Fault detection method and device for electromechanical equipment, server and fault detection equipment

Abstract

The embodiment of the invention provides a fault detection method and device for electromechanical equipment, a server and fault detection device.In the fault detection method, the fault detection equipment collects signals of to-be-detected electromechanical equipment, calculates signal features of the signals according to features of a preset model corresponding to the to-be-detected electromechanical equipment, and sends the calculated signal features to the server; and then comparing the signal features with a threshold value corresponding to the features of a preset model. And if the signal characteristics greater than or equal to the threshold value exist, acquiring historical signals of the to-be-detected electromechanical equipment and historical signals of adjacent equipment of the to-be-detected electromechanical equipment, and determining a fault detection result of the to-be-detected electromechanical equipment according to trend information of the historical signals and trend information of the historical signals of the adjacent equipment. Therefore, fault detection can be carried out on the to-be-detected electromechanical equipment through the corresponding preset model, then equipment abnormity can be eliminated before the to-be-detected electromechanical equipment has a functional fault, and the effect of fault prevention is achieved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and in particular to a fault detection method, device, server and fault detection device for electromechanical equipment. Background Art

[0002] There are a large number of electromechanical equipment in industries such as coal mines, ports and / or power. In order to avoid production losses caused by unplanned equipment downtime, manual inspections are usually used to conduct regular inspections of the equipment and perform maintenance actions on wearing parts, such as oiling or backup, before failures or problems occur.

[0003] With the development of information technology, in order to ensure that the frequency (cost) of operation and maintenance is further reduced on the basis of stable operation of electromechanical equipment, many protection measures will be used in the use of electromechanical equipment, such as: stator current overcurrent, stator voltage overvoltage, stator voltage undervoltage, motor negative phase sequence protection and motor differential protection. By setting various relays to monitor the various parameter values ​​of electromechanical equipment, if some parameter values ​​exceed the set value of the electromechanical equipment, the relay will sound an alarm, and if necessary, the operation of the electromechanical equipment will be directly stopped by cutting off the control loop to prevent the further expansion of the scale of production accidents. On the surface, the protection settings of electromechanical equipment seem to be very perfect, but in actual operation, the relay will only sound an alarm when the electromechanical equipment fails to function, and there is no early warning function. Therefore, the relay can only be used as the last line of defense for the stable operation of electromechanical equipment to prevent major or catastrophic failures. However, the sudden power outage or shutdown of many large equipment will cause huge losses. If the effect of equipment failure prevention cannot be achieved and the equipment abnormality cannot be eliminated before the functional failure, then setting more protective relays will be in vain. Summary of the invention

[0004] The embodiments of the present application provide a fault detection method, device, server and fault detection device for electromechanical equipment. The embodiments of the present application also provide a computer-readable storage medium to implement fault detection of the electromechanical equipment to be detected through a preset model corresponding to the electromechanical equipment to be detected, thereby eliminating equipment abnormalities before a functional fault occurs in the electromechanical equipment to be detected, thereby achieving the effect of fault prevention.

[0005] In a first aspect, an embodiment of the present application provides a method for fault detection of electromechanical equipment, comprising: collecting signals of an electromechanical equipment to be detected; calculating signal characteristics of the signal according to characteristics of a preset model corresponding to the electromechanical equipment to be detected; wherein the characteristics of the preset model include characteristics of each type of characteristics of a predetermined category selected according to accuracy; comparing the signal characteristics with a threshold corresponding to the characteristics of the preset model; if there is a signal characteristic greater than or equal to the threshold, obtaining historical signals of the electromechanical equipment to be detected, as well as historical signals of adjacent devices of the electromechanical equipment to be detected; determining the fault detection result of the electromechanical equipment to be detected according to trend information of the historical signals, as well as trend information of the historical signals of the adjacent devices.

[0006] In the above fault detection method, the fault detection device collects the signal of the electromechanical device to be detected, calculates the signal characteristics of the signal according to the characteristics of the preset model corresponding to the electromechanical device to be detected, and then compares the signal characteristics with the threshold corresponding to the characteristics of the preset model. If there is a signal characteristic greater than or equal to the above threshold, the fault detection device obtains the historical signal of the electromechanical device to be detected, as well as the historical signal of the adjacent device of the electromechanical device to be detected, and determines the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device, so that the electromechanical device to be detected can be detected by the preset model corresponding to the electromechanical device to be detected, and then the abnormality of the device can be eliminated before the functional failure of the electromechanical device to be detected occurs, so as to achieve the effect of fault prevention.

[0007] In one possible implementation, determining the fault detection result of the electromechanical equipment to be detected based on the trend information of the historical signal and the trend information of the historical signal of the adjacent device includes: when the trend information of the historical signal shows an upward trend and the trend information of the historical signal of the adjacent device does not show an upward trend, determining that the electromechanical equipment to be detected has a fault.

[0008] In one possible implementation, before determining the fault detection result of the electromechanical equipment to be detected based on the trend information of the historical signal and the trend information of the historical signal of the adjacent equipment, it also includes: calculating the signal characteristics of the historical signal based on the characteristics of the preset model; and obtaining the trend information of the historical signal based on the signal characteristics of the historical signal.

[0009] In one possible implementation, before determining the fault detection result of the electromechanical equipment to be detected based on the trend information of the historical signal and the trend information of the historical signal of the adjacent device, it also includes: calculating the characteristics of the historical signal of the adjacent device based on the characteristics of the preset model; and obtaining the trend information of the historical signal of the adjacent device based on the characteristics of the historical signal of the adjacent device.

[0010] In one possible implementation, after determining the fault detection result of the electromechanical equipment to be detected based on the trend information of the historical signals and the trend information of the historical signals of the adjacent equipment, it also includes: obtaining the review and inspection results of the electromechanical equipment to be detected, and taking the review and inspection results as the true label; comparing the true label with the fault detection result of the electromechanical equipment to be detected; determining the false alarm rate and the number of missed alarms of the preset model based on the comparison results; when the false alarm rate is greater than a predetermined false alarm rate threshold, updating the threshold corresponding to the feature of the preset model; when the number of missed alarms is greater than or equal to the predetermined missed alarm threshold, updating the feature of the preset model.

[0011] In one possible implementation, when the false alarm rate is greater than a predetermined false alarm rate threshold, updating the threshold corresponding to the feature of the preset model includes: obtaining the historical signal and the field signal of the electromechanical equipment to be detected; calculating the feature matrix of the historical signal and the field signal according to the feature of the preset model; fitting the prior distribution of the feature matrix of the historical signal by column to obtain the prior distribution of the historical signal; sampling the prior distribution of the historical signal a predetermined number of times to obtain the sampling values ​​of the prior distribution; wherein the predetermined number of times is equal to the number of field signals; performing error calculation on the sampling values ​​of the prior distribution and the feature matrix of the field signal to obtain an error matrix; fitting the likelihood distribution of the error matrix to obtain the likelihood distribution of the error matrix; multiplying the prior distribution of the historical signal by the likelihood distribution of the error matrix to obtain the posterior distribution of the feature of the preset model; and updating the threshold corresponding to the feature of the preset model according to the posterior distribution.

[0012] In one possible implementation, when the number of missed alarms is greater than or equal to a predetermined missed alarm threshold, updating the features of the preset model includes: obtaining the sensitivity of each type of features of the predetermined category; replacing the features of the preset model with features with high sensitivity, wherein the features with high sensitivity include features in each type of features whose sensitivity is higher than the current features of the preset model.

[0013] In one possible implementation, the predetermined category of features includes one or a combination of the following: impulse features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features.

[0014] In a second aspect, an embodiment of the present application provides a fault detection method for an electromechanical device, comprising: obtaining historical data of an electromechanical device to be detected; dividing the historical data into a training set and a test set; wherein the training set includes a first proportion of normal samples in the historical data, and the test set includes a second proportion of normal samples in the historical data and faulty samples in the historical data; performing feature calculation on each piece of data in the training set according to a predetermined number of features to obtain a first feature matrix; and performing feature calculation on each piece of data in the test set according to the predetermined number of features to obtain a second feature matrix; wherein each column of the first feature matrix and the second feature matrix is ​​a feature, and the number of columns of the first feature matrix and the second feature matrix is ​​equal to the predetermined number; determining a threshold corresponding to each feature according to the first feature matrix; determining a feature of a preset model corresponding to the electromechanical device to be detected and a threshold corresponding to the feature of the preset model according to the second feature matrix and the threshold corresponding to each feature; sending the feature of the preset model and the corresponding threshold to the fault detection device, so that the fault detection device performs fault detection on the electromechanical device to be detected according to the feature of the preset model and the corresponding threshold.

[0015] In the above-mentioned electromechanical equipment fault detection method, the server obtains historical data of the electromechanical equipment to be detected, divides the above-mentioned historical data into a training set and a test set, performs feature calculation on each piece of data in the above-mentioned training set according to a predetermined number of features, and obtains a first feature matrix; and performs feature calculation on each piece of data in the above-mentioned test set according to the above-mentioned predetermined number of features, and obtains a second feature matrix, and then determines the threshold corresponding to each feature according to the first feature matrix, and determines the features of the preset model corresponding to the above-mentioned electromechanical equipment to be detected and the threshold corresponding to the features of the above-mentioned preset model according to the second feature matrix and the threshold corresponding to each feature. Finally, the server sends the features and corresponding thresholds of the above-mentioned preset model to the fault detection device, so that the fault detection device performs fault detection on the above-mentioned electromechanical equipment to be detected according to the features and corresponding thresholds of the above-mentioned preset model, thereby realizing batch deployment of preset models without manual configuration, improving the efficiency of preset model deployment, and saving deployment costs. In addition, when generating the preset model, the server classifies the features and performs sufficient dimensional characterization in the preset feature set, which can effectively avoid overfitting of a specific data set and ensure the generalization of the preset model.

[0016] In one possible implementation, determining the threshold corresponding to each feature according to the first feature matrix includes: calculating the mean and standard deviation of each column of the first feature matrix; and obtaining the threshold corresponding to each feature according to the mean and the standard deviation.

[0017] In one possible implementation, determining the features of the preset model corresponding to the electromechanical equipment to be tested and the thresholds corresponding to the features of the preset model based on the second feature matrix and the thresholds corresponding to each feature includes: comparing the second feature matrix row by row with the threshold corresponding to each feature; obtaining a result matrix based on the comparison results; obtaining a truth matrix based on the data labels of the test set; generating an error number vector based on the result matrix and the truth matrix; obtaining an accuracy vector based on the error number vector; classifying the predetermined number of features to obtain features of a predetermined category; selecting features of the preset model from the features of the predetermined category based on the accuracy vector, and using the thresholds corresponding to the selected features as the thresholds corresponding to the features of the preset model.

[0018] In one possible implementation, the predetermined category of features includes one or a combination of the following: impulse features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features.

[0019] In a third aspect, an embodiment of the present application provides a fault detection device for an electromechanical device, the device being included in a fault detection device, and the device having the function of realizing the behavior of the fault detection device in the first aspect and the possible implementation of the first aspect. The function can be realized by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules or units corresponding to the above functions. For example, an acquisition module, a calculation module, a comparison module, an acquisition module, and a determination module.

[0020] In a fourth aspect, an embodiment of the present application provides a fault detection device for electromechanical equipment, which is included in a server and has the function of implementing the server behavior in the second aspect and the possible implementation of the second aspect. The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules or units corresponding to the above functions. For example, an acquisition module, a division module, a calculation module, a determination module, and a sending module.

[0021] In a fifth aspect, an embodiment of the present application provides a fault detection device, comprising: one or more processors; a memory; multiple applications; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the fault detection device, enable the fault detection device to perform the method provided in the first aspect.

[0022] In the sixth aspect, an embodiment of the present application provides a server, comprising: one or more processors; a memory; multiple applications; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the server, enable the server to execute the method provided in the second aspect.

[0023] It should be understood that the third aspect and the fifth aspect of the embodiment of the present application are consistent with the technical solution of the first aspect of the embodiment of the present application, and the beneficial effects achieved by each aspect and the corresponding feasible implementation methods are similar and will not be repeated here.

[0024] It should be understood that the fourth and sixth aspects of the embodiments of the present application are consistent with the technical solutions of the second aspect of the embodiments of the present application, and the beneficial effects achieved by each aspect and the corresponding feasible implementation methods are similar and will not be repeated here.

[0025] In a seventh aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the method provided in the first aspect.

[0026] In an eighth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the method provided in the second aspect.

[0027] In a ninth aspect, an embodiment of the present application provides a computer program, which, when executed by a computer, is used to execute the method provided in the first aspect.

[0028] In a tenth aspect, an embodiment of the present application provides a computer program, which, when executed by a computer, is used to execute the method provided in the second aspect.

[0029] In one possible design, the programs in the ninth and tenth aspects may be stored in whole or in part on a storage medium packaged together with the processor, or may be stored in whole or in part on a memory not packaged together with the processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1It is a schematic diagram of the equipment performance degradation curve;

[0031] Figure 2 A schematic diagram of a system architecture provided for an embodiment of the present application;

[0032] Figure 3 A schematic diagram of the structure of a server provided for one embodiment of the present application;

[0033] Figure 4 A schematic diagram of the structure of a fault detection device provided in one embodiment of the present application;

[0034] Figure 5 A schematic diagram of a generation process of a preset model provided in one embodiment of the present application;

[0035] Figure 6 A schematic diagram of vibration signal dimensions provided for one embodiment of the present application;

[0036] Figure 7 A schematic diagram of updating a preset model provided by an embodiment of the present application;

[0037] Figure 8 A schematic diagram of a sensitivity rule generation method provided by an embodiment of the present application;

[0038] Fig. 9 A flow chart of a method for detecting a fault in an electromechanical device provided by one embodiment of the present application;

[0039] Fig.10 A flow chart of a method for detecting a fault in an electromechanical device provided by another embodiment of the present application;

[0040] Fig.11 A flowchart of a method for detecting a fault in an electromechanical device provided in yet another embodiment of the present application;

[0041] Fig.12 A schematic diagram of the structure of a fault detection device provided in another embodiment of the present application;

[0042] Fig.13 A structural schematic diagram of a fault detection device provided in yet another embodiment of the present application;

[0043] Fig.14 A schematic diagram of the structure of a server provided in another embodiment of the present application;

[0044] Fig.15 A schematic diagram of the structure of a server provided in yet another embodiment of the present application. DETAILED DESCRIPTION

[0045] The terms used in the implementation section of this application are only used to explain the specific embodiments of this application and are not intended to limit this application.

[0046] In the existing related technologies, many protection measures are used during the use of electromechanical equipment. On the surface, the protection settings of the electromechanical equipment are very complete, but in actual operation, the relay will only sound an alarm when the electromechanical equipment fails to function, and it does not have an early warning function. Figure 1 is a schematic diagram of the equipment performance degradation curve, such as Figure 1 As shown, the existing protection measures can only be used to detect point A, that is, the fault after the functional failure occurs. When reaching point A, the electromechanical equipment has completely failed and cannot work. The existing protection measures cannot detect PFA, that is, the part between the degradation starting point-potential failure point-mechanical protection alarm. In addition, it should be noted that the potential failure is not a failure, but a degradation phenomenon that appears before the failure, which can be perceived. Functional failure is a failure that will cause shutdown in the traditional sense. Therefore, relays can only be used as the last line of defense for the stable operation of electromechanical equipment to prevent major failures or catastrophic failures. However, if the effect of equipment failure prevention cannot be achieved and the equipment abnormality cannot be eliminated before the functional failure, then it will be in vain to set up more protective relays.

[0047] In order to detect faults in electromechanical equipment, the existing related technologies provide two fault diagnosis schemes, one of which is a fault diagnosis method based on a mechanism model, which specifically includes:

[0048] 1) Experts analyze on-site signals and provide multiple signal features corresponding to faults;

[0049] 2) Perform statistical analysis on the features and set the boundary threshold; the set boundary threshold is the monitoring warning line;

[0050] 3) The algorithm is deployed, and a fault warning is issued when the real-time features exceed the above boundary thresholds at the same time and / or one feature exceeds the boundary threshold.

[0051] However, in this method, when the project is delivered, the configuration parameters of each device need to be determined by experts, and the technical generalization is poor.

[0052] Another fault diagnosis solution provided by the existing related technology is a fault diagnosis method based on artificial intelligence (AI). Based on machine learning technology, the original signal is directly input into a machine learning model (e.g., a deep neural network model) for training to obtain a trained model. Then, detection is performed based on the trained model to achieve fault determination of electromechanical equipment.

[0053] However, due to changes in working conditions or environmental factors between electromechanical equipment, there are differences in data distribution in different scenarios. Models trained with specific data sets cannot be directly reused and need to be manually re-trained.

[0054] In summary, existing technical solutions all have generalization problems, which means that professional R&D personnel are required to participate in batch deployment of algorithms for the same type of devices, resulting in high project delivery costs and time-consuming and labor-intensive projects.

[0055] Based on the above problems, an embodiment of the present application provides a fault detection method for electromechanical equipment, which can solve the problem of batch deployment of fault diagnosis algorithms / models of the same type of electromechanical equipment without the participation of R&D personnel, thereby saving manpower costs in the project delivery process; in addition, with a relatively small number of samples, the model can converge quickly, thereby improving the efficiency of batch deployment of the model.

[0056] The electromechanical equipment fault detection method provided in the embodiment of the present application can be applied to industry scenarios of heavy equipment assets such as coal mines, ports and / or power. For the monitoring problem of electromechanical equipment, its system architecture (edge-cloud collaboration) can be as follows: Figure 2 As shown, Figure 2 A schematic diagram of the system architecture provided for an embodiment of the present application: first, in the initial stage, a preset model of a specified device type is generated on the server side based on historical data, and the model is sent to the fault detection device side; the fault detection device side uses the preset model to detect the data of the electromechanical equipment to be detected, and combines the records of on-site operation and maintenance personnel to realize application feedback of the preset model, and the feedback result is transmitted to the server side; the server side adaptively updates the preset model according to the feedback, and sends the updated result to the fault detection device side to form the above-mentioned custom model of the electromechanical equipment to be detected.

[0057] In the embodiment of the present application, the server may be deployed on the cloud side. Specifically, the server may be deployed on the central cloud side. For example, Figure 3 A schematic diagram of the structure of a server provided in one embodiment of the present application, such as Figure 3 As shown, the server 300 is in the form of a general-purpose computing device. The components of the server 300 may include, but are not limited to: one or more processors 310, a communication interface 320, a memory 330, and a communication bus 340 connecting different components (including the memory 330, the communication interface 320 and the processor 310).

[0058] The communication bus 340 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, or a local bus using any of a variety of bus structures. For example, the communication bus 340 may include, but is not limited to, an industry standard architecture (ISA) bus, a micro channel architecture (MCA) bus, an enhanced ISA bus, a video electronics standards association (VESA) local bus, and a peripheral component interconnection (PCI) bus.

[0059] The server 300 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the server 300, including volatile and non-volatile media, removable and non-removable media.

[0060] The memory 330 may include a computer system readable medium in the form of a volatile memory, such as a random access memory (RAM) and / or a cache memory. The memory 330 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the server-side functions in the fault detection method for electromechanical equipment provided in the embodiment of the present application.

[0061] A program / utility having a set (at least one) of program modules may be stored in the memory 330, such program modules including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination thereof may include an implementation of a network environment. The program module generally executes the server-side functions and / or methods in the fault detection method for electromechanical equipment provided in the embodiments of the present application.

[0062] The processor 310 executes various functional applications and data processing by running the programs stored in the memory 330, such as implementing the server-side functions and / or methods in the fault detection method for electromechanical equipment provided in the embodiment of the present application.

[0063] In the embodiment of the present application, the fault detection device can be deployed on the side of the electromechanical device to be detected, that is, the edge side. For example, Figure 4 A schematic diagram of the structure of a fault detection device provided in one embodiment of the present application is shown in FIG. Figure 4As shown, the fault detection device 400 may include a processor 410 and a communication interface 420. Optionally, the fault detection device 400 may also include a memory 430. The processor 410, the communication interface 420 and the memory 430 may communicate with each other through an internal connection path to transmit control and / or data signals, the memory 430 is used to store a computer program, and the processor 410 is used to call and run the computer program from the memory 430.

[0064] The processor 410 and the memory 430 may be combined into a processing device, or more commonly, they are independent components, and the processor 410 is used to execute the program code stored in the memory 430. In specific implementation, the memory 430 may also be integrated into the processor 410, or may be independent of the processor 410.

[0065] In addition, in order to make the function of the fault detection device 400 more complete, the fault detection device 400 may further include one or more of an input unit 460 and a display unit 470 .

[0066] Optionally, the fault detection device 400 may further include a power supply 450 for providing power to various devices or circuits in the fault detection device 400 .

[0067] It should be understood that Figure 4 The processor 410 in the fault detection device 400 shown may be a system on chip SOC, and the processor 410 may include a central processing unit (CPU), and may further include other types of processors, such as a graphics processing unit (GPU) and the like.

[0068] For ease of understanding, the following embodiments of the present application will be described in detail with Figure 3 The server 300 of the structure shown has Figure 4 Taking the fault detection device 400 of the structure shown in FIG. as an example, Figure 2 The system architecture and application scenarios shown specifically illustrate the fault detection method for electromechanical equipment provided in an embodiment of the present application.

[0069] The overall process of the fault detection method for electromechanical equipment provided in the embodiment of the present application can be divided into two parts: "pre-set model training and distribution" and "edge-side on-site detection and update".

[0070] (I) Preset model training and distribution

[0071] First, a large amount of historical sensor data of a certain component is obtained based on experiments or field collection (hereinafter referred to as historical data). The normal samples in the historical data are divided into 7:3, of which 70% are used as training data and 30% are used as test data. The data features are calculated for each training data item, and the distribution statistics of each feature are performed (for example, using the 3sigma method, that is, calculating the mean + 3× standard deviation). Based on the statistical results, the thresholds corresponding to each feature of a certain component under normal conditions are obtained. The feature calculation results of the test data are compared with the thresholds corresponding to each feature calculated. If they are within the threshold, it is judged to be normal, otherwise it is judged to be faulty. The above calculation results are compared with the true value results (sample labels), and the accuracy is calculated using the formula "1-test samples with wrong judgment / total test samples". The feature with the highest accuracy in each dimension and its corresponding threshold are used as the preset model of the component. When the component needs to be analyzed for faults on site, the above preset model is issued. In the feature selection of the preset model, the embodiment of the present application does not use the top 5 features with the highest accuracy as the preset features, but classifies the features by dimension and uses the features with the highest accuracy in each category of the classified features as the preset features.

[0072] (II) On-site detection and update at the edge:

[0073] Phase 1: Detection using preset models: Collect a group of signals from the electromechanical equipment to be detected on site, perform basic filtering and / or noise reduction on the signal, calculate the signal characteristics of the signal using the preset model, and compare the calculated signal characteristics with the threshold corresponding to the characteristic in the preset model. If the threshold is not exceeded, it is judged to be normal; if the threshold is exceeded, read multiple groups of data from adjacent time and adjacent measuring points, calculate the characteristic values ​​of the above multiple groups of data using the preset model, and then perform the following analysis: 1) Perform trend judgment on the characteristics of adjacent time to determine whether the characteristics show an upward trend; 2) Perform correlation judgment on the characteristics of adjacent measuring points to determine whether the characteristics also show an upward trend; if 1) is yes and 2) is no, it is determined that the electromechanical equipment to be detected has a fault; otherwise, it is determined to be normal. The above fault judgment results are recorded (hereinafter referred to as calculation labels). Among them, the adjacent measuring points are the adjacent equipment of the electromechanical equipment to be detected.

[0074] Phase 2, algorithm self-assessment: During fault review or regular inspection, the operation and maintenance personnel will enter the review and inspection results into the system (hereinafter referred to as the real label). The algorithm regularly reads the calculated label and the real label, compares them, and calculates the false alarm rate and the number of missed alarms. If the false alarm rate is ≤10% and the number of missed alarms is less than 1, return to phase one for detection; otherwise, enter phase three. Among them, false alarm rate = number of false alarm samples / total number of real label samples; among them, the above number of false alarm samples is the number of samples that are faulty in the calculated label but not faulty in the real note; the number of missed alarms is the number of samples that are faulty in the real label but not faulty in the calculated label.

[0075] Phase 3, online model update: If the false alarm rate is > 10%, the threshold update model is used to update the feature threshold; if the number of missed alarms is greater than or equal to 1, the feature set is updated, and the feature threshold uses the threshold of the corresponding feature in the preset model. According to the above update process, the model features and their thresholds are updated to form an updated online model. Use this online model to continue to perform fault detection on the equipment to be detected according to the process of Phase 1.

[0076] The fault detection method for electromechanical equipment provided in the embodiment of the present application adopts a cloud-edge collaborative data processing and analysis architecture of "preset model + model adaptive update" to achieve efficient integration of information extracted from historical data (empirical knowledge) and field data information; and the preset model feature set considers multiple fault feature dimensions to avoid the decrease in generalization caused by over-emphasis on a certain dimension; in addition, the time dimension and space dimension are added to the preset model to improve the distinction between normal interference and faults and achieve stable fault judgment; through Bayesian reasoning (for example: 3sigma method), it is possible to obtain accurate feature thresholds when accessing a small amount of new equipment data, and it is also possible to achieve feature adaptive optimization based on field data.

[0077] The following takes bearing vibration monitoring as an example to illustrate the fault detection method of electromechanical equipment provided by the embodiment of the present application. In the example of bearing vibration monitoring, the electromechanical equipment to be detected is a bearing, and the sensor is a vibration sensor.

[0078] The following first introduces the generation process of the preset model. The generation of the preset model is to generate the preset model based on historical data, and the preset model may include a feature set and thresholds of corresponding features.

[0079] Figure 5 A schematic diagram of a generation process of a preset model provided in one embodiment of the present application, Figure 5 The process of obtaining a preset model from historical data without field data is shown, which may include:

[0080] 1) Data sample preparation: Historical data can include public data sets, field data collected from previous projects, and other samples (a total of 43,000 samples, including 40,000 normal samples and 3,000 fault samples). Figure 5 , the normal samples in the historical data set are divided, 70% of which constitute the training set (40000×0.7=28000), and 30% of the normal samples + fault samples constitute the test set (40000×0.3+3000=15000).

[0081] 2) Signal characteristics and characteristic threshold calculation: see Figure 5 , the training set data is calculated one by one, and 18 eigenvalues ​​are obtained (pulse index, kurtosis, gini coefficient, harmonic noise ratio, entropy spectrum, harmonic product spectrum peak, envelope energy ratio, fault frequency band energy proportion, envelope signal fault frequency band energy proportion, peak, root mean square value, variance, center of gravity frequency, mean square frequency, frequency variance, symmetry basis, fluctuation coefficient and Gaussian correlation), and each data is calculated to obtain a 1×18 one-dimensional vector. After all training data are calculated, a 28000×18 feature matrix is ​​obtained; then, the threshold is calculated column by column, that is, the mean (mean) and standard deviation (std) of each column of the above matrix (that is, according to feature statistics) are calculated, and "mean + 3×standard deviation" is recorded as the feature threshold, and 1×18 feature threshold vectors are obtained.

[0082] 3) Preset model generation and distribution: see Figure 5 , calculate the test set data one by one to get 18 eigenvalues, then each data is calculated to get a 1×18 one-dimensional vector. After all the test data are calculated, a 15000×18 feature matrix is ​​obtained; compare the above matrix with the feature threshold vector row by row. If the calculation result is greater than the threshold at the corresponding position, it is judged to be faulty; otherwise, it is judged to be normal. Record fault as 1 and normal as 0; according to the above comparison results, a 15000×18 test result matrix (0-1 matrix) can be obtained; the true value matrix (15000×18 dimensional 0-1 matrix) is given using the test set data labels; the test result matrix is ​​subtracted from the true value matrix, and the absolute value is calculated. The above matrix is ​​summed by column to obtain a 1×18 calculation error vector (a matrix after subtracting the absolute value and summing the columns: 0 represents that the calculation result is consistent with the true value, and 1 represents that the calculation result is inconsistent with the true value). In this way, the accuracy vector = 1-calculation error vector / 15000. The 18 features are classified as follows, and the feature with the highest accuracy in each category is used as the feature of the preset model, and the value in the corresponding threshold vector is used as the threshold corresponding to the feature of the preset model. When the on-site project needs to diagnose the bearing fault, the above preset model (feature + corresponding threshold) is issued.

[0083] Figure 6A schematic diagram of vibration signal dimensions provided for an embodiment of the present application is shown in FIG. Figure 6 As shown, the feature classification is as follows:

[0084] a) Impact characteristics: pulse index, kurtosis, gini coefficient;

[0085] b) Harmonic characteristics: harmonic-to-noise ratio, entropy spectrum, and harmonic product spectrum peak;

[0086] c) Modulation characteristics: envelope energy ratio, fault frequency band energy ratio, envelope signal fault frequency band energy ratio;

[0087] d) Energy characteristics: peak value, root mean square value, variance;

[0088] e) Spectrum distribution characteristics: centroid frequency, mean square frequency, frequency variance;

[0089] f) Signal fluctuation characteristics: symmetry basis, fluctuation coefficient, Gaussian correlation.

[0090] In this embodiment, the features are classified and fully dimensionalized in the preset feature set, which can effectively avoid overfitting of a specific data set and ensure the generalization of the preset model.

[0091] Next, the detection process on the edge side is introduced. The edge side detection is to use the preset model to detect the bearing to be detected, and to optimize the model online based on the comparison between the detection result and the actual result. In this example, the preset model feature set may include: kurtosis, entropy spectrum, fault frequency band energy ratio, root mean square value, mean square frequency and fluctuation coefficient. Specifically, the detection process on the edge side may include:

[0092] 1) Multi-dimensional feature extraction: A group of signals are collected on-site, and band-pass filtering is performed on the signals (for example, 3000Hz to 5000Hz band-pass filtering). The signal characteristics of the group of signals, namely, kurtosis, entropy spectrum, fault frequency band energy ratio, root mean square value, mean square frequency and fluctuation coefficient, are calculated according to the characteristics of the preset model to obtain a 1×6-dimensional vector, which is compared with the 1×6-dimensional feature threshold. If the threshold is not exceeded, it can be determined that the bearing is normal; if a certain feature (or multiple features) exceeds the corresponding threshold, multivariate spatiotemporal statistical analysis is performed.

[0093] 2) Multivariate spatiotemporal statistical analysis: Criterion 1, take the historical signals from the 12 time periods before the current moment to the current moment, and obtain the trend information through differential operation, that is, diff(n)=a(n+1)-a(n), n=1:12. If diff(1:12)≥0, the trend judgment result is recorded as trend=1, otherwise trend=0; Criterion 2, take the signals of the 5 adjacent devices of the bearing to be tested. If the bearing to be tested has no adjacent devices, no calculation is performed, and space=1 is directly set; if there are adjacent devices, the following calculation is performed: calculate the trend information of the 5 adjacent devices and obtain 5 diff(1:12). If the data of more than half (that is, 3) of the adjacent devices have diff(1:n)≥0, it is judged to be normal, that is, space=0; otherwise, it is judged to be faulty, that is, space=1.

[0094] Fault determination: When trend = 1 and space = 1, it is determined to be a fault, and the detection result of the bearing is set to 1; otherwise, it is determined to be normal, and the detection result of the bearing is set to 0. That is to say, when the trend information of the historical signal of the bearing shows an upward trend, and the trend information of the signal of the above-mentioned adjacent device does not show an upward trend, it is determined that the bearing has a fault; when the trend information of the historical signal of the bearing shows an upward trend, and the trend information of the signal of the above-mentioned adjacent device also shows an upward trend, it can be determined that the bearing is normal. In this embodiment, by introducing the time trend and the trend of the adjacent device for fault determination, false alarms can be effectively reduced.

[0095] 3) Edge-side model update: Generally speaking, the pre-set model can only achieve a moderate level of accuracy. If the accuracy needs to be improved, it needs to be updated based on field data. The core issue is how to improve the efficiency of model update convergence, that is, to update under fewer sample conditions. Edge-side model updates can include:

[0096] a. Algorithm self-assessment: Enter the review and inspection results of the operation and maintenance personnel into the system. The review and inspection results of the operation and maintenance personnel are referred to as true labels below. The algorithm regularly reads the calculated labels and true labels and compares them. The calculated labels are the fault detection results of the preset model for the bearing. Then, the false alarm rate and the number of missed alarms are calculated. If the false alarm rate is >10% or the number of missed alarms is ≥1, it indicates that the preset model needs to be updated. The false alarm rate is the ratio of the number of false alarm samples to the total number of true label samples. The number of false alarm samples is the number of samples that are faulty in the calculated labels but not faulty in the true labels; the number of missed alarms is the number of samples that are faulty in the true labels but not faulty in the calculated labels.

[0097] b. False alarm processing: When the false alarm rate is > 10%, the threshold corresponding to each feature in the preset model is updated. Figure 7 A schematic diagram of updating a preset model provided in one embodiment of the present application is provided. Figure 7 Taking the updating of the threshold corresponding to the kurtosis feature as an example, the updating method of the thresholds corresponding to other features is the same. Figure 7 As shown, the update method is as follows:

[0098] (1) Prior distribution fitting: Using 40,000 normal samples in history, feature calculation is performed to obtain a 40,000×6 feature matrix (only the eigenvalues ​​related to the preset model are calculated), and Gaussian distribution fitting is performed column by column to obtain 6 prior Gaussian distributions, p1(x)~p6(x); the following takes p1(x) as an example for explanation, and the update methods of the remaining p2(x)~p6(x) are the same.

[0099] (2) Likelihood distribution fitting: Take feature 1 (e.g. kurtosis) as an example: there are 200 pieces of current field data, then calculate the calculation vector corresponding to the 200 pieces of data and feature 1, and obtain a 1×200 feature 1 calculation vector; select a sampling method, sample p1(x) 200 times, and obtain 200 sampling values ​​(1×200 vector); subtract the sampling value vector from the feature 1 calculation vector to obtain an error vector (1×200 vector); perform Gaussian distribution fitting on the above error vector to obtain the likelihood Gaussian distribution p1(y|x) of feature 1. Gaussian distribution sampling methods can include: inverse transformation method, Box-Muller algorithm and rejection sampling method, etc. Here, you can select one sampling method.

[0100] (3) Calculation of the posterior distribution: The posterior distribution of feature 1 is obtained by multiplication: p1(x|y)=p1(x)×p1(y|x), and the result is a Gaussian distribution. Specifically, since both the prior distribution and the likelihood distribution are Gaussian distributions, the posterior distribution must be a Gaussian distribution.

[0101] (4) Update the threshold corresponding to the feature: Use the sampling method to sample p1(x|y) 100,000 times, calculate the mean and standard deviation of the 100,000 samples, and update the threshold corresponding to the feature according to the mean and standard deviation. Specifically, the threshold corresponding to the feature can be updated according to formula (1).

[0102] bound = mean + 3×std (1)

[0103] In formula (1), bound is the threshold corresponding to the feature, mean is the mean of the posterior distribution sampling, and std is the standard deviation of the posterior distribution sampling.

[0104] (5) Threshold update test: Use the updated threshold to test the results of 200 field data. If the threshold is exceeded, it is a fault; if the threshold is not exceeded, it is normal. Compare the results with the real labels of the data. If the false alarm rate is less than 10% and there is no missed alarm, the test passes and the threshold corresponding to the feature in the preset model is updated to the above threshold; otherwise, it is switched to manual analysis.

[0105] c. False negative processing: When a false negative occurs, the feature set is updated. The updating method is as follows:

[0106] (1) Feature update rule table: Take a certain type of feature for sensitivity analysis. The impact type feature is taken as an example below. The processing methods for other types of features are the same. Figure 8 A schematic diagram of a sensitivity rule generation method provided by an embodiment of the present application is shown as follows: Figure 8 As shown, five groups of data are generated, each with 3000 data points. Each data point is sampled according to a Gaussian distribution with amplitude strengths of 1, 5, 10, 20, and 50, resulting in 5×3000 data points. The Gaussian noise features (pulse index, kurtosis, and gini coefficient) of the above 5×3000 data points are calculated to obtain 5×3×3000 eigenvalues. The 3000 fault samples are added to the five groups of data to obtain 5×3000 superimposed data points, and the superimposed features (pulse index, kurtosis, and gini coefficient) are calculated to obtain 5×3×3000 eigenvalues. The superimposed features are subtracted from the Gaussian noise features to obtain 5×3×3000 difference eigenvalues. The mean of the difference eigenvalues ​​is calculated for each group of 3000. The mean reflects the sensitivity of the feature to noise, totaling 5×3. Based on the above 5×3 sensitivity matrix, the sensitivity relationship of the feature is obtained; see Figure 8 The sensitivity of the pulse index is 3, 1.1, 0.01, -2, -3; the sensitivity of the kurtosis is 6, 3, 2.1, 0.8, -0.4; the sensitivity of the gini coefficient is 9, 6, 4, 2, 1.1; therefore, sensitivity: gini coefficient>kurtosis>pulse index.

[0107] (2) Feature set update: If there is a missed report on site, the feature set and corresponding threshold of the preset model are replaced with a feature set and corresponding threshold with a sensitivity level one higher. For example, if the impact feature in the original feature set of the preset model is kurtosis, it is replaced with the gini coefficient and its threshold. This embodiment establishes the "optimization direction" of feature selection through sensitivity analysis, which can meet the personalized operation and maintenance needs of customers without manual configuration.

[0108] (3) Update test: Use the updated feature values ​​and corresponding thresholds to test the results of 200 on-site data. If the threshold is exceeded, it is a fault; if the threshold is not exceeded, it is normal. Compare the results with the real labels of the data. If the false alarm rate is less than 10% and there are no missed alarms, the test passes. If the false alarm rate is greater than 10% and there are no missed alarms, the false alarm processing is started and the threshold corresponding to each feature is updated. In other cases, manual analysis is performed.

[0109] Through the above steps, it is possible to realize adaptive optimization of the preset model without human intervention, integrate the information of a small amount of on-site data with the information of a large amount of historical data, and realize manual-free deployment and rapid convergence of models between devices of the same type.

[0110] As described above, the present application embodiment provides a method for detecting a fault in an electromechanical device. Fig. 9 A flowchart of a method for detecting a fault in an electromechanical device provided by an embodiment of the present application is shown in FIG. Fig. 9 As shown, the above fault detection method may include:

[0111] Step 901 : The fault detection device 400 collects signals from the electromechanical device to be detected.

[0112] Step 902 : The fault detection device 400 calculates the signal characteristics of the above signal according to the characteristics of the preset model corresponding to the above electromechanical device to be detected.

[0113] The features of the preset model include features selected according to accuracy in each feature of the predetermined category. The features of the predetermined category may include one or a combination of the following: impulse features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features. Figure 6 As shown, each of the above-mentioned features may include at least two features, and at least one feature in each feature category may be selected as the feature of the preset model according to the accuracy. In specific implementation, the feature with the highest accuracy may be selected from each feature category of the predetermined category as the feature of the preset model. That is to say, the feature of the above-mentioned preset model may include the feature with the highest accuracy in each feature category of the predetermined category.

[0114] In step 903, the fault detection device 400 compares the signal feature with the threshold value corresponding to the feature of the preset model.

[0115] Step 904: If there is a signal feature greater than or equal to the threshold, the fault detection device 400 obtains the historical signal of the electromechanical device to be detected and the historical signal of the adjacent device of the electromechanical device to be detected.

[0116] For example, the historical signals of the electromechanical equipment to be detected may include the historical signals from 12 time periods before the current moment to the current moment, and the historical signals of the adjacent devices of the electromechanical equipment to be detected may include the historical signals of 5 adjacent devices of the electromechanical equipment to be detected, and the historical signals of these 5 adjacent devices may also be the historical signals from 12 time periods before the current moment to the current moment.

[0117] Step 905 : The fault detection device 400 determines the fault detection result of the electromechanical device to be detected according to the trend information of the historical signals and the trend information of the historical signals of the adjacent devices.

[0118] Furthermore, before step 905, the fault detection device 400 may first obtain the trend information of the historical signal and the trend information of the historical signal of the adjacent device.

[0119] The trend information of the historical signal may be obtained by: the fault detection device 400 calculates the signal characteristics of the historical signal according to the characteristics of the preset model, and obtains the trend information of the historical signal according to the signal characteristics of the historical signal.

[0120] Obtaining trend information of historical signals of the adjacent devices may be as follows: the fault detection device 400 calculates features of historical signals of the adjacent devices according to features of the preset model, and obtains trend information of historical signals of the adjacent devices according to features of the historical signals of the adjacent devices.

[0121] In specific implementation, when obtaining trend information of historical signals according to the characteristics of historical signals, it can be achieved through differential operation. Of course, the embodiments of the present application are not limited to this, and it can also be achieved by using methods such as moving average difference or exponential moving average to calculate and analyze the monotonicity of data.

[0122] In this way, the fault detection device 400 determines the fault detection result of the electromechanical device to be detected based on the trend information of the above historical signals and the trend information of the historical signals of the above adjacent devices. The result can be: when the trend information of the above historical signals shows an upward trend, and the trend information of the historical signals of the above adjacent devices does not show an upward trend, it is determined that the electromechanical device to be detected has a fault. For example, assuming that there are 5 adjacent devices to the electromechanical device to be detected, if the trend information of the historical signals of the electromechanical device to be detected shows an upward trend, and the trend information of the historical signals of the above adjacent devices shows an upward trend of less than 3 of the 5 adjacent devices, this means that the trend information of the historical signals of the electromechanical device to be detected shows an upward trend, and the trend information of the historical signals of the above adjacent devices does not show an upward trend. At this time, it can be determined that the electromechanical device to be detected has a fault. If the trend information of the historical signal of the electromechanical equipment to be detected shows an upward trend, and the trend information of the five adjacent devices showing an upward trend is greater than or equal to 3, this means that the trend information of the historical signal of the electromechanical equipment to be detected shows an upward trend, and the trend information of the historical signal of the above-mentioned adjacent devices also shows an upward trend. In this case, the historical signal of the electromechanical equipment to be detected shows an upward trend, which may be caused by the cooperation and / or resonance between the devices, and it cannot be determined that the electromechanical equipment to be detected has a fault.

[0123] In the above fault detection method, the fault detection device 400 collects the signal of the electromechanical device to be detected, calculates the signal characteristics of the signal according to the characteristics of the preset model corresponding to the electromechanical device to be detected, and then compares the signal characteristics with the threshold corresponding to the characteristics of the preset model. If there is a signal characteristic greater than or equal to the threshold, the fault detection device 400 obtains the historical signal of the electromechanical device to be detected and the historical signal of the adjacent device of the electromechanical device to be detected, and determines the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device, so that the electromechanical device to be detected can be detected by the preset model corresponding to the electromechanical device to be detected, and then the abnormality of the device can be eliminated before the functional failure of the electromechanical device to be detected occurs, so as to achieve the effect of fault prevention.

[0124] Fig.10 A flowchart of a method for detecting a fault in an electromechanical device provided by another embodiment of the present application is shown in FIG. Fig.10 As shown, this application Fig. 9 In the illustrated embodiment, after step 905, the following may also be included:

[0125] Step 1001 , the fault detection device 400 obtains the review and inspection results of the electromechanical equipment to be detected, and uses the review and inspection results as true labels.

[0126] The above-mentioned review and inspection results of the electromechanical equipment to be tested may be obtained after the operation and maintenance personnel review and inspect the electromechanical equipment to be tested.

[0127] Step 1002 , the fault detection device 400 compares the real label with the fault detection result of the electromechanical device to be detected.

[0128] Specifically, the fault detection device 400 can regularly or periodically compare the real tag with the fault detection result of the electromechanical device to be detected.

[0129] Step 1003: The fault detection device 400 determines the false alarm rate and the number of missed alarms of the preset model according to the comparison result.

[0130] Among them, the false alarm rate is the ratio of the number of false alarm samples to the total number of true label samples. The number of false alarm samples is the number of samples with faults in the calculated labels but not faults in the true labels; the number of missed alarms is the number of samples with faults in the true labels but not faults in the calculated labels.

[0131] Step 1004, when the false alarm rate is greater than the predetermined false alarm rate threshold, the fault detection device 400 updates the threshold corresponding to the feature of the preset model; when the number of missed alarms is greater than or equal to the predetermined missed alarm threshold, the fault detection device 400 updates the feature of the preset model.

[0132] Specifically, the above-mentioned predetermined false alarm rate threshold can be set by oneself according to system performance and / or implementation requirements during specific implementation. The present embodiment does not limit the size of the above-mentioned predetermined false alarm rate threshold. For example, the predetermined false alarm rate threshold can be 10%; the above-mentioned predetermined missed alarm number threshold can be set by oneself according to system performance and / or implementation requirements during specific implementation. The present embodiment does not limit the size of the above-mentioned predetermined missed alarm number threshold. For example, the predetermined missed alarm number threshold can be 1.

[0133] In this embodiment, when the false alarm rate is greater than the predetermined false alarm rate threshold, the fault detection device 400 updates the threshold corresponding to the feature of the preset model as follows: the fault detection device 400 obtains the historical signal and the field signal of the electromechanical device to be detected, and calculates the feature matrix of the historical signal and the field signal according to the feature of the preset model. Then, the fault detection device 400 performs prior distribution fitting on the feature matrix of the historical signal by column to obtain the prior distribution of the historical signal, and then samples the prior distribution of the historical signal for a predetermined number of times to obtain the sampling value of the prior distribution; wherein the predetermined number of times is equal to the number of the field signal. Next, the fault detection device 400 performs error calculation on the sampling value of the prior distribution and the feature matrix of the field signal to obtain the error matrix, and performs likelihood distribution fitting on the error matrix to obtain the likelihood distribution of the error matrix. Finally, the fault detection device 400 multiplies the prior distribution of the historical signal by the likelihood distribution of the error matrix to obtain the posterior distribution of the feature of the preset model, and updates the threshold corresponding to the feature of the preset model according to the posterior distribution. This embodiment can achieve manual-free deployment and rapid model convergence between devices of the same type by effectively fusing historical signals and field signals.

[0134] In a specific implementation, according to the above posterior distribution, updating the threshold corresponding to the feature of the above preset model can be: using a sampling method to sample the above posterior distribution, calculating the mean and standard deviation of the above sampling values, and then updating the threshold corresponding to the feature of the above preset model according to the above mean and standard deviation. Specifically, the threshold corresponding to the feature of the above preset model can be updated according to formula (1).

[0135] In this embodiment, when the above-mentioned number of missed alarms is greater than or equal to the predetermined missed alarm threshold, the fault detection device 400 may update the features of the preset model as follows: the fault detection device 400 obtains the sensitivity of each type of features of the above-mentioned predetermined category, and then replaces the features of the preset model with features with high sensitivity. The features with high sensitivity may be features of each type of features with higher sensitivity than the current features of the preset model. For example, assuming that the sensitivity gini coefficient of the impact feature is greater than the kurtosis and the pulse index, and the impact feature in the current feature set of the preset model is the kurtosis, then when the number of missed alarms is greater than or equal to the predetermined missed alarm threshold, the impact feature in the current feature set of the preset model may be changed to the gini coefficient. Of course, after the feature is updated, the threshold is also updated accordingly. In specific implementation, the fault detection device 400 may be configured as follows: Figure 8 The method provided in the illustrated embodiment obtains the sensitivity of each feature of the above predetermined category, which will not be described in detail here. This embodiment establishes the "optimization direction" of feature optimization through sensitivity analysis, which can meet the personalized operation and maintenance needs of customers without manual configuration.

[0136] Fig.11 A flowchart of a method for detecting a fault in an electromechanical device provided in another embodiment of the present application is shown in FIG. Fig.11 As shown, the above-mentioned fault detection method of electromechanical equipment may include:

[0137] Step 1101 : the server 300 obtains historical data of the electromechanical device to be detected.

[0138] Step 1102: The server 300 divides the historical data into a training set and a test set.

[0139] The training set includes a first proportion of normal samples in the historical data, and the test set includes a second proportion of normal samples in the historical data and fault samples in the historical data. The first proportion and the second proportion can be set according to system performance and / or implementation requirements during specific implementation. This embodiment does not limit the size of the first proportion and the second proportion. For example, the first proportion can be 70% and the second proportion can be 30%.

[0140] In step 1103, the server 300 performs feature calculation on each piece of data in the training set according to a predetermined number of features to obtain a first feature matrix; and performs feature calculation on each piece of data in the test set according to the predetermined number of features to obtain a second feature matrix.

[0141] Among them, each column of the first characteristic matrix and the second characteristic matrix is ​​a feature, and the number of columns of the first characteristic matrix and the second characteristic matrix is ​​equal to the above-mentioned predetermined number. Specifically, the above-mentioned predetermined number can be set according to the system performance and / or implementation requirements during the specific implementation. The present embodiment does not limit the size of the above-mentioned predetermined number. For example, the above-mentioned predetermined number can be 18. For example, the 18 features may include: pulse index, kurtosis, gini coefficient, harmonic noise ratio, entropy spectrum, harmonic product spectrum peak, envelope energy ratio, fault band energy ratio, envelope signal fault band energy ratio, peak value, root mean square value, variance, center of gravity frequency, mean square frequency, frequency variance, symmetry basis, fluctuation coefficient and Gaussian correlation.

[0142] Step 1104: The server 300 determines a threshold corresponding to each feature according to the first feature matrix.

[0143] Specifically, the server 300 determines the threshold corresponding to each feature according to the first feature matrix by calculating the mean and standard deviation of each column of the first feature matrix, and then obtaining the threshold corresponding to each feature according to the mean and standard deviation.

[0144] Step 1105 : The server 300 determines the features of the preset model corresponding to the electromechanical device to be detected and the thresholds corresponding to the features of the preset model according to the second feature matrix and the thresholds corresponding to each feature.

[0145] Specifically, the server 300 determines the features of the preset model corresponding to the electromechanical equipment to be tested and the thresholds corresponding to the features of the preset model based on the second feature matrix and the thresholds corresponding to each feature. The server 300 compares the second feature matrix row by row with the thresholds corresponding to each feature, and obtains a result matrix based on the comparison results; and obtains a truth matrix based on the data labels of the test set; then, the server 300 generates an error count vector based on the result matrix and the truth matrix, and then obtains an accuracy vector based on the error count vector; finally, the predetermined number of features are classified to obtain features of a predetermined category, and features of the preset model are selected from the features of the predetermined category based on the accuracy vector, and the thresholds corresponding to the selected features are used as the thresholds corresponding to the features of the preset model.

[0146] In specific implementation, the server 300 may generate an error number vector according to the result matrix and the truth matrix by subtracting the result matrix from the truth matrix and calculating the absolute value, and then summing the absolute value matrix by column to obtain the error number vector.

[0147] The server 300 may obtain the accuracy vector according to the above error number vector as follows: The server obtains the accuracy vector according to formula (2).

[0148] Accuracy vector = 1-error vector / number of data in the test set (2)

[0149] The server 300 may select the feature of the preset model from the features of the predetermined category according to the accuracy vector as follows: the server 300 may select the feature with the highest accuracy from each feature of the predetermined category as the feature of the preset model.

[0150] In this embodiment, the above-mentioned predetermined categories of features include one or a combination of the following: impulse features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features. Each of the above-mentioned features may include at least two features, and the features included in each feature may be as follows: Figure 6 shown.

[0151] In step 1106, the server 300 sends the characteristics of the preset model and the corresponding threshold to the fault detection device 400, so that the fault detection device 400 can perform fault detection on the electromechanical device to be detected according to the characteristics of the preset model and the corresponding threshold.

[0152] In the above-mentioned fault detection method for electromechanical equipment, the server 300 obtains historical data of the electromechanical equipment to be detected, divides the above-mentioned historical data into a training set and a test set, and performs feature calculation on each data in the above-mentioned training set according to a predetermined number of features to obtain a first feature matrix; and performs feature calculation on each data in the above-mentioned test set according to the above-mentioned predetermined number of features to obtain a second feature matrix, and then determines the threshold corresponding to each feature according to the first feature matrix, and determines the features of the preset model corresponding to the above-mentioned electromechanical equipment to be detected and the threshold corresponding to the features of the above-mentioned preset model according to the second feature matrix and the threshold corresponding to each feature. Finally, the server 300 sends the features and corresponding thresholds of the above-mentioned preset model to the fault detection device 400, so that the fault detection device 400 can perform fault detection on the above-mentioned electromechanical equipment to be detected according to the features and corresponding thresholds of the above-mentioned preset model, thereby realizing batch deployment of preset models without manual configuration, improving the efficiency of preset model deployment, and saving deployment costs. In addition, when generating the preset model, the server 300 classifies the features and performs sufficient dimensional characterization in the preset feature set, which can effectively avoid overfitting of specific data sets and ensure the generalization of the preset model.

[0153] It is to be understood that some or all of the steps or operations in the above embodiments are merely examples, and the present application embodiments may also perform other operations or variations of various operations. In addition, the various steps may be performed in different orders presented in the above embodiments, and it is possible that not all of the operations in the above embodiments need to be performed.

[0154] It is understandable that, in order to realize the above functions, the fault detection device includes hardware and / or software modules corresponding to the execution of each function. In combination with the algorithm steps of each example described in the embodiments disclosed in this application, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application in combination with the embodiments, but such implementation should not be considered to be beyond the scope of this application.

[0155] This embodiment can divide the functional modules of the fault detection device according to the above method embodiment. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0156] Fig.12This is a structural diagram of a fault detection device provided by another embodiment of the present application, in which each functional module is divided according to each function. Fig.12 A possible schematic diagram of the composition of the fault detection device 1200 involved in the above embodiment is shown. Fig.12 As shown, the fault detection device 1200 may include: a collection module 1201, a calculation module 1202, a comparison module 1203, an acquisition module 1204 and a determination module 1205;

[0157] The acquisition module 1201 is used to acquire signals of the electromechanical equipment to be detected.

[0158] A calculation module 1202 is used to calculate the signal characteristics of the above-mentioned signal according to the characteristics of the preset model corresponding to the above-mentioned electromechanical device to be detected; wherein the characteristics of the above-mentioned preset model include characteristics selected according to the accuracy rate in each category of characteristics of the predetermined category;

[0159] A comparison module 1203 is used to compare the signal feature with a threshold value corresponding to the feature of the preset model;

[0160] The acquisition module 1204 is used to acquire the historical signal of the electromechanical device to be detected and the historical signal of the adjacent device of the electromechanical device to be detected when there is a signal feature greater than or equal to the threshold;

[0161] The determination module 1205 is used to determine the fault detection result of the electromechanical device to be detected based on the trend information of the historical signal and the trend information of the historical signal of the adjacent device. In this embodiment, the determination module 1205 is specifically used to determine that the electromechanical device to be detected has a fault when the trend information of the historical signal shows an upward trend and the trend information of the historical signal of the adjacent device does not show an upward trend.

[0162] It should be noted that this application Fig. 9 All relevant contents of each step involved in the illustrated method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0163] The fault detection device 1200 provided in this embodiment is used to execute the present application Fig. 9 The fault detection method for electromechanical equipment provided by the illustrated embodiment can therefore achieve the same effect as the above method.

[0164] Fig.13 A structural diagram of a fault detection device provided in another embodiment of the present application, Fig.12 Compared with the fault detection device shown in the figure, the difference is that Fig.13In the fault detection device 1200 shown, the calculation module 1202 is also used to calculate the signal characteristics of the above-mentioned historical signal according to the characteristics of the above-mentioned preset model before the determination module 1205 determines the fault detection result of the electromechanical equipment to be detected; and obtain the trend information of the above-mentioned historical signal according to the signal characteristics of the above-mentioned historical signal.

[0165] The calculation module 1202 is also used to calculate the characteristics of the historical signals of the above-mentioned adjacent devices according to the characteristics of the above-mentioned preset model before the determination module 1205 determines the fault detection result of the electromechanical equipment to be detected; and obtain the trend information of the historical signals of the above-mentioned adjacent devices according to the characteristics of the historical signals of the above-mentioned adjacent devices.

[0166] Furthermore, the fault detection device 1200 may also include: an updating module 1206;

[0167] The acquisition module 1204 is further used to obtain the review and inspection results of the electromechanical equipment to be detected after the determination module 1205 determines the fault detection result of the electromechanical equipment to be detected, and use the review and inspection results as the true label;

[0168] The comparison module 1203 is further used to compare the real label with the fault detection result of the electromechanical device to be detected;

[0169] The determination module 1205 is further used to determine the false alarm rate and the number of missed alarms of the preset model according to the comparison result;

[0170] The updating module 1206 is used to update the threshold corresponding to the feature of the preset model when the false alarm rate is greater than the preset false alarm rate threshold; when the number of missed alarms is greater than or equal to the preset missed alarm threshold, update the feature of the preset model.

[0171] In this embodiment, the updating module 1206 may include: a signal acquisition submodule 12061, a matrix calculation submodule 12062, an acquisition submodule 12063 and a model updating submodule 12064;

[0172] The signal acquisition submodule 12061 is used to acquire the historical signal and the on-site signal of the electromechanical equipment to be detected;

[0173] The matrix calculation submodule 12062 is used to calculate the characteristic matrix of the historical signal and the field signal according to the characteristics of the preset model;

[0174] The acquisition submodule 12063 is used to fit the prior distribution of the feature matrix of the historical signal by column to obtain the prior distribution of the historical signal; sample the prior distribution of the historical signal a predetermined number of times to obtain the sampling value of the prior distribution; wherein the predetermined number of times is equal to the number of the field signal; perform error calculation on the sampling value of the prior distribution and the feature matrix of the field signal to obtain the error matrix; perform likelihood distribution fitting on the error matrix to obtain the likelihood distribution of the error matrix; and multiply the prior distribution of the historical signal by the likelihood distribution of the error matrix to obtain the posterior distribution of the features of the preset model;

[0175] The model updating submodule 12064 is used to update the threshold corresponding to the feature of the preset model according to the posterior distribution.

[0176] In this embodiment, the obtaining submodule 12063 is used to obtain the sensitivity of each type of feature of the above predetermined category;

[0177] The model updating submodule 12064 is used to replace the features of the preset model with features with high sensitivity, where the features with high sensitivity include features of each type of features whose sensitivity is higher than the current features of the preset model.

[0178] It should be noted that this application Figure 9-10 All relevant contents of each step involved in the illustrated method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0179] The fault detection device 1200 provided in this embodiment is used to execute the present application Figure 9-10 The fault detection method for electromechanical equipment provided by the illustrated embodiment can therefore achieve the same effect as the above method.

[0180] It should be understood that the fault detection device 1200 may correspond to Figure 4 The fault detection device 400 shown in FIG. 1 is a device for detecting a fault in a computer system. The functions of the acquisition module 1201, the calculation module 1202, the comparison module 1203, the acquisition module 1204, the determination module 1205 and the update module 1206 can be Figure 4 The processor 410 in the fault detection device 400 is shown to be implemented.

[0181] In the case of adopting an integrated unit, the fault detection device 1200 may include a processing module, a storage module, and a communication module.

[0182] Among them, the processing module can be used to control and manage the actions of the fault detection device 1200, for example, it can be used to support the fault detection device 1200 to execute the steps executed by the above-mentioned acquisition module 1201, calculation module 1202, comparison module 1203, acquisition module 1204, determination module 1205 and update module 1206. The storage module can be used to support the fault detection device 1200 to store program codes and data, etc. The communication module can be used to support the communication between the fault detection device 1200 and other devices.

[0183] Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logic boxes, modules and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, and so on. The storage module can be a memory. The communication module can specifically be a device that interacts with other electronic devices, such as a radio frequency circuit, a Bluetooth chip and / or a Wi-Fi chip.

[0184] In one embodiment, when the processing module is a processor and the storage module is a memory, the fault detection device 1200 involved in this embodiment can be a Figure 4 Device of the structure shown.

[0185] Similarly, it is understandable that, in order to implement the above functions, the server includes hardware and / or software modules corresponding to the execution of each function. In combination with the algorithm steps of each example described in the embodiments disclosed in this application, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application in combination with the embodiments, but such implementation should not be considered to be beyond the scope of this application.

[0186] This embodiment can divide the server into functional modules according to the above method embodiment. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one module. The above integrated module can be implemented in the form of hardware. It should be noted that the division of modules in this embodiment is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0187] Fig.14 A schematic diagram of the structure of a server provided in another embodiment of the present application, in which each functional module is divided according to each function, Fig.14A possible composition diagram of the server 1400 involved in the above embodiment is shown. Fig.14 As shown, the server 1400 may include: an acquisition module 1401, a division module 1402, a calculation module 1403, a determination module 1404 and a sending module 1405;

[0188] Wherein, the acquisition module 1401 is used to acquire historical data of the electromechanical equipment to be detected;

[0189] A division module 1402 is used to divide the historical data into a training set and a test set; wherein the training set includes a first proportion of normal samples in the historical data, and the test set includes a second proportion of normal samples in the historical data and fault samples in the historical data;

[0190] The calculation module 1403 is used to perform feature calculation on each piece of data in the training set according to a predetermined number of features to obtain a first feature matrix; and perform feature calculation on each piece of data in the test set according to the predetermined number of features to obtain a second feature matrix; wherein each column of the first feature matrix and the second feature matrix is ​​a feature, and the number of columns of the first feature matrix and the second feature matrix is ​​equal to the predetermined number;

[0191] The determination module 1404 is used to determine the threshold corresponding to each feature according to the first feature matrix; determine the feature of the preset model corresponding to the electromechanical device to be detected and the threshold corresponding to the feature of the preset model according to the second feature matrix and the threshold corresponding to each feature;

[0192] The sending module 1405 is used to send the characteristics of the preset model and the corresponding threshold to the fault detection device, so that the fault detection device can perform fault detection on the electromechanical device to be detected according to the characteristics of the preset model and the corresponding threshold.

[0193] It should be noted that this application Fig.11 All relevant contents of each step involved in the illustrated method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0194] The server 1400 provided in this embodiment is used to execute the application Fig.11 The fault detection method for electromechanical equipment provided by the illustrated embodiment can therefore achieve the same effect as the above method.

[0195] Fig.15 A schematic diagram of the structure of a server provided in another embodiment of the present application, Fig.14 Compared to the server shown, the difference is that Fig.15In the server 1400 shown, the determination module 1404 may include: a matrix calculation submodule 14041, an acquisition submodule 14042, a comparison submodule 14043, a generation submodule 14044, a classification submodule 14045 and a selection submodule 14046;

[0196] The matrix calculation submodule 14041 is used to calculate the mean and standard deviation of each column of the first feature matrix;

[0197] The obtaining submodule 14042 is used to obtain a threshold value corresponding to each feature according to the above mean value and the above standard deviation.

[0198] A comparison submodule 14043 is used to compare the second feature matrix row by row with the threshold value corresponding to each feature;

[0199] The obtaining submodule 14042 is further used to obtain a result matrix according to the comparison result; and obtain a truth matrix according to the data labels of the above test set;

[0200] A generating submodule 14044 is used to generate an error number vector according to the above result matrix and the above truth value matrix;

[0201] The obtaining submodule 14042 is further used to obtain an accuracy rate vector according to the above error number vector;

[0202] The classification submodule 14045 is used to classify the predetermined number of features to obtain features of a predetermined category;

[0203] The selection submodule 14046 is used to select features of the preset model from the features of the predetermined category according to the accuracy vector, and use the threshold corresponding to the selected feature as the threshold corresponding to the feature of the preset model.

[0204] It should be noted that this application Fig.11 All relevant contents of each step involved in the illustrated method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0205] The server 1400 provided in this embodiment is used to execute the application Fig.11 The fault detection method for electromechanical equipment provided by the illustrated embodiment can therefore achieve the same effect as the above method.

[0206] It should be understood that the server 1400 may correspond to Figure 3 The server 300 shown in FIG. 1400 includes a sending module 1405 and a sending module 1405. Figure 3 The processor 310 and the communication interface 320 in the server 300 are implemented; the functions of the acquisition module 1401, the division module 1402, the calculation module 1403 and the determination module 1404 can be implemented by Figure 3 The processor 310 in the server 300 is shown implemented.

[0207] In the case of adopting an integrated unit, the server 1400 may include a processing module, a storage module, and a communication module.

[0208] The processing module can be used to control and manage the actions of the server 1400, for example, it can be used to support the server 1400 in executing the steps performed by the acquisition module 1401, the division module 1402, the calculation module 1403, the determination module 1404 and the sending module 1405. The storage module can be used to support the server 1400 in storing program codes and data, etc. The communication module can be used to support the communication between the server 1400 and other devices.

[0209] Among them, the processing module can be a processor or a controller, which can implement or execute various exemplary logic boxes, modules and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, and so on. The storage module can be a memory. The communication module can specifically be a device that interacts with other electronic devices, such as a radio frequency circuit, a Bluetooth chip and / or a Wi-Fi chip.

[0210] In one embodiment, when the processing module is a processor and the storage module is a memory, the server 1400 involved in this embodiment may be a server having Figure 3 Device of the structure shown.

[0211] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the present application. Figure 9-10 The method provided by the illustrated embodiment.

[0212] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the present application. Fig.11 The method provided by the illustrated embodiment.

[0213] The present application also provides a computer program product, which includes a computer program, which, when executed on a computer, enables the computer to execute the present application. Figure 9-10 The method provided by the illustrated embodiment.

[0214] The present application also provides a computer program product, which includes a computer program, which, when executed on a computer, enables the computer to execute the present application. Fig.11 The method provided by the illustrated embodiment.

[0215] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0216] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel 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 this application.

[0217] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0218] In several embodiments provided in the present application, any function can be stored in a computer-readable storage medium if it is implemented in the form of a software functional unit and sold or used as an independent product. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0219] The above is only a specific implementation of the present application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for detecting a fault of an electromechanical device, characterized in that: include: Collect signals from electromechanical equipment to be tested; Calculating the signal characteristics of the signal according to the characteristics of the preset model corresponding to the electromechanical device to be detected; wherein the characteristics of the preset model include characteristics selected according to the accuracy rate in each category of the predetermined category; Comparing the signal feature with a threshold value corresponding to the feature of the preset model; If there is a signal feature greater than or equal to the threshold, acquiring historical signals of the electromechanical device to be detected and historical signals of adjacent devices of the electromechanical device to be detected; The fault detection result of the electromechanical device to be detected is determined according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device.

2. The method according to claim 1, characterized in that After determining the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device, the method further includes: Obtaining the review and inspection results of the electromechanical equipment to be detected, and using the review and inspection results as true labels; Comparing the real label with the fault detection result of the electromechanical device to be detected; According to the comparison results, determine the false alarm rate and the number of missed alarms of the preset model; When the false alarm rate is greater than a predetermined false alarm rate threshold, the threshold corresponding to the feature of the preset model is updated; when the number of missed alarms is greater than or equal to a predetermined missed alarm number threshold, the feature of the preset model is updated.

3. The method according to claim 2, characterized in that When the false alarm rate is greater than a predetermined false alarm rate threshold, updating the threshold corresponding to the feature of the preset model includes: Acquire historical signals and field signals of the electromechanical equipment to be detected; Calculating a characteristic matrix of the historical signal and the field signal according to the characteristics of the preset model; Performing prior distribution fitting on the feature matrix of the historical signal by column to obtain the prior distribution of the historical signal; Sampling the prior distribution of the historical signal a predetermined number of times to obtain sampling values ​​of the prior distribution; wherein the predetermined number of times is equal to the number of the field signals; Perform error calculation on the sampling values ​​of the prior distribution and the characteristic matrix of the field signal to obtain an error matrix; Performing likelihood distribution fitting on the error matrix to obtain the likelihood distribution of the error matrix; Multiplying the prior distribution of the historical signal by the likelihood distribution of the error matrix to obtain the posterior distribution of the features of the preset model; According to the posterior distribution, the threshold corresponding to the feature of the preset model is updated.

4. The method according to claim 2, characterized in that: When the number of missed reports is greater than or equal to a predetermined missed report number threshold, updating the features of the preset model includes: Obtaining the sensitivity of each feature of the predetermined category; The features of the preset model are replaced with features with high sensitivity, where the features with high sensitivity include features in each type of features whose sensitivity is higher than the current features of the preset model.

5. The method according to claim 1, characterized in that Determining the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device includes: When the trend information of the historical signal presents an upward trend, and the trend information of the historical signal of the adjacent device does not present an upward trend, it is determined that the electromechanical device to be detected fails.

6. The method according to claim 1, characterized in that Before determining the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device, the method further includes: Calculating the signal characteristics of the historical signal according to the characteristics of the preset model; According to the signal characteristics of the historical signal, trend information of the historical signal is obtained.

7. The method according to claim 1, characterized in that Before determining the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device, the method further includes: Calculating the characteristics of the historical signal of the adjacent device according to the characteristics of the preset model; According to the characteristics of the historical signals of the neighboring devices, trend information of the historical signals of the neighboring devices is obtained.

8. The method according to any one of claims 1 to 7, characterized in that: The predetermined category of features includes one or a combination of the following: impact features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features.

9. A method for detecting a fault in an electromechanical device, characterized in that: include: Obtain historical data of the electromechanical equipment to be tested; Dividing the historical data into a training set and a test set; wherein the training set includes a first proportion of normal samples in the historical data, and the test set includes a second proportion of normal samples in the historical data and fault samples in the historical data; According to a predetermined number of features, feature calculation is performed on each piece of data in the training set to obtain a first feature matrix; and according to the predetermined number of features, feature calculation is performed on each piece of data in the test set to obtain a second feature matrix; wherein each column of the first feature matrix and the second feature matrix is ​​a feature, and the number of columns of the first feature matrix and the second feature matrix is ​​equal to the predetermined number; Determine a threshold value corresponding to each feature according to the first feature matrix; Determining, according to the second feature matrix and the threshold corresponding to each feature, the feature of the preset model corresponding to the electromechanical device to be detected and the threshold corresponding to the feature of the preset model; The characteristics of the preset model and the corresponding threshold are sent to a fault detection device, so that the fault detection device performs fault detection on the electromechanical device to be detected according to the characteristics of the preset model and the corresponding threshold.

10. The method according to claim 9, characterized in that The determining, according to the first feature matrix, a threshold value corresponding to each feature comprises: Calculate the mean and standard deviation of each column of the first feature matrix; A threshold corresponding to each feature is obtained according to the mean and the standard deviation.

11. The method according to claim 9, characterized in that Determining the feature of the preset model corresponding to the electromechanical device to be detected and the threshold value corresponding to the feature of the preset model according to the second feature matrix and the threshold value corresponding to each feature includes: Comparing the second feature matrix row by row with the threshold value corresponding to each feature; According to the comparison results, a result matrix is ​​obtained; According to the data labels of the test set, a true value matrix is ​​obtained; Generate an error number vector according to the result matrix and the truth value matrix; Obtaining an accuracy rate vector according to the error number vector; Classifying the predetermined number of features to obtain features of a predetermined category; According to the accuracy vector, a feature of a preset model is selected from the features of the predetermined category, and a threshold corresponding to the selected feature is used as a threshold corresponding to the feature of the preset model.

12. The method according to claim 11, characterized in that The predetermined category of features includes one or a combination of the following: impact features, harmonic features, modulation features, energy features, spectrum distribution features, and signal fluctuation features.

13. A fault detection device for electromechanical equipment, characterized in that: include: An acquisition module, used for acquiring signals of electromechanical equipment to be detected; A calculation module, used to calculate the signal characteristics of the signal according to the characteristics of the preset model corresponding to the electromechanical device to be detected; wherein the characteristics of the preset model include characteristics selected according to the accuracy rate in each category of characteristics of a predetermined category; A comparison module, used for comparing the signal feature with a threshold value corresponding to the feature of the preset model; an acquisition module, configured to acquire, when there is a signal feature greater than or equal to the threshold, a historical signal of the electromechanical device to be detected and a historical signal of an adjacent device of the electromechanical device to be detected; The determination module is used to determine the fault detection result of the electromechanical device to be detected according to the trend information of the historical signal and the trend information of the historical signal of the adjacent device.

14. A fault detection device for electromechanical equipment, characterized in that: include: An acquisition module, used for acquiring historical data of the electromechanical equipment to be detected; A division module, used to divide the historical data into a training set and a test set; wherein the training set includes a first proportion of normal samples in the historical data, and the test set includes a second proportion of normal samples in the historical data and fault samples in the historical data; A calculation module, configured to perform feature calculation on each piece of data in the training set according to a predetermined number of features to obtain a first feature matrix; and perform feature calculation on each piece of data in the test set according to the predetermined number of features to obtain a second feature matrix; wherein each column of the first feature matrix and the second feature matrix is ​​a feature, and the number of columns of the first feature matrix and the second feature matrix is ​​equal to the predetermined number; A determination module, configured to determine a threshold value corresponding to each feature according to the first feature matrix; and to determine a feature of a preset model corresponding to the electromechanical device to be detected and a threshold value corresponding to the feature of the preset model according to the second feature matrix and the threshold value corresponding to each feature; The sending module is used to send the characteristics of the preset model and the corresponding threshold value to the fault detection device, so that the fault detection device can perform fault detection on the electromechanical device to be detected according to the characteristics of the preset model and the corresponding threshold value.

15. A fault detection device, characterized in that: include: one or more processors; Memory; Multiple applications; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the fault detection device, enable the fault detection device to perform the method according to any one of claims 1 to 8.

16. A server, characterized in that: include: one or more processors; Memory; Multiple applications; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the server, cause the server to perform the method according to any one of claims 9 to 12.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the method according to any one of claims 1 to 8.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the method according to any one of claims 9 to 12.

Images