, display the fault point and fault type of the target mobile device through the terminal.

[0309] In an embodiment of the present application, the executing body of the embodiment of the present application may be the dynamic equipment monitoring module 1221 of the functional layer 122, which may obtain the basic operating parameter data of the target dynamic equipment stored in the basic layer 121, and use the dynamic equipment fault monitoring method provided in the embodiment of the present application to perform fault monitoring on the target dynamic equipment.

[0310] The following is a further explanation of the method for monitoring faults of moving equipment provided by the present application with reference to specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation of the embodiments of the present invention.

[0311] See Figure 13, which is a flow chart of an embodiment of a method for monitoring a fault in a moving device provided in an embodiment of the present application. The process shown in Figure 13 can be applied to a moving device fault monitoring system, such as the moving device fault monitoring system 120 shown in Figure 12. As shown in Figure 13, the process may include the following steps:

[0312] Step 1301: Determine multiple test points of a target mobile device and a parameter matching rule for each test point.

[0313] The target moving equipment mentioned above refers to the moving equipment to be monitored for faults. It may include, but is not limited to: power units, transmission units, and actuators. Power units include motors, piston engines, and steam turbines; transmission units include couplings and gearboxes; and actuators include centrifugal compressors, centrifugal pumps, reciprocating compressors, and reciprocating pumps.

[0314] The above-mentioned test point refers to the location where the target moving equipment under test may fail, or the key location of the target moving equipment. By testing the operating status of the key location, it can be determined whether the target moving equipment fails. For example, the test point can be the motor, piston engine, and reciprocating pump of the moving equipment.

[0315] The above parameter matching rule refers to the correspondence between each test point and the corresponding parameter data. Through the correspondence, the target parameter data corresponding to each test point can be determined.

[0316] In one embodiment, the user may predetermine the test points of the target mobile device to be monitored, and input the location information of the test points of the target mobile device and the parameter matching rules of each test point into the execution body of the embodiment of the present application through a visual interface.

[0317] Based on this, the execution subject of the embodiment of the present application can determine multiple test points of the target dynamic device and the corresponding relationship between the position information of each test point and the parameter matching rule through the visual interface.

[0318] Based on this, the execution subject of the embodiment of the present application can obtain the position information of each test point on the target mobile device, and determine the parameter matching rule of each test point from the above-mentioned corresponding relationship according to the position information.

[0319] Step 1302: For each test point, match the target parameter data of the test point according to the corresponding parameter matching rule.

[0320] The above-mentioned target parameter data is the parameter data corresponding to each test point, which may include but is not limited to: static attribute data of the target moving equipment, maintenance policy data of the target moving equipment, operation data of the target moving equipment, and sensor signal data of the target moving equipment collected by sensors, etc.

[0321] In one embodiment, the execution subject of the embodiment of the present application can match the target parameter data of each test point according to the corresponding parameter matching rules for each test point after determining multiple test points of the target dynamic device and the parameter matching rules for each test point.

[0322] As an exemplary embodiment, the execution subject of the embodiment of the present application can obtain the target parameter data corresponding to each test point from a preset database (e.g., data source module 1211 shown in FIG1 ) according to the corresponding parameter matching rules. The database can pre-store the following parameter data of the target mobile device: static attribute data of the target mobile device, maintenance and warranty data of the target mobile device, operating data of the target mobile device, and sensor signal data of the target mobile device collected by sensors.

[0323] Specifically, the static attribute data of the target dynamic equipment may include but is not limited to: equipment name, equipment production date, equipment commissioning date, various rated parameters of the equipment and other static file attribute data.

[0324] The maintenance policy data of the target moving equipment may include but is not limited to: equipment daily maintenance time, maintenance person in charge, maintenance parts, degree of damage to accessories and other equipment maintenance information data.

[0325] The operating data of the above-mentioned target moving equipment may include but is not limited to: instrument operation record data such as speed, sand ratio, total pressure, displacement, power, etc. during equipment operation. The collection frequency is generally low, but it can directly respond to the current start-up, shutdown or operating status of the target moving equipment.

[0326] The sensor signal data of the target moving equipment collected by the sensor may be sensor signals used for moving equipment operation monitoring and analysis of physical parameter changes, which may include but are not limited to: vibration, temperature, pressure, sound print, etc.

[0327] Optionally, since there are differences in the types of monitoring sensors for different equipment such as rotating and reciprocating equipment, the above-mentioned sensors may include but are not limited to: vibration sensors for rotating parts, key phase, vibration, temperature, pressure sensors for reciprocating parts, and other sensor hardware that can be used for monitoring equipment operating status signals.

[0328] Furthermore, the target moving device may be pre-installed with multiple sensors, which may be installed at multiple test points of the target moving device, or may be installed at other locations of the target moving device, and the embodiment of the present application does not impose any restrictions on this. Based on this, the execution subject of the embodiment of the present application may obtain the preset sensor acquisition rules, and control the corresponding sensors to acquire the corresponding sensor signal data in accordance with the sensor acquisition rules. Afterwards, the collected sensor signal data may be stored in the above-mentioned database, and the data stored in the database may be updated in a timely manner. Among them, the above-mentioned sensor acquisition rules may be an acquisition mode pre-set by the user: for example, synchronous acquisition at 60-second intervals, triggering acquisition when the crankshaft speed is greater than 10 revolutions or the vibration characteristic value is greater than 1mm / s; sampling frequency: 25600Hz; number of sampling points: 102400; sensor sensitivity: 100mv / g.

[0329] For example, the following describes how to obtain sensor data, using the installation of vibration, pressure, and temperature sensors on a reciprocating piston pump as an example:

[0330] First, the vibration sensor can be installed on the radial and axial parts of each bearing seat of the reduction box (or other rotating parts) by means of thread or magnetism.

[0331] Second, the key phase gear disc is mounted on the power end crankshaft or a rotating component connected to the crankshaft and rotates with the crankshaft. The key phase sensor is mounted on the power end crankshaft or a supporting component connected to the crankshaft via a bracket and does not rotate with the crankshaft. The key phase gear disc has N teeth, (N-1) of which are evenly distributed, with an angle between the teeth of 360° / N. The remaining tooth has a different shape, either concave or convex. This tooth is used to define the zero point position, referred to as the zero point tooth. This method can be used to obtain the crankshaft speed and the angle of crankshaft rotation (this angle is related to the piston stroke). The obtained angle value is used as the horizontal coordinate to calibrate the changes in other signals (vibration, temperature, pressure, etc.).

[0332] The cranking process brings cylinder 1 to its top dead center, aligning the zero-point gear precisely with the key phase sensor. Since the angular difference between each cylinder is fixed, each cylinder has a zero-point angle reference when collecting sensor signals.

[0333] Third, the vibration sensor is installed on the radial and axial parts of the crankcase bearing seat, the crosshead load area, and the vertical direction of the packing cavity through threads or magnetic attraction.

[0334] Fourth, the temperature sensor is installed in the crosshead load area by means of threads or magnets to monitor the temperature of the crosshead components. A temperature-vibration integrated sensor can also be used.

[0335] Fifth, the pressure sensor can be installed on the suction gland through threads or other means to monitor the real-time dynamic pressure in the valve box cavity.

[0336] Step 1303: Determine the fault symptom quantity of the test point based on the above target parameter data.

[0337] The aforementioned fault symptom quantity refers to a fault symptom variable or factor that can have a strong correlation with a certain type of equipment fault. Each test point may correspond to one or more fault symptom quantities, which is not limited in the present embodiment.

[0338] In one embodiment, after determining the target parameter data corresponding to each test point of the target dynamic device, the execution subject of the embodiment of the present application can determine the fault symptom quantity of each test point according to the target parameter data corresponding to the test point.

[0339] As an exemplary embodiment, the execution entity of the embodiment of the present application may determine the location information of each test point and, based on the location information of the test point, determine the characteristic value calculation rule and the fault symptom quantity calculation rule corresponding to the test point. The characteristic value calculation rule and the fault symptom quantity calculation rule may be calculation rules pre-entered by the user.

[0340] Based on this, for each test point, the execution subject of the embodiment of the present application can calculate the target parameter data corresponding to the test point according to the above-mentioned characteristic value calculation rules to obtain at least one characteristic value corresponding to the test point.

[0341] The above-mentioned characteristic values ​​refer to the indicator data that can reflect the main distribution characteristics of the data, calculated or extracted by performing a series of numerical transformations on the target parameter data. They may include but are not limited to the following:

[0342] First, in each operating cycle of the target moving device, the time domain characteristic values ​​are calculated, which may include but are not limited to: effective value, maximum value, minimum value, mean value, average amplitude, peak-to-peak value, kurtosis value, skewness value, root mean square amplitude, peak factor, pulse factor, waveform factor, margin factor, etc.

[0343] Second, during each operating cycle of the target moving equipment, the characteristic values ​​of each angle segment are calculated according to the angle domain segmentation, which may include but are not limited to: effective value, maximum value, minimum value, mean value, average amplitude, peak-to-peak value, kurtosis value, skewness value, root amplitude, peak factor, pulse factor, waveform factor, margin factor, etc.

[0344] For example, based on the initial phase zero point position of the key phase signal, the full-cycle vibration signal of one working cycle of a 1V measuring point of a cylinder packing can be extracted; through the interpolation algorithm (including but not limited to the multi-strip interpolation algorithm, Lagrange interpolation algorithm, etc.), the equally spaced vibration time domain signal is converted into a vibration angle domain signal with equal intervals of 0 to 360° according to the crankshaft angle; the angular domain effective value is calculated according to every 10° or other angle segment.

[0345] Third, perform fast Fourier transform on the sensor signal to obtain the frequency domain signal, filter it according to different fixed frequency bands, and calculate the eigenvalues ​​of different frequency bands.

[0346] For example, assuming the original vibration signal analysis frequency range is 0-12800Hz, bandpass filtering can be performed at 0-2Hz, 2-1000Hz, and 1000-12800Hz to calculate the effective value of the corresponding frequency band. The 0-2Hz band can indicate whether the sensor is operating normally; the 2-1000Hz band can indicate low-frequency faults such as imbalance and misalignment; and the 1000-12800Hz band can indicate high-frequency faults such as bearing and gear wear.

[0347] Fourth, based on the transmission chain ratio and bearing model, the corresponding motor speed, gearbox parallel stage speed, gearbox planetary stage speed, crankshaft speed, and gearbox meshing frequency can be calculated according to the crankshaft speed; the gearbox and crankcase vibration sensor signals are fast Fourier transformed to obtain frequency domain signals, and the frequency domain signals are filtered by windowing function according to the above frequencies. For the filtered signals, the characteristic values ​​of the corresponding frequencies are calculated.

[0348] Subsequently, the target parameter data and / or characteristic values ​​can be combined according to the aforementioned fault symptom calculation rules to generate a fault symptom for the test point. Such fault symptom may include, but is not limited to, rotating component fault symptom, leakage fault symptom, sensor anomaly symptom, data-driven modeling fault identification symptom, and baseline space limit violation symptom.

[0349] The aforementioned rotating component fault symptom quantities are obtained by combining and calculating the bearing rotation frequency, gear meshing frequency, frequency sidebands, and other frequencies and their multiples or different frequency bands (combined calculation methods include but are not limited to addition, difference, integration, differentiation, etc.) to obtain fault symptom quantities that respond to different types of rotating component faults. The creation of rotating fault symptom quantities includes but is not limited to: bearing factor (which can reflect fault conditions such as bearing imbalance); bearing pad factor (which can reflect large head bearing fault conditions); parallel level load factor and planetary level load factor (which can reflect shaft misalignment fault conditions); sideband factor (which can reflect gear damage conditions), etc.

[0350] The aforementioned leakage fault indicators are calculated by combining specific frequency bands or frequencies of the in-cylinder vibration signal during the equipment's reciprocating motion (combined calculation methods include, but are not limited to, addition, subtraction, integration, and differentiation) to obtain indicators of in-cylinder leakage. Examples of leakage fault indicators include, but are not limited to, effective leakage factors, energy leakage factors, and acceleration leakage factors (which can reflect fault conditions caused by leakage from seals such as in-cylinder check valves).

[0351] The above-mentioned sensor abnormality sign quantity is obtained by calculating the characteristic indicators that have a strong correlation with the sensor state, such as the energy value and bias voltage value of a specified frequency segment of a signal frequency domain data, and performing combined calculations (the combined calculation method includes but is not limited to addition, difference, integration, differentiation, etc.) to obtain the sign quantity that can respond to the sensor abnormality.

[0352] The aforementioned data-driven modeling of fault identification symptoms is based on normal unit signals collected in the laboratory or at the field, and is developed using algorithms such as unsupervised machine learning. This allows for the output of fault state and health state identification errors and judgment results for designated measurement points.

[0353] For example, historical vibration signal data can be obtained from a measuring point on the input side of the large gear of a piston pump reducer when the equipment is in a healthy state (data collection is performed at 60-second intervals, a sampling frequency of 25,600 Hz, and 102,400 sampling points). A fast Fourier transform is performed on each minute of sample data. The raw frequency domain data is used as training sample data, and a symmetric autoencoder neural network model based on a one-dimensional convolutional neural network layer is constructed. This training sample data is then input and the model is iteratively trained in multiple batches. By inputting the vibration data to be measured into the model, the fault identification error is output, which is the fault identification sign quantity for that measuring point.

[0354] The baseline space out-of-limit indicators described above can be constructed from normal unit signals collected in the laboratory or on-site, creating a probability distribution space in the time or frequency domain, referred to as the baseline space. By inputting a signal into the baseline space, the difference distribution between the input signal and the baseline space is obtained. This difference distribution is then calculated to obtain a characteristic value, which reflects the unit's operating status.

[0355] For example, a normal vibration signal from a gearbox input 1H measurement point can be collected, along with historical vibration signal data when the equipment is in a healthy state (data collection is performed at 60-second intervals, a sampling frequency of 25,600 Hz, and 102,400 sampling points). A Fast Fourier Transform (FFT) is then performed on each minute of sample data. The amplitudes of all frequency-domain samples are summed and averaged according to their corresponding frequencies to obtain an average amplitude spectrum, which is the frequency-domain baseline space. The frequency-domain signal to be measured is then input and subtracted from the baseline space. When the amplitude of a frequency is less than the corresponding amplitude in the baseline space, the difference is reset to zero. When it is greater than the amplitude in the baseline space, the difference is accumulated and summed. The sum of these differences represents the baseline space overrun indicator for that measurement point.

[0356] Step 1304: Determine the operating status of each test point based on the fault symptom quantity corresponding to each test point.

[0357] Step 1305: When it is determined that the target mobile device has a fault through the operating status of each test point, the fault point and fault type of the target mobile device are determined according to the operating status of each test point.

[0358] The following is a unified description of step 1304 and step 1305:

[0359] The above-mentioned operating status refers to the corresponding status of the test point during the operation of the target dynamic equipment, which may include a fault state and a non-fault state. The non-fault state is a normal operating state. Furthermore, the above-mentioned fault state can be further divided into a low fault state, a medium fault state, and a high fault state. The fault levels increase in sequence. The higher the fault level, the greater the fault.

[0360] In one embodiment, the execution subject of the embodiment of the present application may determine the operating status corresponding to each test point in the target dynamic device according to one or more fault symptom quantities corresponding to the test point.

[0361] As for how to determine the operating status corresponding to each test point based on the fault symptom quantity corresponding to each test point, it can be explained below through the process shown in Figure 14 and will not be described in detail here.

[0362] Afterwards, the operating status of each test point can be used to determine whether the target mobile device has a fault. If it is determined that the target mobile device has a fault, the fault point and fault type of the target mobile device can be determined based on the operating status of each test point.

[0363] As an exemplary embodiment, when it is determined that the operating status of any at least one test point among the multiple test points of the target mobile device is a faulty state, it can be determined that the target mobile device has a fault.

[0364] As for how to determine the fault point and fault type of the target moving equipment based on the operating status of each test point, it can be explained below through the process shown in Figure 18 and will not be described in detail here.

[0365] In addition, in order to enable users to promptly understand whether there is a fault in the target mobile device, the executive body of the embodiment of the present application can, when determining that there is a fault in the target mobile device, issue an alarm through a preset alarm method to remind the user that there is a fault in the target mobile device. The above-mentioned alarm method can be an alarm information sound and light prompt, or SMS push, etc.

[0366] Furthermore, after determining the fault point and fault type of the target moving equipment, the fault point and fault type of the target moving equipment can be output through a visual interface, and the technician can determine whether the fault point and fault type of the target moving equipment are accurate. After receiving the technician's confirmation, the fault point and fault type are stored in the fault case library.

[0367] Furthermore, to facilitate users' further understanding of the operating status of the target dynamic equipment, the execution entity of the embodiment of the present application can perform signal processing and feature transformation on the parameter data of the target dynamic equipment stored in the database (e.g., data source module 1211) according to preset signal processing rules to obtain various graph data. Such graph data may include, but is not limited to: unit status diagrams, vibration monitoring diagrams, envelope demodulation diagrams, order ratio diagrams, angular domain monitoring diagrams, multi-trend diagrams, multi-parameter analysis diagrams, and baseline space diagrams.

[0368] In addition, in order to facilitate users to understand the fault point and fault type of the target moving device in more detail, the executing entity of the embodiment of the present application can generate a three-dimensional image of the target moving device after determining the fault point and fault type of the target moving device, and mark the fault point and fault type of the target moving device in the three-dimensional image (for example, mark the fault point of the target moving device in red) to obtain the target three-dimensional image.

[0369] Afterwards, the three-dimensional image of the target can be output and an early warning can be issued through a preset early warning method.

[0370] The technical solution provided by the embodiments of the present application determines multiple test points for a target mobile device and parameter matching rules for each test point. For each test point, the target parameter data for that test point is matched according to the corresponding parameter matching rules. Based on the target parameter data, the fault symptom quantity for that test point is determined. Based on the corresponding fault symptom quantity for each test point, the operating status of each test point is determined. When the operating status of each test point determines that a fault exists in the target mobile device, the fault point and fault type of the target mobile device are determined based on the operating status of each test point. This technical solution, by pre-setting multiple test points for the target mobile device and determining the operating status of each test point based on the fault symptom quantity of each test point, performs fault detection on the target mobile device. Compared to discovering faults in the target mobile device based on expert experience or comparative analysis, this technical solution can achieve real-time fault monitoring of the mobile device. When a mobile device operational fault is detected, the fault point and fault type of the mobile device are determined, thereby enabling timely discovery and diagnosis of mobile device operational faults, reducing diagnostic and maintenance costs and improving user experience.

[0371] See Figure 14, which is a flowchart of another embodiment of a method for monitoring faults in a moving device provided in an embodiment of the present application. The process shown in Figure 14, based on the process shown in Figure 13, specifically describes how to determine the operating status corresponding to each test point based on the fault symptom corresponding to each test point. As shown in Figure 14, the process may include the following steps:

[0372] Step 1401: Obtain the highest speed and the lowest speed of the target moving device in multiple operation cycles.

[0373] The above-mentioned operation cycle refers to the operation cycle of the target mobile device within a preset historical time period. The above-mentioned preset historical time period may be one minute or two minutes before the current moment, and the embodiment of the present application does not impose any limitation on this.

[0374] The above-mentioned maximum speed refers to the maximum speed at which the target moving equipment operates within the above-mentioned preset historical time period.

[0375] The minimum rotational speed refers to the lowest rotational speed at which the target device operates within the preset historical time period.

[0376] In one embodiment, the execution subject of the embodiment of the present application may obtain the maximum speed and the minimum speed of the target moving equipment in the above-mentioned multiple operation cycles from a preset database.

[0377] In another embodiment, the execution subject of the embodiment of the present application may obtain the maximum speed and the minimum speed of the target moving device input by the user in multiple operation cycles through a visual interface.

[0378] Step 1402: Determine the speed difference between the maximum speed and the minimum speed, and determine whether the speed difference is less than a preset difference threshold. If so, execute step 303; if not, end the process.

[0379] Step 1403: When it is determined that the speed difference is less than the difference threshold, the operating state of each test point is determined based on the preset dynamic warning model and the fault symptom quantity corresponding to each test point.

[0380] The following is a unified description of step 1402 and step 1403:

[0381] In an embodiment of the present application, when fault monitoring is performed on a target moving device, the maximum speed and minimum speed of the target moving device can be used to first determine whether the target moving device meets the fault monitoring conditions. If so, the process can continue to determine whether the target moving device has a fault; if not, the process can be terminated.

[0382] In one embodiment, after determining the maximum speed and minimum speed of the target moving device in multiple operating cycles, the executing entity of the embodiment of the present application may subtract the minimum speed from the maximum speed to determine the speed difference corresponding to the target moving device, and further determine whether the speed difference is less than a preset difference threshold.

[0383] Optionally, if it is determined that the above speed difference is greater than or equal to the difference threshold, it means that the target dynamic device is unstable at this time and fault monitoring cannot be performed. Therefore, the process can be directly ended and an alarm message indicating that the current target dynamic device is unstable can be output.

[0384] Conversely, if the speed difference is determined to be less than the difference threshold, it indicates that the target moving equipment operating temperature is under control and fault monitoring can be performed. Therefore, the operating status of each test point can be determined based on a preset dynamic early warning model and the fault symptom quantity corresponding to each test point. The dynamic early warning model can be a pre-trained model for predicting early warning results for each fault symptom quantity.

[0385] As an exemplary embodiment, the execution entity of the embodiment of the present application may first obtain the historical operating data of the target dynamic equipment during a preset historical time period, as well as the configuration parameters of the preset dynamic early warning model. The preset historical time period may be the past day, two days, or a week, and the embodiment of the present application does not impose any restrictions on this. The configuration parameters of the dynamic early warning model may be parameters involved in the dynamic early warning model, such as multiple thresholds for each fault symptom quantity.

[0386] Afterwards, the above historical operation data, the fault symptom quantity corresponding to each test point, and the above configuration parameters can be input into the dynamic early warning model to obtain the early warning results corresponding to each fault symptom quantity output by the dynamic early warning model, such as 1 (normal), 2 (low fault risk), 3 (medium fault risk), and 4 (high fault risk).

[0387] Finally, the operating state of each test point may be determined based on the early warning result corresponding to each fault symptom quantity, wherein the operating state may include a fault state and a non-fault state.

[0388] In one embodiment, the aforementioned fault symptom quantity may include a sensor abnormality symptom quantity, which can be used to indicate whether a sensor collecting parameter data of the target moving equipment is abnormal. Based on this, when determining the operating status of each test point based on the warning result corresponding to each fault symptom quantity, the execution subject of this embodiment of the application may first determine whether the warning result corresponding to the aforementioned sensor abnormality symptom quantity is a fault warning.

[0389] Optionally, if it is determined that the warning result corresponding to the abnormal sign of the sensor is a fault warning, it means that there is an abnormality in the sensor used to collect the parameter data of the target moving equipment at this time, and the parameter data used as the basis for fault monitoring of the target moving equipment may be wrong. Therefore, the fault monitoring judgment can be terminated directly and the alarm information of the sensor abnormality can be output.

[0390] On the contrary, if it is determined that the warning result corresponding to the abnormal sign of the sensor is a non-fault warning, the sensor signal data collected by the current sensor is further obtained, and the sensor signal data is input into a preset sensor fault identification model to obtain the sensor abnormality identification result output by the sensor fault identification model.

[0391] The sensor fault identification model can be a neural network model constructed based on historical normal device data and sensor fault sample data, and sensor fault identification is achieved through model training. The neural network model can be an unsupervised learning model, a supervised classification learning model, etc., and the network structure can be, but is not limited to, an autoencoder neural network structure, a convolutional neural network structure, a fully connected neural network model structure, or other network structures that are combinations of the above neural network model structures.

[0392] Furthermore, if the above sensor abnormality identification result indicates that the sensor is abnormal, it means that the parameter data collected by the target moving device through the sensor is inaccurate. Therefore, it can be determined that the operating status of each test point is a non-fault state.

[0393] Optionally, if the above sensor abnormality identification result indicates that the sensor is normal, it means that the sensor is operating normally at this time. Therefore, the operating status of the test point can be determined by the warning result level of each fault symptom quantity corresponding to each test point.

[0394] As an exemplary embodiment, since each test point of the target dynamic equipment may include one or more fault symptom quantities, and the warning result of each fault symptom quantity may include multiple levels, such as 1 (normal), 2 (low fault risk), 3 (medium fault risk), and 4 (high fault risk), the warning result level of each symptom quantity corresponding to each test point can be determined, and the fault state corresponding to the warning result with the highest warning result level can be determined as the operating state of the corresponding test point.

[0395] For example, assuming that test point 1 includes three fault signs, and their corresponding warning results are 1 (normal), 2 (low fault risk), and 4 (high fault risk), it can be seen that the warning result with the highest warning level is 4 (high fault risk). Continuing to assume that the fault state corresponding to 4 (high fault risk) is high-risk fault operation, then the operating state corresponding to test point 1 is high-risk fault operation.

[0396] The following uses the fault warning of a certain plunger pump hydraulic end input side measurement point fault identification sign quantity (hereinafter referred to as ae sign quantity) as an example to illustrate how to determine the operating status corresponding to each test point based on the fault sign quantity corresponding to each test point in the embodiment of the present application. Referring to Figure 15, there is a flowchart of another embodiment of the dynamic equipment fault monitoring method provided in the embodiment of the present application. As shown in Figure 15, the process may include:

[0397] First, input the current motor speed and speed difference (i.e., the difference between the highest and lowest speeds of multiple equipment operating cycles within one minute) to determine whether the speed difference is greater than the speed difference threshold α. If the speed difference is greater than or equal to α, exit the early warning logic; otherwise, input the current ae sign quantity and its historical data at the previous n moments into the dynamic early warning model, and output the initial early warning result result_label.

[0398] Afterwards, determine whether the sensor abnormality sign quantity in result_label is label 2 / 3 / 4 (that is, the warning result is 2 (low fault risk), 3 (medium fault risk), or 4 (high fault risk)). If so, exit the warning logic; otherwise, input the current raw signal data into the sensor fault identification model.

[0399] Next, the sensor fault identification model outputs a sensor fault status label, which is categorized as 1 (normal), 2 (low fault risk), 3 (medium fault risk), and 4 (high fault risk). If the label is 2 / 3 / 4, the initial warning result for the sign quantity labeled 2 / 3 / 4 in result_label is corrected to 1. Otherwise, the initial warning result is directly output as the final warning result.

[0400] Among them, the warning threshold of the above-mentioned dynamic fault warning model involves the configuration of multi-level warning thresholds for all symptom quantities. It is equipped with a front-end interface that can independently perform the above-mentioned threshold configuration function interface. The front-end manually configures the background to synchronously update the threshold configuration table. The threshold configuration data function interface is shown in Figure 16. See Figure 16, which is a schematic diagram of a threshold configuration data function interface provided in an embodiment of the present application. As shown in Figure 16, the threshold configuration data function interface may include: device ID, measurement point location, symptom quantity, symptom quantity name, first-level threshold, whether to enable, trigger alarm category, second-level threshold, whether to enable, trigger alarm category, third-level threshold, whether to enable, first-level trigger alarm category and other parameters.

[0401] In addition, the execution body of the embodiment of the present application may also include a historical alarm log, which is an important tool. All alarms that have occurred and whether these alarms have been confirmed can be clearly viewed, which can make users sure that they have mastered all alarms. The historical alarm log data function interface is shown in Figure 17. Referring to Figure 17, a schematic diagram of a historical alarm log data function interface provided by an embodiment of the present application is shown. As shown in Figure 17, the historical alarm log data function interface may include: event ID, alarm category, device ID, measurement point location, symptom quantity, symptom quantity name, symptom quantity value, symptom quantity over-limit threshold, rule fault tree fault label, data model fault label, confirmation status, alarm start time, alarm confirmation time, confirmation personnel, manual judgment results, and remarks and other parameters.

[0402] The technical solution provided by the embodiment of the present application obtains the maximum speed and minimum speed of the target dynamic equipment in multiple operating cycles, determines the speed difference between the maximum speed and the minimum speed, and determines whether the speed difference is less than a preset difference threshold. If so, then when it is determined that the speed difference is less than the difference threshold, the operating status of each test point is determined based on a preset dynamic early warning model and the fault symptom quantity corresponding to each test point; if not, the process ends. This technical solution determines the operating status of the test point based on a pre-trained dynamic early warning model and the fault symptom quantity corresponding to each test point when it is determined that the target dynamic equipment is operating stably. By training the dynamic early warning model, the operating status of each test point can be determined more accurately, thereby achieving more accurate determination of the operating status of each test point of the target dynamic equipment, thereby achieving timely discovery of dynamic equipment operating faults and diagnosis of the faults, reducing diagnostic and maintenance costs, and improving user experience.

[0403] See Figure 18, which is a flowchart of another embodiment of a method for monitoring a fault in a moving device provided in an embodiment of the present application. The process shown in Figure 18, based on the process shown in Figure 14, further describes how to determine the fault point and fault type of the target moving device based on the operating status of each test point. As shown in Figure 18, the process may include the following steps:

[0404] Step 1801: When it is determined that a target moving device has a fault, determine whether the operating state of each test point is a fault state.

[0405] Step 1802: Determine the test point whose operating status is a fault state as the initial fault point.

[0406] The following is a unified description of step 1801 and step 1802:

[0407] As shown in the process of Figure 14, the early warning results of the fault sign quantity corresponding to each test point of the target dynamic equipment can include: 1 (normal), 2 (low fault risk), 3 (medium fault risk), and 4 (high fault risk). Based on this, the operating state of each test point can be determined as a fault state or a non-fault state. Its fault state can be further divided into low-risk fault operation, medium-risk fault operation, and high-risk fault operation.

[0408] Based on this, when the execution subject of the embodiment of the present application determines that the operating status of any test point of the target moving device is a faulty state, it is determined that the target moving device has a fault.

[0409] Furthermore, when it is determined that the target device has a fault, it may be determined whether the operating status of each test point is a fault state, and the test point whose operating status is a fault state is determined as the initial fault point.

[0410] Step 1803: For each initial fault point, input the early warning result of the initial fault point into a preset fault detection model to obtain the fault point and fault type of the target moving equipment output by the fault detection model.

[0411] The above-mentioned fault detection model may be a pre-trained model for detecting the fault point and fault type of the target moving equipment.

[0412] In the embodiment of the present application, the early warning result corresponding to the initial fault point can be input into the above-mentioned fault detection model, so as to obtain the fault point and fault type of the target moving equipment output by the fault detection model.

[0413] Furthermore, the above-mentioned fault detection model may include a typical fault classification model and a fault analysis model. The above-mentioned typical fault classification model can be used to determine whether the fault type of the target moving equipment is a typical fault, and the above-mentioned fault analysis model can be used to analyze the fault point and basic fault type of the target moving equipment.

[0414] Based on this, when the fault point and fault type of the target dynamic equipment are determined through the fault detection model, the early warning results corresponding to the initial fault point can be input into the typical fault classification model and the fault analysis model respectively to obtain the typical fault classification results output by the typical fault classification model and the fault analysis results output by the fault analysis model.

[0415] Afterwards, the typical fault classification results and fault analysis results can be weighted and summed to obtain the fault point and fault type corresponding to the target moving equipment.

[0416] Furthermore, in order to provide users with more accurate fault types and corresponding solutions, thereby improving user experience, the execution entity of the embodiment of the present application may determine whether the above-mentioned fault type is a typical fault type.

[0417] Optionally, when it is determined that the fault type is a typical fault type, the fault type can be matched with a preset typical fault case library to obtain a target solution corresponding to the fault type, wherein the above-mentioned typical fault case library can be used to store typical faults and the solutions corresponding to each typical fault.

[0418] The target solution can then be output via a visualization interface.

[0419] In addition, the above fault types and target solutions are manually confirmed. Business personnel or diagnostic personnel can manually confirm the current fault diagnosis results on the front-end operation interface. After manual confirmation, the fault case events and solutions will be automatically transferred to the fault case library.

[0420] For example, the expert judgment rules for coupling misalignment faults in a plunger pump rule-based fault diagnosis model are shown in Table 1:

[0421] Table 1

[0422] Continuing to assume that the confidence level of the rule is 95%, the failure probability of "coupling misalignment" = 95%*(0.3s1+0.4s2+0.4s3).

[0423] The technical solution provided by the embodiment of the present application determines whether the operating status of each test point is a faulty state when a fault is determined in the target moving device. The test point with a faulty operating status is determined as the initial fault point. For each initial fault point, the early warning result of the initial fault point is input into a preset fault detection model to obtain the fault point and fault type of the target moving device output by the fault detection model. This technical solution, through the preset fault detection model and based on the operating status of each test point, can accurately predict the fault point and fault type of the target moving device, thereby achieving more accurate determination of the fault point and fault type of the target moving device, thereby enabling timely discovery of moving device operating faults and diagnosis of the faults, reducing diagnostic and maintenance costs, and improving user experience.

[0424] See Figure 19, which is a schematic diagram of the structure of another dynamic equipment fault monitoring system provided in an embodiment of the present application. As shown in Figure 19, the dynamic equipment fault monitoring system may include: a data source, a data acquisition module, a data processing module, an early warning center module, a data storage module, a data communication module, a graph analysis module, and a fault diagnosis module.

[0425] Among them, the data source can be used to store including but not limited to the following four types of parameter data: static attribute data of equipment archives, equipment maintenance work order data, equipment operation control data, and sensor data.

[0426] The data acquisition module can collect analog signals from a variety of sensors and convert them into digital signals. The acquisition module can set the acquisition mode, sampling frequency, number of sampling points, sensor type, sensor sensitivity, and more. It supports multi-channel synchronous acquisition, interval acquisition, and acquisition triggered by speed or characteristic indicators.

[0427] The aforementioned data processing module primarily involves speed calculation, eigenvalue calculation, and symptom quantity calculation. Eigenvalues ​​are indicators that reflect the primary distribution characteristics of data, calculated or extracted through a series of numerical transformations of raw signals such as static attribute data, equipment maintenance data, operational control data, and sensor data. Symptom quantities are fault symptom variables or factors that are strongly correlated with specific types of equipment failures, generated by combining eigenvalues ​​or raw data.

[0428] The above-mentioned early warning center module, also known as the alarm module, realizes the main functions of dynamic fault early warning, including: stable speed working condition identification function, equipment dynamic early warning function, and sensor anomaly identification function. The input data of the early warning center is the symptom quantity to be tested after data processing, the motor speed difference and the dynamic early warning model configuration parameters; the output data is the current early warning status of each symptom quantity, and the status level is divided into 1 (normal), 2 (low fault risk), 3 (medium fault risk), and 4 (high fault risk). The early warning center implements dynamic multi-level fault early warning through a series of data flows, function item calls and control logic, and can prevent abnormal point false alarms, repeated alarms of the same alarm event, etc. At the same time, the front-end interface provides custom configuration functions for the thresholds of the dynamic early warning function and the parameters of the early warning model. Users can customize the early warning model according to actual business needs.

[0429] Among them, alarm information can be displayed on large-screen monitoring, PC terminals, handheld terminal apps, etc., and sound and light alarm prompts can be issued. Alarm information can also be pushed via SMS, message prompts, emails, etc. In addition, for historical alarm information, the system provides historical alarm log query, storage and analysis functions.

[0430] The aforementioned typical fault diagnosis module primarily implements fault classification prediction, fault location, and solution recommendations based on fault warnings. This module includes AI-based fault diagnosis and rule-based fault diagnosis. AI-based fault diagnosis primarily builds a neural network model based on historical data from normal equipment and typical fault samples of varying levels. Through model training, it develops an AI fault diagnosis model for typical faults and achieves classification predictions for typical faults. Rule-based fault diagnosis relies on expert experience rules, including various structured storage, to locate and classify faults in warning data. The expert case library contains structured storage results of various typical fault case events, along with expert-recommended solutions.

[0431] The details are as follows: For the alarm results of the early warning center, the fault diagnosis module is input for further fault location and classification. If the early warning result of a certain measuring point is 2 / 3 / 4, the AI ​​fault diagnosis model and the rule-based fault diagnosis model of the measuring point will be called, and the prediction results of the two types of models will be weighted and fused to output the predicted fault type of the measuring point. At the same time, the expert recommended solution for the fault is output based on the fault case library associated with the fault type. Among them, the AI ​​fault diagnosis model is a pre-trained model, which targets typical faults of various types of equipment, including but not limited to valve leakage fault classification prediction models, bearing fault classification prediction models, gear fault classification prediction models, etc. The rule-based model is generated by structured processing and storage of expert judgment rules for various historical fault events.

[0432] The fault diagnosis module also provides a manual confirmation interface for prediction results. Business personnel or diagnostic personnel can manually confirm the current fault diagnosis results on the front-end operation interface. After manual confirmation, the fault case event and solution will be automatically transferred to the fault case library.

[0433] The aforementioned storage module can store data from the acquisition, calculation, and alarm modules, meeting data capacity requirements for at least six months or other periodic data. Regarding the dilution strategy for sensor data collection: alarm data is permanently stored; normal data is diluted based on importance by time period, such as year, month, day, and hour, to ensure data is available for every time period. The longer the time period, the more diluted the data.

[0434] For example, the storage of alarm data of a plunger pump fault dynamic warning system includes: all symptom quantity original data and all symptom quantity status label data.

[0435] The aforementioned communication module can be divided into three parts: internal system communication, external system communication, and system communication link. Internal system communication is primarily achieved through communication methods such as UDP / HTTP; external system communication uses standard communication protocols (ModBus / OPC / TCP / IP...) or custom protocols. Internal and external communication as a whole includes hardware layer protocols, network layer transmission protocols, and application layer protocols. This completes data transmission and exchange between different information systems (PLCs, operational data acquisition systems, etc.) and between different units and modules within the system. Furthermore, the system's data communication link consists of sensors, data collectors, edge industrial computers, industrial gateways, and remote servers.

[0436] The above-mentioned spectrum analysis module can be used to process the original signal and transform its features to obtain professional spectrum data, which can assist engineers in diagnosing unit faults. It mainly includes the following functions:

[0437] Unit status diagram: displays the real-time operating status of each reciprocating pump and sub-components, status: normal, warning, alarm, high alarm;

[0438] Vibration monitoring diagram: displays the real-time and historical characteristic value trends, waveforms, and spectra of each vibration measurement point.

[0439] Envelope demodulation diagram: The high-frequency resonance response wave generated by the fault impact is amplified and converted into a low-frequency waveform with fault characteristic information through the envelope detection method, and then the characteristic frequency of the fault is found using the spectrum analysis method.

[0440] Optionally, the algorithm steps may be: first, the filtering characteristics of the bandpass filter are used to filter out the target frequency component; second, the modulation signal is extracted from the signal using the Hilbert transform and the change of the modulation signal is analyzed; finally, the low-frequency signal component is separated from the vibration signal through the Fourier transform.

[0441] Order ratio diagram: Combined with key phase data, the time domain waveform of equal time sampling is converted into the time domain waveform of equal angle sampling, and then Fourier transform and other transformations are performed;

[0442] Angle domain monitoring diagram: displays the full cycle signal of each reciprocating measuring point, including the angle domain waveform diagram, angle domain envelope diagram, and angle domain histogram.

[0443] Optionally, the algorithm steps may be: first, intercept the entire cycle key phase data and vibration data or other parameter data; second, interpolate the key phase data to obtain the time corresponding to each 1° or other equally spaced angles, that is, the set {angle, t}; then interpolate the vibration data, with the interpolation point being angle, to obtain the amplitude corresponding to angle, that is, the set {angle, amp}, which is the angular domain waveform; perform Hilbert transform on the angular domain waveform to obtain the angular domain envelope diagram; perform segmented cutting calculation on the angular domain waveform to obtain the angular domain histogram.

[0444] Multi-trend graph: displays real-time and historical characteristic value trends of state parameters such as speed, vibration, pressure, and temperature;

[0445] Multi-parameter analysis diagram: Synchronous analysis of the full-cycle vibration waveform, cavity pressure waveform, crosshead temperature, etc. with the same horizontal coordinate;

[0446] Baseline space diagram: displays the comparison and difference results with the baseline space.

[0447] The dynamic equipment fault monitoring system provided in the embodiment of the present application can monitor the real-time operating status and fault warning of various components of the dynamic equipment. It contains different types of equipment fault signs to realize equipment fault response. The early warning center calls the dynamic early warning model and the sensor fault identification model through the early warning center logic to realize equipment fault early warning and sensor fault judgment. It can also realize the output of typical fault classification prediction results and fault solution output, as well as structured storage of fault cases. Furthermore, the visual interface can realize the whole process fault event identification and tracing from early warning model configuration, early warning alarm to manual confirmation, graph analysis, etc., which greatly reduces the professional threshold of users and improves production operation efficiency.

[0448] See Figure 20, which is a block diagram of an embodiment of a moving equipment fault monitoring device provided in an embodiment of the present application. The device shown in Figure 20 can be applied to the moving equipment fault monitoring system shown in Figure 12. As shown in Figure 20, the device may include:

[0449] The first determination module 2001 is used to determine multiple test points of the target mobile device and a parameter matching rule for each test point;

[0450] A matching module 2002 is configured to match target parameter data of each test point according to corresponding parameter matching rules;

[0451] A second determining module 2003 is configured to determine a fault sign quantity of the test point according to the target parameter data;

[0452] A third determining module 2004 is configured to determine the operating status of each test point based on the fault symptom quantity corresponding to each test point;

[0453] The fourth determining module 2005 is configured to determine the fault point and fault type of the target moving device according to the operating status of each test point when it is determined that the target moving device has a fault through the operating status of each test point.

[0454] In the related art, in order to ensure the safe operation of mechanical equipment, a large number of different sensors are usually set in the mechanical equipment, so that during the operation of the mechanical equipment, relevant data can be collected through the sensors, and the operating status of the mechanical equipment can be analyzed based on the collected data, thereby ensuring the safe operation of the mechanical equipment. It can be seen that the sensors set in the mechanical equipment are crucial to ensuring the safe operation of the mechanical equipment. However, the sensors will be affected by various factors during the working process, which will cause sensor failure, and sensor failure will affect the safe operation of the mechanical equipment. At present, in order to avoid the impact of sensor failure on the safe operation of mechanical equipment, the sensors in the mechanical equipment are usually manually identified regularly, so that the faulty sensors are replaced when they are identified. However, due to the large number of sensors in the mechanical equipment, identifying sensor failures in the above manner will not only waste a lot of manpower and material resources, but will also affect the operation of the mechanical equipment, and thus affect production efficiency.

[0455] In this regard, the present application provides a sensor fault identification method to solve the technical problem that identifying sensor faults in the above manner not only wastes a lot of manpower and material resources, but also affects the operation of mechanical equipment, thereby affecting production efficiency.

[0456] Referring to FIG21 , FIG21 is a flow chart of a sensor fault identification method provided in an embodiment of the present application. The sensor fault identification method provided in an embodiment of the present application includes the following steps:

[0457] S2101: Acquire a historical sensor data set corresponding to the target sensor to be identified.

[0458] S2102: Determine a fault identification model corresponding to the target sensor based on the historical sensor data set.

[0459] For steps S2101 and S2102, the target sensor is installed in the mechanical equipment. The type of the target sensor can be selected according to actual needs, and the type of the target sensor is not specifically limited in this embodiment. For example, the target sensor can be a vibration sensor, a temperature sensor, a pressure sensor, etc. The historical sensor data set includes a plurality of historical sensor data, and the historical sensor data is the sensor data before the fault of the target sensor is identified. The historical sensor data specifically refers to the return data of the historical sensor. Among them, the historical sensor data set includes a plurality of normal historical sensor data, and the historical sensor data set also includes a plurality of historical sensors of different fault types. The historical fault data set can be obtained from the logs recorded during the operation of the mechanical equipment.

[0460] It should be noted that target sensors of the same type are placed at different positions in the mechanical equipment. When using the historical sensor data set to determine the fault identification model, in order to ensure the accuracy of the obtained fault identification model, historical sensor data of all target sensors of the same type in the mechanical equipment can be obtained, and the historical sensor data set can be obtained after screening all the obtained historical sensor data.

[0461] Specifically, in order to achieve sensor fault identification, a preset classification model can be pre-selected. After obtaining the historical sensor data set, the preset classification model can be trained to obtain the final fault identification model. When the fault identification model is used to identify the fault of the target sensor, it can be identified whether the target sensor is faulty and the corresponding fault type of the target sensor. Specifically, determining the fault identification model corresponding to the target sensor based on the historical sensor data set includes:

[0462] For each historical sensor data in the historical sensor data set, the historical sensor data is labeled by fault classification to obtain training sample data corresponding to the historical sensor data;

[0463] The preset classification model is trained based on the training sample data corresponding to all historical sensor data in the historical sensor data set to obtain a fault recognition model corresponding to the target sensor.

[0464] In this embodiment, after obtaining a historical sensor dataset corresponding to the target sensor to be identified, the fault type of each historical sensor in the historical sensor dataset is annotated to distinguish whether the historical sensor data is normal data or faulty data. If the historical sensor data is faulty data, the fault type is also annotated. For example, the fault type may be electromagnetic interference fault, short circuit fault, or open circuit fault. After fault classification and annotating each historical sensor data in the historical sensor dataset, training sample data corresponding to each historical sensor data in the historical sensor dataset is obtained, and further, a training sample dataset corresponding to the historical sensor dataset is obtained. All training sample data in the training sample dataset is divided according to a preset ratio to obtain a training set, a validation set, and a test set. A preset classification model is trained based on the training set, validation set, and test set to obtain a fault identification model corresponding to the target sensor. The preset ratio can be set according to actual needs, and the specific value of the preset ratio is not specifically limited in this embodiment. For example, the preset ratio can be 20:3:2. The preset classification model can also be selected according to actual needs, and the specific form of the preset classification model is not limited in this embodiment. For example, the preset classification model can be a convolutional neural network model.

[0465] S2103: When actual sensor data of the target sensor is obtained, at least one preset mechanism model corresponding to the target sensor is obtained. The preset mechanism model is used to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor itself when the target sensor itself has a fault.

[0466] In this embodiment, the actual sensor data of the target sensor is actually the data actually transmitted back by the target sensor during operation. Types of target sensor faults primarily include target sensor disconnection, target sensor reverse power supply, target sensor short circuit, and target sensor overload. For each target sensor fault, a corresponding preset mechanism model can be pre-set. During the target sensor's operation, the preset mechanism model can be used to identify the target sensor fault, thereby determining whether the target sensor itself is faulty and, if so, the corresponding fault type.

[0467] Specifically, at least one preset mechanism model includes a bias voltage model, an output signal model, and a spectrum model. The actual sensor data includes an actual bias voltage, an actual output signal value, and a spectrum model. Each preset mechanism model stores a preset threshold value and a corresponding relationship between the comparison result and the fault identification result. The comparison result is the comparison result between the actual sensor data and the preset threshold value. That is, for the bias voltage model, the bias voltage model stores the corresponding preset threshold value and a corresponding relationship between the comparison result and the fault identification result; for the output signal model, the output signal model stores the corresponding preset threshold value and a corresponding relationship between the comparison result and the fault identification result; and for the spectrum model, the spectrum model stores the corresponding preset threshold value and a corresponding relationship between the comparison result and the fault identification result. It should be noted that the preset threshold values ​​in the bias voltage model, the output signal model, and the spectrum model can be set according to actual needs, and the specific values ​​of the preset threshold values ​​are not specifically limited in this embodiment.

[0468] In the above, after obtaining at least one preset mechanism model and actual sensor data of the target sensor, fault identification of the target sensor can be achieved based on the correspondence between the preset thresholds and comparison results stored in each preset mechanism model and the fault identification results.

[0469] S2104: Perform fault identification on the target sensor according to the fault identification model, at least one preset mechanism model, and actual sensor data to obtain a target fault identification result corresponding to the target sensor.

[0470] In this embodiment, during the operation of the target sensor, both the fault identification model and at least one preset mechanism model can identify faults in the target sensor. Combining the fault identification model and at least one preset mechanism model ultimately enables target sensor fault identification, avoiding the drawbacks of periodic manual fault identification of the target sensor.

[0471] The sensor fault identification method provided in this embodiment, based on determining the fault identification model corresponding to the target sensor and obtaining the preset mechanism model corresponding to the pre-set target sensor, combines the fault identification model and the preset mechanism model to perform fault identification on the target sensor during the operation of the target sensor to obtain a final target fault identification result, thereby realizing fault identification of the target sensor, avoiding the disadvantages of manual and regular fault identification of the target sensor, reducing the expenditure of manpower and material resources, and improving the production efficiency of mechanical equipment equipped with the target sensor.

[0472] Referring to FIG22 , FIG22 is a flow chart of another sensor fault identification method provided in an embodiment of the present application. A sensor fault identification method provided in an embodiment of the present application includes the following steps:

[0473] S2201: Acquire a historical sensor data set corresponding to the target sensor to be identified.

[0474] In this embodiment, step S2201 is consistent with step S2101. For details, please refer to the above-mentioned step S2101, which will not be described in detail in this embodiment.

[0475] S2202: Determine a fault identification model corresponding to the target sensor based on the historical sensor data set, where the fault identification model is used to identify whether the target sensor fails due to environmental interference and the type of environmental fault corresponding to the target sensor when the target sensor fails due to environmental interference.

[0476] In this embodiment, the fault identification model is used only to identify target sensor faults caused by environmental interference and to determine the corresponding environmental fault type for the target sensor when such faults are caused by environmental interference. During target sensor fault identification, if at least one pre-set mechanism model identifies a fault in the target sensor itself, there is no need to run the fault identification model for further target sensor fault identification, thereby reducing resource waste. Environmental interference types include electromagnetic interference, among others.

[0477] S2203: When actual sensor data of the target sensor is obtained, at least one preset mechanism model corresponding to the target sensor is obtained. The preset mechanism model is used to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor itself when the target sensor itself has a fault.

[0478] In this embodiment, step S2203 is consistent with step S2103. For details, please refer to the above-mentioned step S2103, which will not be described in detail in this embodiment.

[0479] S2204: For each preset mechanism model in the at least one preset mechanism model, perform fault identification on the target sensor according to the preset mechanism model and actual sensor data to obtain a first fault identification result corresponding to the target sensor.

[0480] S2205: When all first fault identification results indicate that the target sensor itself does not have a fault, actual sensor data is input into the fault identification model, so that the fault identification model outputs a target fault identification result corresponding to the target sensor.

[0481] With respect to the above-mentioned steps S2204 and S2205, since each preset mechanism model only stores preset thresholds, comparison results, and fault identification results, and the preset mechanism model can identify whether the target sensor itself has a fault and the target sensor's own fault type when the target sensor itself has a fault, only fewer hardware resources are required to realize the fault identification of the target sensor. Therefore, when performing fault identification on the target sensor, the target sensor can be firstly identified through at least one preset mechanism model set by the preset settings, so that when it is determined that the target sensor itself has a fault and the target sensor's own fault type is determined, there is no need to further perform fault identification on the target sensor through the fault identification model. That is to say, when the first fault identification result is that the target sensor itself has a fault and the target sensor's own fault type is determined when the target sensor itself has a fault, the target sensor itself has a fault and the target sensor's own fault type is determined when the target sensor itself has a fault, the target sensor itself has a fault and the target sensor's own fault type is determined when the target sensor itself has a fault as the target fault identification result, and the actual sensor data does not need to be input into the fault identification model for further fault identification of the target sensor. Only when the preset mechanism model identifies that the target sensor itself is not faulty (i.e., all first fault identification results indicate that the target sensor itself is not faulty) is the fault identification model used to further identify the target sensor to determine whether the target sensor fault is caused by environmental interference and, if so, the corresponding environmental fault type. Because the fault identification model requires significant hardware resources to run, this approach can reduce unnecessary resource waste.

[0482] In this embodiment, the preset mechanism model is consistent with the above description, and the preset mechanism model is not described in detail in this embodiment. In the above description, fault identification is performed on the target sensor based on the preset mechanism model and actual sensor data to obtain a first fault identification result corresponding to the target sensor, including:

[0483] When the preset mechanism model is a bias voltage model, the actual bias voltage is compared with a preset threshold value in the bias voltage model to obtain a first comparison result; and a first fault identification result corresponding to the first comparison result is determined based on a correspondence between the comparison result in the bias voltage model and the fault identification result;

[0484] When the preset mechanism model is an output signal model, the actual output signal value is compared with a preset threshold value in the output signal model to obtain a second comparison result; and the first fault identification result corresponding to the second comparison result is determined based on a correspondence between the comparison result in the output signal model and the fault identification result;

[0485] When the preset mechanism model is a spectrum model, a ski slope factor is determined based on the actual spectrum; the ski slope factor is compared with a preset threshold in the spectrum model to obtain a third comparison result; and based on the correspondence between the comparison result in the spectrum model and the fault identification result, a first fault identification result corresponding to the third comparison result is determined.

[0486] Specifically, the bias voltage model may store multiple preset thresholds. For each preset threshold, the bias voltage model also stores a corresponding relationship between the comparison result and the fault identification result corresponding to each preset threshold. For example, if the preset thresholds stored in the bias voltage model include a first preset threshold and a second preset threshold, the bias voltage model stores a corresponding relationship between the comparison result and the fault identification result corresponding to the first preset threshold, and further stores a corresponding relationship between the comparison result and the fault identification result corresponding to the second preset threshold. Since bias voltage is a common characteristic of sensors, the bias voltage within a sensor is typically set to a preset level. For example, in accelerometers, the bias voltage is typically set to 12V. Regardless of how the supply voltage changes, due to the sensor's internal characteristics, the bias voltage should always fluctuate within a specified range. If it exceeds this range, for example, when the bias voltage equals the supply voltage, as shown in Figure 24, this indicates that the sensor is disconnected or reversely powered, and the connector or cable needs to be checked. For another example, when the bias voltage is zero volts, it is generally considered to be a short circuit within the sensor. Therefore, based on the internal characteristic of the sensor's bias voltage, and based on the preset mechanism model corresponding to the preset bias voltage, it is possible to determine the actual bias voltage of the target sensor to identify the fault of the target sensor. It should be noted that the preset threshold stored in the preset mechanism model corresponding to the bias voltage and the corresponding relationship between the comparison result and the fault identification result can be set according to actual needs, and are not specifically limited in this embodiment. When the actual bias voltage is determined by the preset threshold stored in the preset mechanism model corresponding to the bias voltage to identify the fault of the target sensor, the fault identification result obtained includes not only whether the target sensor itself has a fault, but also the type of fault corresponding to the target sensor itself (i.e., disconnection fault, short circuit fault, etc.) when the target sensor itself has a fault.

[0487] The output signal typically includes an output voltage signal and an output current signal. The specific form of the output signal can be set according to actual needs and is not specifically limited in this embodiment. The output signal model can also store multiple preset thresholds. For each preset threshold, the output signal model also stores a corresponding relationship between the comparison result and the fault identification result. For example, when the preset thresholds stored in the output signal model include a third preset threshold and a fourth preset threshold, the output signal model stores a corresponding relationship between the comparison result and the fault identification result corresponding to the third preset threshold, and also stores a corresponding relationship between the comparison result and the fault identification result corresponding to the fourth preset threshold. Because the sensor has a characteristic output current signal or output voltage signal, for example, the output current signal of a pressure sensor can be converted into a desired pressure signal or temperature signal through a specific conversion. The internal output current of the pressure sensor is 4 to 20 mA. When the output current is less than a certain preset current (e.g., 1.185 mA), the pressure sensor is short-circuited and the circuit needs to be checked. When the output current is greater than a certain preset current (e.g., 22.81 mA), the pressure sensor is faulty and the sensor needs to be checked and replaced. It should be noted that the preset threshold values ​​stored in the preset mechanism model corresponding to the output signal, the corresponding relationship between the comparison result and the fault identification result can be set according to actual needs and are not specifically limited in this embodiment. When the actual output signal is judged using the preset threshold values ​​stored in the preset mechanism model corresponding to the output signal to achieve fault identification of the target sensor, the resulting fault identification result includes not only whether the target sensor itself has a fault, but also, if the target sensor itself has a fault, the corresponding fault type of the target sensor itself (i.e., a circuit breaker fault, etc.).

[0488] The actual spectrum of the target sensor is actually the FFT spectrum, which can quickly indicate signal quality. The spectrum model can also store multiple preset thresholds. For each preset threshold, the spectrum model also stores a corresponding relationship between the comparison result and the fault identification result. For example, the spectrum model may store a fifth preset threshold corresponding to frequency and a sixth preset threshold corresponding to time. The fifth preset threshold may be zero, and the spectrum model may store a corresponding relationship between the comparison result and the fault identification result. When both time and frequency in the spectrum are zero, a sensor disconnection fault is indicated. As shown in Figure 26, the lowest frequency line in the FFT spectrum has an unexpectedly high (usually the highest) amplitude. This amplitude typically decreases with increasing frequency, resulting in a ski slope profile. The presence of a large ski slope profile indicates deformation caused by sensor overload. Therefore, based on these characteristics, corresponding preset thresholds can be set to identify whether the corresponding sensor fault type is deformation caused by overload. Specifically, the spectrum model can store preset thresholds corresponding to multiple ski slope factors, as well as corresponding relationships between comparison results corresponding to different preset thresholds and fault identification results, such as a sixth preset threshold, a seventh preset threshold, and an eighth preset threshold. Different preset thresholds represent different levels of deformation caused by overload. Based on the actual spectrum, the ski slope factor can be determined. The final fault identification result is then determined based on the corresponding relationships between the ski slope factor and the preset thresholds corresponding to the multiple ski slope factors stored in the spectrum model, as well as the corresponding relationships between comparison results corresponding to different preset thresholds and fault identification results.

[0489] In the above, the ski slope factor is determined based on the actual spectrum, specifically including:

[0490] Obtain multiple preset frequency intervals corresponding to the target sensor;

[0491] For each preset frequency interval, determine the passing frequency within the preset frequency interval based on the actual spectrum;

[0492] Based on all pass frequencies, determine the ski slope factor.

[0493] In this embodiment, since different sensor types result in different sensor characteristics and thus different preset frequency intervals, multiple preset frequency intervals corresponding to the target sensor can be set according to actual needs. The ski slope factor can be determined based on all passing frequencies using the following formula:

[0494] In the above formula, ski_factor represents the ski slope factor, sp1_rms represents the minimum preset frequency interval; sp2_rms represents the second minimum preset frequency interval; spi_rms represents the maximum preset frequency interval, and i represents the number of preset frequency intervals.

[0495] The sensor fault identification method provided in this embodiment, based on determining the fault identification model corresponding to the target sensor and obtaining the preset mechanism model corresponding to the pre-set target sensor, combines the fault identification model and the preset mechanism model to perform fault identification on the target sensor during the operation of the target sensor to obtain a final target fault identification result, thereby realizing fault identification of the target sensor, avoiding the disadvantages of manual and regular fault identification of the target sensor, reducing the expenditure of manpower and material resources, and improving the production efficiency of mechanical equipment equipped with the target sensor.

[0496] Referring to FIG23 , FIG23 is another sensor fault identification method provided by an embodiment of the present application. The sensor fault identification method provided by this embodiment includes the following methods:

[0497] S2301: Acquire a historical sensor data set corresponding to the target sensor to be identified.

[0498] In this embodiment, step S2301 is consistent with step S2101. For details, please refer to the above description of step S2101, which will not be repeated in this embodiment.

[0499] S2302: Determine the fault identification model corresponding to the target sensor based on the historical sensor data set. The fault identification model is used to identify whether the target sensor itself has a fault, the type of fault corresponding to the target sensor itself when the target sensor itself has a fault, whether the target sensor is faulted by environmental interference, and the type of environmental fault corresponding to the target sensor when the target sensor is faulted by environmental interference.

[0500] In this embodiment, the fault identification model can be used to identify faults in the target sensor caused by environmental interference and the type of environmental fault corresponding to the target sensor when the target sensor fails due to environmental interference. It can also be used to identify faults in the target sensor itself and the type of fault corresponding to the target sensor itself when the target sensor fails due to environmental interference, so as to identify faults in the target sensor. If at least one preset mechanism model is used to identify that the target sensor itself has a fault, the fault identification model can be used to identify whether the target sensor itself has a fault, and the final fault identification result can be determined by combining the identification results of the fault identification model and the preset mechanism model, thereby improving the accuracy of sensor fault identification. Moreover, when the preset mechanism model does not identify that the target sensor itself has a fault, the fault identification model can also be used to identify the fault of the target sensor caused by environmental interference to achieve fault identification of the target sensor.

[0501] S2303: When actual sensor data of the target sensor is obtained, at least one preset mechanism model corresponding to the target sensor is obtained. The preset mechanism model is used to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor itself when the target sensor itself has a fault.

[0502] S2304: For each preset mechanism model in the at least one preset mechanism model, perform fault identification on the target sensor data according to the preset mechanism model and actual sensor data to obtain a first fault identification result corresponding to the target sensor.

[0503] Regarding the above-mentioned steps S2303 and S2304, step S2303 is consistent with the above-mentioned step S2203, and step S2304 is consistent with the above-mentioned step S2204. For details, please refer to steps S2203 and S2204, which will not be described in detail in this embodiment.

[0504] S2305: Input the actual sensor data into the fault recognition model, so that the fault recognition model outputs a second fault recognition result corresponding to the target sensor.

[0505] In this embodiment, during the execution of step S2304, step S2305 is executed synchronously to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor when the target sensor itself has a fault, and to identify whether the target sensor has a fault caused by environmental interference and the corresponding environmental fault type of the target sensor when the target sensor has a fault caused by environmental interference.

[0506] S2306: Determine a target fault identification result corresponding to the target sensor based on all the first fault identification results and the second fault identification results.

[0507] In this embodiment, after all the first fault identification results and the second fault identification results are obtained, all the first fault identification results and the second fault identification results may be combined to obtain a final target fault identification result corresponding to the target sensor.

[0508] Specifically, determining a target fault identification result corresponding to a target sensor according to all first fault identification results and second fault identification results includes:

[0509] determining whether there is a fault identification result in the second fault identification result that is consistent with all the first fault identification results;

[0510] When there is a fault identification result consistent with all the first fault identification results in the second fault identification results, determining the second fault identification result as the target fault identification result corresponding to the target sensor;

[0511] When there is a fault identification result in the second fault identification result that is inconsistent with at least one first fault identification result, generating an alarm prompt information according to the inconsistent at least one first fault identification result;

[0512] Push the alarm prompt information to the target terminal corresponding to the target sensor.

[0513] In this embodiment, the target terminal may be a mobile phone, tablet, or the like. The specific form of the target terminal can be selected based on actual needs and is not limited in this embodiment. When a fault identification result among the second fault identification results is consistent with all first fault identification results, it indicates that the identification results of the fault identification model and the preset mechanism model are consistent. Since the fault identification model can also identify faults caused by environmental interference, to improve the accuracy of sensor fault identification, the second fault identification result is determined as the target fault identification result. When a fault identification result among the second fault identification results is inconsistent with at least one first fault identification result, it may be due to inaccurate preset threshold settings stored in a preset mechanism model or inaccurate training parameters in the fault identification model. To further improve the accuracy of sensor fault identification, an alarm message can be generated based on the at least one inconsistent first fault identification result and pushed to the target terminal corresponding to the target sensor to update the preset mechanism model or fault identification model. The specific form of the alarm message can be set based on actual needs and is not limited in this embodiment.

[0514] The sensor fault identification method provided in this embodiment, based on determining the fault identification model corresponding to the target sensor and obtaining the preset mechanism model corresponding to the pre-set target sensor, combines the fault identification model and the preset mechanism model to perform fault identification on the target sensor during the operation of the target sensor to obtain a final target fault identification result, thereby realizing fault identification of the target sensor, avoiding the disadvantages of manual and regular fault identification of the target sensor, reducing the expenditure of manpower and material resources, and improving the production efficiency of mechanical equipment equipped with the target sensor.

[0515] Referring to Figure 27, Figure 27 is a schematic diagram of the structure of a sensor fault identification device provided in an embodiment of the present application. A sensor fault identification device provided in an embodiment of the present application includes: an acquisition module 10, a determination module 20, and an identification model 30. The acquisition module 10 is used to acquire a historical sensor data set corresponding to a target sensor to be identified; the determination module 20 is used to determine a target fault identification model corresponding to the target sensor based on the historical sensor data set; the acquisition module 10 is also used to acquire at least one preset mechanism model corresponding to the target sensor when actual sensor data of the target sensor is acquired, the preset mechanism model being used to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor itself when the target sensor itself has a fault; the identification module 30 is used to perform fault identification on the target sensor based on the fault identification model, at least one preset mechanism model, and the actual sensor data to obtain a target fault identification result corresponding to the target sensor.

[0516] In this embodiment, the fault identification model is used to identify whether the target sensor fails due to environmental interference and the type of environmental fault corresponding to the target sensor when the target sensor fails due to environmental interference.

[0517] In this embodiment, the identification module 30 is further configured to:

[0518] For each of the at least one preset mechanism model, performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor;

[0519] When all the first fault identification results indicate that the target sensor itself does not have a fault, the actual sensor data is input into the fault identification model, so that the fault identification model outputs a target fault identification result corresponding to the target sensor.

[0520] In this embodiment, the fault identification model is used to identify whether the target sensor itself has a fault, the type of fault corresponding to the target sensor itself when the target sensor itself has a fault, whether the target sensor has a fault caused by environmental interference, and the type of environmental fault corresponding to the target sensor when the target sensor has a fault caused by environmental interference.

[0521] In this embodiment, the identification module 30 is further configured to:

[0522] For each of the at least one preset mechanism model, performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor;

[0523] Inputting the actual sensor data into the fault identification model so that the fault identification model outputs a second fault identification result corresponding to the target sensor;

[0524] A target fault identification result corresponding to the target sensor is determined according to all the first fault identification results and the second fault identification results.

[0525] In this embodiment, the identification module 30 is further configured to:

[0526] determining whether there is a fault identification result in the second fault identification results that is consistent with all the first fault identification results;

[0527] When there is a fault identification result in the second fault identification results that is consistent with all the first fault identification results, determining the second fault identification result as the target fault identification result corresponding to the target sensor;

[0528] When there is a fault identification result in the second fault identification result that is inconsistent with at least one of the first fault identification results, generating an alarm prompt information according to the inconsistent at least one first fault identification result;

[0529] The alarm prompt information is pushed to the target terminal corresponding to the target sensor.

[0530] In this embodiment, the determining module 20 is further configured to:

[0531] For each historical sensor data in the historical sensor data set, performing fault classification and labeling on the historical sensor data to obtain training sample data corresponding to the historical sensor data;

[0532] The preset classification model is trained according to the training sample data corresponding to all the historical sensor data in the historical sensor data set to obtain a fault recognition model corresponding to the target sensor.

[0533] In this embodiment, at least one of the preset mechanism models includes a bias voltage model, an output signal model and a spectrum model, and the actual sensor data includes an actual bias voltage, an actual output signal value and an actual spectrum. Each of the preset mechanism models stores a preset threshold and a correspondence between the comparison result and the fault identification result, and the comparison result is a comparison result between the actual sensor data and the preset threshold.

[0534] In this embodiment, the identification module 30 is further configured to:

[0535] When the preset mechanism model is the bias voltage model, comparing the actual bias voltage with a preset threshold in the bias voltage model to obtain a first comparison result; and determining a first fault identification result corresponding to the first comparison result based on a correspondence between the comparison result in the bias voltage model and the fault identification result;

[0536] When the preset mechanism model is the output signal model, comparing the actual output signal value with a preset threshold in the output signal model to obtain a second comparison result; and determining a first fault identification result corresponding to the second comparison result based on a correspondence between the comparison result in the output signal model and the fault identification result;

[0537] When the preset mechanism model is the spectrum model, a ski slope factor is determined based on the actual spectrum; the ski slope factor is compared with a preset threshold in the spectrum model to obtain a third comparison result; and based on a correspondence between the comparison result in the spectrum model and the fault identification result, a first fault identification result corresponding to the third comparison result is determined.

[0538] In this embodiment, the recognition model 30 is further used to:

[0539] Acquire multiple preset frequency intervals corresponding to the target sensor;

[0540] For each of the preset frequency intervals, determining a passing frequency within the preset frequency interval according to the actual frequency spectrum;

[0541] From all said pass frequencies, a ski slope factor is determined.

[0542] The sensor fault identification device provided in this embodiment, based on determining the fault identification model corresponding to the target sensor and obtaining the preset mechanism model corresponding to the pre-set target sensor, combines the fault identification model and the preset mechanism model to perform fault identification on the target sensor during the operation of the target sensor to obtain a final target fault identification result, thereby realizing fault identification of the target sensor, avoiding the disadvantages of manual and regular fault identification of the target sensor, reducing the expenditure of manpower and material resources, and improving the production efficiency of mechanical equipment equipped with the target sensor.

[0543] Figure 28 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 2800 shown in Figure 28 includes: at least one processor 2801, a memory 2802, at least one network interface 2804 and other user interfaces 2803. The various components in the electronic device 2800 are coupled together via a bus system 2805. It will be understood that the bus system 2805 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 2805 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, various buses are labeled as bus system 2805 in Figure 28.

[0544] The user interface 2803 may include a display, a keyboard, or a pointing device (eg, a mouse, a trackball, a touchpad, or a touch screen).

[0545] It is understood that the memory 2802 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DRRAM). Memory 2802 described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0546] In some embodiments, the memory 2802 stores the following elements, executable units, or data structures, or a subset thereof, or an extended set thereof: an operating system 28021 and application programs 28022 .

[0547] Among them, operating system 28021 includes various system programs, such as the framework layer, core library layer, and driver layer, which are used to implement various basic services and handle hardware-based tasks. Application 28022 includes various application programs, such as media players and browsers, which are used to implement various application services. Programs that implement the methods of the embodiments of the present application can be included in application 28022.

[0548] In an embodiment of the present application, by calling the program or instruction stored in the memory 2802, specifically, the program or instruction stored in the application 28022, the processor 2801 is used to execute the method steps provided by each method embodiment, for example, including: obtaining a historical sensor data set corresponding to the target sensor to be identified; determining a fault identification model corresponding to the target sensor based on the historical sensor data set; when obtaining the actual sensor data of the target sensor, obtaining at least one preset mechanism model corresponding to the target sensor, the preset mechanism model is used to identify whether the target sensor itself has a fault and the target sensor's own fault type when the target sensor itself has a fault; based on the fault identification model, at least one preset mechanism model and the actual sensor data, the target sensor is fault identified to obtain a target fault identification result corresponding to the target sensor.

[0549] The methods disclosed in the above embodiments of the present application can be applied to or implemented by processor 2801. Processor 2801 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 2801 or by software instructions. The above processor 2801 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software units in the decoding processor. The software units can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory 2802 , and the processor 2801 reads the information in the memory 2802 and completes the steps of the above method in combination with its hardware.

[0550] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or a combination thereof.

[0551] For software implementation, the technology described herein can be implemented by a unit that performs the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0552] The electronic device provided in this embodiment can be an electronic device as shown in Figure 28, which can execute all steps of the fault prediction method in Figure 1, all steps of the fault monitoring method in Figure 6, Figures 13 to 15, and all steps of the dynamic equipment fault monitoring method in Figure 18, or all steps of the sensor fault identification method in Figures 21 to 23, thereby achieving the technical effects of the methods shown in the above figures. Please refer to the relevant descriptions in the above figures for details. For the sake of brevity, they are not repeated here.

[0553] The present application also provides a storage medium (computer-readable storage medium). The storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; and the memory may also include a combination of the aforementioned types of memory.

[0554] When one or more programs in the storage medium can be executed by one or more processors to implement the above-mentioned fault prediction method, fault monitoring method, dynamic equipment fault monitoring method, or sensor fault identification method executed on the sensor fault identification device side.

[0555] The processor is used to execute a sensor fault identification program stored in a memory to implement the following steps of a sensor fault identification method executed on the sensor fault identification device side: obtaining a historical sensor data set corresponding to a target sensor to be identified; determining a fault identification model corresponding to the target sensor based on the historical sensor data set; when obtaining actual sensor data of the target sensor, obtaining at least one preset mechanism model corresponding to the target sensor, the preset mechanism model being used to identify whether the target sensor itself has a fault and the type of fault corresponding to the target sensor itself when the target sensor itself has a fault; performing fault identification on the target sensor based on the fault identification model, at least one preset mechanism model and the actual sensor data to obtain a target fault identification result corresponding to the target sensor.

[0556] Professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0557] It should be noted that references in this specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," and the like indicate that the described embodiments may include a particular feature, structure, or characteristic, but not necessarily every embodiment includes that particular feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same embodiment. Furthermore, when a particular feature, structure, or characteristic is described in conjunction with an embodiment, it is within the knowledge of those skilled in the art to implement such feature, structure, or characteristic in conjunction with other embodiments, whether explicitly described or not.

[0558] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0559] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A fault prediction method, characterized in that: include: Acquire a vibration signal collected by a sensor installed on a target component of a target device, and clean and segment the vibration signal to obtain a target vibration signal; Input the target vibration signal into a target abnormality judgment model to obtain an output result, and determine whether the target vibration signal is abnormal according to the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is obtained by training the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder; When the abnormal judgment result indicates that the target vibration signal is abnormal, the target vibration signal is input into a target fault prediction model to obtain a fault prediction result, wherein the target fault prediction model is obtained by combining an encoder of the target abnormal judgment model and a newly added fully connected layer, and the target fault prediction model is obtained by training with multiple fault types and vibration signals of each fault type as samples.

2. The method according to claim 1, characterized in that The vibration signal is cleaned and segmented to obtain a target vibration signal including: Segmenting the vibration signal according to the time dimension to obtain a plurality of first vibration signals, wherein each first vibration signal includes vibration signals of the same duration, and the duration includes a plurality of continuous vibration cycles; Determine in sequence whether there is an error signal in each first vibration signal, wherein the error signal indicates that the sensor is operating abnormally; In the case of a first vibration signal having an error signal, deleting the first vibration signal having the error signal from the multiple first vibration signals to obtain multiple second vibration signals; Each second vibration signal is sequentially input into a preset filter to obtain a plurality of target vibration signals.

3. The method according to claim 1, characterized in that The target abnormality judgment model is trained in the following way: Acquire a sample vibration signal set of the target component, wherein the sample vibration signal set includes a plurality of sample vibration signals generated by the target component under normal operation; Inputting the sample vibration signals in the sample vibration signal set into the initial abnormality judgment model, and processing to obtain an output signal corresponding to each sample vibration signal, wherein the initial abnormality judgment model is composed of an encoder and a decoder, the encoder is used to extract features of the sample vibration signals and reduce data dimensions, and the decoder is used to restore data dimensions and restore data features; Determine a difference value between each output signal and a corresponding sample vibration signal to obtain a plurality of difference values, and determine whether there is a difference value greater than a preset threshold value among the plurality of difference values; When there is a difference value greater than the preset threshold value among the multiple difference values, changing the neuron connection weights in the initial abnormality judgment model, and retraining the changed initial abnormality judgment model until there is no difference value greater than the preset threshold value among the multiple difference values; In the case that there is no difference value greater than the preset threshold value among the multiple difference values, the target abnormality judgment model is obtained.

4. The method according to claim 3, characterized in that Determining whether the target vibration signal is abnormal according to the output result, and obtaining the abnormality judgment result includes: Determine a difference value between the output result and the target vibration signal to obtain a target difference value, and determine whether the target difference value is greater than the preset threshold value; When the target difference value is greater than the preset threshold, determining that the target vibration signal is abnormal; When the target difference value is less than or equal to the preset threshold, it is determined that there is no abnormality in the target vibration signal.

5. The method according to claim 1, characterized in that The target fault prediction model is trained in the following way: Acquire multiple fault types and historical vibration signals under each fault type to obtain multiple groups of historical vibration signals; Add a label to each historical vibration signal in each group of historical vibration signals according to the fault type to obtain multiple groups of updated historical vibration signals; The multiple groups of updated historical vibration signals are used as samples to train the initial fault prediction model to obtain the target fault prediction model.

6. The method according to claim 1, characterized in that After obtaining the fault prediction result, the method further includes: Acquire a historical abnormal vibration signal collected by the sensor when an abnormality occurs in the target component; Determining characteristic information of the vibration signal of the target component according to the historical abnormal vibration signal, obtaining a plurality of characteristic information, and acquiring a characteristic value corresponding to each characteristic information; Determine a judgment standard for each feature information according to the feature value corresponding to each feature information, and obtain multiple judgment standards; Obtaining a vibration data value corresponding to each characteristic information in the target vibration signal, and performing abnormality judgment on the vibration data value using a judgment standard with the same characteristic information to obtain a judgment result; When the judgment result indicates that there is no abnormality in the target vibration signal, it is determined that there is an abnormality in the target abnormality judgment model, and an alarm message is issued, wherein the alarm message indicates that there is an abnormality in the target abnormality judgment model.

7. The method according to claim 6, characterized in that The vibration data values ​​are judged to be abnormal using the same judgment criteria as the characteristic information, and the judgment results include: Acquire the vibration data value under each feature information and the judgment standard under the same feature information, and determine whether the vibration data value meets the corresponding judgment standard to obtain multiple sub-judgment results, wherein the sub-judgment results are used to characterize whether the vibration data value meets the corresponding judgment standard; If there is a sub-judgment result that does not meet the judgment criterion among the multiple sub-judgment results, determining that the judgment result is that the target vibration signal is abnormal; In the case that there is no sub-judgment result that does not meet the judgment criterion among the multiple sub-judgment results, it is determined that the judgment result is that there is no abnormality in the target vibration signal.

8. A fault prediction device, characterized in that: include: A first acquisition unit is used to acquire a vibration signal collected by a sensor installed on a target component of a target device, and clean and segment the vibration signal to obtain a target vibration signal; A first determination unit is used to input the target vibration signal into a target abnormality judgment model to obtain an output result, and determine whether the target vibration signal is abnormal according to the output result to obtain an abnormality judgment result, wherein the target abnormality judgment model is obtained by training the sample vibration signal of the target component, and the target abnormality judgment model is composed of an encoder and a decoder; A prediction unit is used to input the target vibration signal into a target fault prediction model to obtain a fault prediction result when the abnormal judgment result indicates that the target vibration signal is abnormal, wherein the target fault prediction model is obtained by combining an encoder of the target abnormal judgment model and a newly added fully connected layer, and the target fault prediction model is obtained by training with multiple fault types and vibration signals of each fault type as samples.

9. A fault monitoring method, characterized in that: include: Acquire an initial vibration signal collected by a preset sensor, wherein the preset sensor is arranged at a target position of a target device, and the preset sensor is used to collect the vibration signal at the target position; Filtering the initial vibration signal by a zero-phase filtering method to obtain an initial high-frequency signal; Acquire a reference key phase signal of a reference device, and determine a mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and change the abscissa of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device; Acquire standard data of the target device, detect the target angular vibration signal through the standard data to obtain a detection result, and determine the fault state of the target device according to the detection result, wherein the standard data is data in the vibration signal of the target device during normal operation.

10. The method according to claim 9, characterized in that The initial vibration signal is filtered by a zero-phase filtering method to obtain an initial high-frequency signal including: Obtaining a preset high-pass filter, and inputting the initial vibration signal into the preset high-pass filter to obtain a first time domain signal; Flipping the first time domain signal around the center line of the X-axis, and inputting the flipped first time domain signal into the preset high-pass filter to obtain a second time domain signal; The second time domain signal is flipped around the center line of the X-axis to obtain the initial high-frequency signal.

11. The method according to claim 9, characterized in that After obtaining the reference key phase signal of the reference device, the method further includes: Determine the rotation period of the reference device according to the rotation speed of the reference device, and intercept the signal within a preset time period in the reference key phase signal to obtain an updated reference key phase signal, wherein the preset time period includes a plurality of continuous rotation periods; intercepting the signal within the preset time period from the initial high-frequency signal to obtain an updated initial high-frequency signal; Determining the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device, and changing the abscissa of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship, and obtaining the target angular domain vibration signal includes: The mapping relationship between the time in the updated reference key phase signal and the rotation angle of the reference device is determined according to the rotation speed of the reference device, and the updated initial high-frequency signal is converted according to the mapping relationship to obtain a target angular domain vibration signal.

12. The method according to claim 11, characterized in that The reference device is provided with a code disk, the rotation speed of the reference device is the same as the rotation speed of the code disk, and determining the mapping relationship between the time in the reference key phase signal and the rotation angle of the reference device according to the rotation speed of the reference device includes: Determine the number of revolutions of the code disk within the preset time period according to the rotation speed to obtain a preset number of revolutions; The rotation angle corresponding to the preset number of revolutions is determined, and the duration required for each rotation of 1 degree is determined according to the rotation angle and the preset time period to obtain the mapping relationship.

13. The method according to claim 9, characterized in that According to the mapping relationship, the abscissa of the initial high-frequency signal is changed from a time value to a corresponding angle value, and the target angular domain vibration signal is obtained, including: Converting the time value in the time axis of the initial high-frequency signal into a preset angle value according to the mapping relationship to obtain an initial angular domain vibration signal; Determine whether each preset angle value has a corresponding amplitude value in the initial angular domain vibration signal, and if the target preset angle value does not have a corresponding amplitude value, obtain multiple preset angle values ​​adjacent to the target preset angle value and the amplitude value of each preset angle value, use interpolation to calculate the amplitude value of the target preset angle value according to the amplitude value of each preset angle value, and add the amplitude to the corresponding position in the initial angular domain vibration signal to obtain the target angular domain vibration signal.

14. The method according to claim 9, characterized in that Detecting the target angular vibration signal by using the standard data to obtain a detection result, and determining the fault state of the target device according to the detection result includes: Acquire a calibration angle and a standard amplitude at the calibration angle from the standard data; Determining the amplitude at the calibration angle in the target angle domain vibration signal to obtain a target amplitude, and determining whether the target amplitude is greater than the standard amplitude; When the target amplitude is greater than the standard amplitude, determining that the target device has a fault; When the target amplitude is less than or equal to the standard amplitude, it is determined that the target device has no fault.

15. The method according to claim 14, characterized in that Determining the amplitude at the calibration angle in the target angle domain vibration signal to obtain the target amplitude includes: Acquire the angle range in the target angular domain vibration signal, and acquire the number of rotation periods contained in the angle range to obtain the number of periods, wherein the angle range is greater than 360 degrees; Calculate the angle value corresponding to the calibration angle in each cycle to obtain multiple preset angles; The target angle corresponding to each preset angle in the target angle domain vibration signal is determined to obtain a plurality of target angles, and the amplitude at each target angle is acquired to obtain a plurality of target amplitudes.

16. A fault monitoring device, characterized in that: include: An acquisition unit, configured to acquire an initial vibration signal acquired by a preset sensor, wherein the preset sensor is disposed at a target position of a target device, and the preset sensor is configured to acquire a vibration signal at the target position; A filtering unit, used for filtering the initial vibration signal by a zero-phase filtering method to obtain an initial high-frequency signal; a changing unit, configured to obtain a reference key phase signal of a reference device, and determine a mapping relationship between a time in the reference key phase signal and a rotation angle of the reference device according to a rotation speed of the reference device, and change a horizontal coordinate of the initial high-frequency signal from a time value to a corresponding angle value according to the mapping relationship, so as to obtain a target angular domain vibration signal, wherein the rotation speed of the reference device is the same as the rotation speed of the target device; The first determination unit is used to obtain standard data of the target device, detect the target angular vibration signal through the standard data to obtain a detection result, and determine the fault state of the target device according to the detection result, wherein the standard data is data in the vibration signal of the target device during normal operation.

17. A method for monitoring a fault of a moving device, characterized in that: Applied to a dynamic equipment fault monitoring system, the method comprises: Determine multiple test points of the target device and the parameter matching rules for each test point; For each test point, matching the target parameter data of the test point according to the corresponding parameter matching rule; Determining the fault symptom quantity of the test point according to the target parameter data; Determine the operating status of each of the test points based on the fault symptom quantity corresponding to each of the test points; When it is determined that the target moving device has a fault through the operating status of each test point, the fault point and fault type of the target moving device are determined according to the operating status of each test point.

18. The method according to claim 17, characterized in that The step of matching target parameter data of each test point according to a corresponding parameter matching rule includes: For each test point, target parameter data corresponding to the test point is obtained from a preset database according to a parameter matching rule corresponding to the test point, wherein the database pre-stores the following parameter data of the target moving device: static attribute data of the target moving device, maintenance policy data of the target moving device, operation data of the target moving device, and sensor signal data of the target moving device collected by a sensor; The method further comprises: Get the preset sensor collection rules; According to the sensor acquisition rule, the corresponding sensor is controlled to acquire the corresponding sensor signal data.

19. The method according to claim 17, characterized in that Determine the parameter matching rules for each test point, including: Obtaining position information of each test point on the target moving device; Determining a parameter matching rule for each of the test points according to the location information; Determining the fault symptom quantity of the test point according to the target parameter data includes: Determine, according to the location information of the test point, a characteristic value calculation rule and a fault symptom quantity calculation rule corresponding to the test point; Calculate the target parameter data according to the characteristic value calculation rule to obtain at least one characteristic value corresponding to the test point; The target parameter data and / or the characteristic value are combined according to the fault symptom calculation rule to generate the fault symptom of the test point.

20. The method according to claim 17, characterized in that The determining the operating status of each of the test points based on the fault symptom quantity corresponding to each of the test points includes: Obtaining the highest speed and the lowest speed of the target moving device in multiple operation cycles; Determine a speed difference between the maximum speed and the minimum speed, and determine whether the speed difference is less than a preset difference threshold; When it is determined that the rotation speed difference is less than the difference threshold, the operating state of each of the test points is determined based on a preset dynamic warning model and a fault symptom quantity corresponding to each of the test points.

21. The method according to claim 20, characterized in that The step of determining the operating status of each test point based on a preset dynamic warning model and a fault sign quantity corresponding to each test point includes: Acquire historical operation data of the target dynamic equipment in a preset historical time period, and configuration parameters of a preset dynamic early warning model; Input the historical operation data, the fault symptom quantity corresponding to each of the test points, and the configuration parameters into the dynamic early warning model to obtain the early warning result corresponding to each fault symptom quantity output by the dynamic early warning model; Based on the early warning result corresponding to each fault symptom quantity, the operating state of each test point is determined, wherein the operating state includes a fault state and a non-fault state.

22. The method according to claim 21, characterized in that The fault sign quantity includes a sensor abnormality sign quantity, and the determining the operating state of each test point based on the early warning result corresponding to each fault sign quantity includes: Determining whether the warning result corresponding to the abnormal sign quantity of the sensor is a fault warning; If it is determined that the warning result corresponding to the sensor abnormality sign quantity is a fault warning, the alarm information of the sensor abnormality is output and the process ends; If it is determined that the warning result corresponding to the abnormal sign quantity of the sensor is a non-fault warning, obtaining sensor signal data collected by the sensor; Inputting the sensor signal data into a preset sensor fault identification model to obtain a sensor abnormality identification result output by the sensor fault identification model; If the sensor abnormality identification result indicates that the sensor is abnormal, determining that the operating state of each test point is a non-fault state; If the sensor abnormality identification result indicates that the sensor is normal, the warning result level of each fault symptom quantity corresponding to each test point is determined, and the fault state corresponding to the warning result with the highest warning result level is determined as the operating state of the test point.

23. The method according to claim 22, characterized in that Determining that the target moving device has a fault by the operating status of each test point includes: In the case where it is determined that the operating state of any test point is a fault state, determining that the target moving device has a fault; Determining the fault point and fault type of the target moving device according to the operating status of each test point includes: Determine whether the operating status of each test point is a fault state; Determine the test point whose operating state is a fault state as the initial fault point; For each of the initial fault points, the early warning result corresponding to the initial fault point is input into a preset fault detection model to obtain the fault point and fault type of the target moving equipment output by the fault detection model.

24. The method according to claim 23, characterized in that The fault detection model includes a typical fault classification model and a fault analysis model. The typical fault classification model is used to determine whether the fault type of the target moving device is a typical fault. The fault analysis model is used to analyze the fault point and basic fault type of the target moving device. The early warning result corresponding to the initial fault point is input into the preset fault detection model to obtain the fault point and fault type of the target moving device output by the fault detection model, including: Inputting the early warning result corresponding to the initial fault point into the typical fault classification model and the fault analysis model respectively, to obtain the typical fault classification result output by the typical fault classification model and the fault analysis result output by the fault analysis model; The typical fault classification result and the fault analysis result are weightedly summed to obtain the fault point and fault type corresponding to the target moving equipment.

25. The method according to claim 24, characterized in that The method further comprises: Determining whether the fault type is a typical fault type; In the case where it is determined that the fault type is a typical fault type, the fault type is matched with a preset typical fault case library to obtain a target solution corresponding to the fault type, wherein the typical fault case library is used to store typical faults and solutions corresponding to each of the typical faults; The target solution is output through a visualization interface.

26. The method according to claim 18, characterized in that The method further comprises: According to preset signal processing rules, signal processing and feature transformation are performed on the parameter data of the target moving device stored in the database to obtain a variety of atlas data; The atlas data is output through a visual interface to analyze the target dynamic device according to the atlas data.

27. The method according to claim 17, characterized in that After determining the fault point and fault type of the target moving equipment, the method further includes: generating a three-dimensional image of the target moving device; Marking the fault point and fault type of the target moving equipment in the three-dimensional image to obtain a target three-dimensional image; The target three-dimensional image is output, and an early warning is issued through a preset early warning method.

28. A moving equipment fault monitoring device, characterized in that: Applied to a dynamic equipment fault monitoring system, the device comprises: A first determination module is used to determine multiple test points of the target mobile device and a parameter matching rule for each test point; A matching module, used for matching target parameter data of each test point according to a corresponding parameter matching rule; A second determination module, configured to determine the fault symptom quantity of the test point according to the target parameter data; A third determination module, configured to determine the operating state of each of the test points based on the fault symptom quantity corresponding to each of the test points; The fourth determination module is used to determine the fault point and fault type of the target moving device according to the running status of each test point when it is determined that the target moving device has a fault through the running status of each test point.

29. A sensor fault identification method, characterized in that: include: Obtain a historical sensor data set corresponding to the target sensor to be identified; Determining a fault identification model corresponding to the target sensor according to the historical sensor data set; When actual sensor data of the target sensor is obtained, at least one preset mechanism model corresponding to the target sensor is obtained, wherein the preset mechanism model is used to identify whether the target sensor itself has a fault and the fault type of the target sensor itself when the target sensor itself has a fault; According to the fault identification model, at least one of the preset mechanism models and the actual sensor data, fault identification is performed on the target sensor to obtain a target fault identification result corresponding to the target sensor.

30. The method according to claim 29, characterized in that The fault identification model is used to identify whether the target sensor fails due to environmental interference and the type of environmental fault corresponding to the target sensor when the target sensor fails due to environmental interference; The performing fault identification on the target sensor according to the fault identification model, at least one of the preset mechanism models and the actual sensor data to obtain a target fault identification result corresponding to the target sensor includes: For each of the at least one preset mechanism model, according to the preset mechanism model and the actual sensor data, performing fault identification on the target sensor to obtain a first fault identification result corresponding to the target sensor; When all the first fault identification results indicate that the target sensor itself does not have a fault, the actual sensor data is input into the fault identification model, so that the fault identification model outputs a target fault identification result corresponding to the target sensor.

31. The method according to claim 29, characterized in that The fault identification model is used to identify whether the target sensor itself has a fault, the type of fault of the target sensor itself when the target sensor itself has a fault, whether the target sensor is faulted by environmental interference, and the type of environmental fault of the target sensor when the target sensor is faulted by environmental interference; The performing fault identification on the target sensor according to the fault identification model, at least one of the preset mechanism models and the actual sensor data to obtain a target fault identification result corresponding to the target sensor includes: For each of the at least one preset mechanism model, according to the preset mechanism model and the actual sensor data, performing fault identification on the target sensor to obtain a first fault identification result corresponding to the target sensor; Inputting the actual sensor data into the fault identification model so that the fault identification model outputs a second fault identification result corresponding to the target sensor; A target fault identification result corresponding to the target sensor is determined according to all the first fault identification results and the second fault identification results.

32. The method according to claim 31, characterized in that The determining, according to all the first fault identification results and the second fault identification results, a target fault identification result corresponding to the target sensor includes: determining whether there is a fault identification result in the second fault identification results that is consistent with all the first fault identification results; When there is a fault identification result consistent with all the first fault identification results in the second fault identification results, determining the second fault identification result as the target fault identification result corresponding to the target sensor; When there is a fault identification result in the second fault identification result that is inconsistent with at least one of the first fault identification results, generating an alarm prompt information according to the inconsistent at least one of the first fault identification results; The alarm prompt information is pushed to the target terminal corresponding to the target sensor.

33. The method according to claim 29, characterized in that The determining, based on the historical sensor data set, a fault identification model corresponding to the target sensor includes: For each historical sensor data in the historical sensor data set, fault classification and labeling are performed on the historical sensor data to obtain training sample data corresponding to the historical sensor data; The preset classification model is trained according to the training sample data corresponding to all the historical sensor data in the historical sensor data set to obtain a fault recognition model corresponding to the target sensor.

34. The method according to claim 30 or 31, characterized in that At least one of the preset mechanism models includes a bias voltage model, an output signal model and a spectrum model, the actual sensor data includes an actual bias voltage, an actual output signal value and an actual spectrum, each of the preset mechanism models stores a preset threshold value and a corresponding relationship between a comparison result and a fault identification result, and the comparison result is a comparison result between the actual sensor data and the preset threshold value; The performing fault identification on the target sensor according to the preset mechanism model and the actual sensor data to obtain a first fault identification result corresponding to the target sensor includes: When the preset mechanism model is the bias voltage model, the actual bias voltage is compared with a preset threshold in the bias voltage model to obtain a first comparison result; and a first fault identification result corresponding to the first comparison result is determined according to a corresponding relationship between the comparison result in the bias voltage model and the fault identification result; When the preset mechanism model is the output signal model, the actual output signal value is compared with a preset threshold value in the output signal model to obtain a second comparison result; according to the corresponding relationship between the comparison result in the output signal model and the fault identification result, a first fault identification result corresponding to the second comparison result is determined; When the preset mechanism model is the spectrum model, a ski slope factor is determined according to the actual spectrum; the ski slope factor is compared with a preset threshold in the spectrum model to obtain a third comparison result; and a first fault identification result corresponding to the third comparison result is determined according to a correspondence between the comparison result in the spectrum model and the fault identification result.

35. The method according to claim 34, characterized in that Determining a ski slope factor according to the actual frequency spectrum includes: Acquire multiple preset frequency intervals corresponding to the target sensor; For each of the preset frequency intervals, determining a passing frequency within the preset frequency interval according to the actual frequency spectrum; Based on all said pass frequencies, a ski slope factor is determined.

36. A sensor fault identification device, characterized in that: include: An acquisition module, used to acquire a historical sensor data set corresponding to a target sensor to be identified; A determination module, configured to determine a target fault identification model corresponding to the target sensor according to the historical sensor data set; The acquisition module is further used to acquire at least one preset mechanism model corresponding to the target sensor when the actual sensor data of the target sensor is acquired, and the preset mechanism model is used to identify whether the target sensor itself has a fault and the corresponding fault type of the target sensor itself when the target sensor itself has a fault; The identification module is used to perform fault identification on the target sensor according to the fault identification model, at least one of the preset mechanism models and the actual sensor data to obtain a target fault identification result corresponding to the target sensor.

37. An electronic device, characterized in that: include: A processor and a memory, wherein the processor is connected to the memory, and the processor is used to execute a fault prediction program, a fault monitoring program, a moving equipment fault monitoring program, or a sensor fault identification program stored in the memory, so as to implement the fault prediction method described in any one of claims 1 to 7, the fault monitoring method described in any one of claims 9 to 15, the moving equipment fault monitoring method described in any one of claims 17 to 27, or the sensor fault identification method described in any one of claims 29 to 36.

38. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the fault prediction method described in any one of claims 1 to 7, the fault monitoring method described in any one of claims 9 to 15, the moving equipment fault monitoring method described in any one of claims 17 to 27, or the sensor fault identification method described in any one of claims 29 to 36.

Images