Industrial equipment fault prediction and health management method and system

Abstract

The invention discloses an industrial equipment fault prediction and health management method and system, and relates to the technical field of industrial control, and the method comprises the steps: sending a time synchronization instruction to equipment, recording a time synchronization moment, setting a power-off waiting time length, calculating an RTC time error after the equipment is powered on again, and recording environment parameters; acquiring the RTC time of the equipment and the time drift distance of the last time synchronization moment in real time, updating an abnormal criterion based on the time drift distance and an RTC time error, and entering a time correction mode; the method comprises the following steps: identifying a key event based on a machine learning model, extracting a time anchor point of the key event, constructing an LSTM model based on environmental parameters and RTC time, predicting estimated time intervals, respectively correcting the estimated time intervals including the time anchor point and not including the time anchor point, and updating an RTC time error, thereby realizing adaptive equipment time anomaly detection and repair.

Description

Technical Field

[0001] The present application relates to the field of industrial control technology, and in particular to a method and system for industrial equipment fault prediction and health management. Background Art

[0002] With the continuous advancement of industrial automation, smart manufacturing, and the Industrial Internet of Things (IIoT), a large number of devices are continuously operating at high intensity in key areas such as production lines, energy storage, and transportation. In this context, continuous monitoring of device operating status, fault trend prediction, and health assessment have become core technical requirements for ensuring system safety, stability, and cost-effectiveness.

[0003] Currently, existing industrial equipment fault prediction and health management systems primarily utilize distributed sensors to collect operating parameters such as voltage, current, temperature, insulation resistance, and relay status. These systems then use data processing algorithms to determine health status and issue fault warnings. These systems typically include multiple submodules, such as chip self-test, voltage / current / temperature accuracy testing, and communication interface verification. Some solutions also integrate cloud-based management platforms or edge computing capabilities to support more complex model analysis. Time information is fundamental to each of these core functional modules. During operation, equipment may experience system time offsets, resets, or drift due to abnormalities such as power outages, restarts, and communication interruptions. This can lead to log corruption, data discontinuities, and analysis errors.

[0004] Chinese patent application publication number CN117687333A discloses a BLVDS-based IO bus time synchronization method for an industrial control system, comprising the following steps: a master device obtains reference time synchronization information and latches the reference time synchronization information; performs a time compensation algorithm to obtain a precise time compensation value; reads the latched reference time synchronization information and accumulates the precise time compensation value to obtain a time synchronization frame; the master device sends the time synchronization frame to the slave device, which resets the time compensation algorithm to perform time compensation calculations and accumulates data in the time synchronization frame for time synchronization; without adding a new time synchronization bus, the master control device is used to decide data transmission and time synchronization permissions on the BLVDS bus, thereby ensuring the stability of the time synchronization mechanism; and the time synchronization can be changed to a broadcast or precise mode according to actual conditions, making it applicable to more scenarios and more flexible and practical.

[0005] However, existing technologies cannot solve the problem of system time loss caused by device power outage or soft restart. The timestamps of collected data may be distorted, seriously affecting the availability and timing consistency of data. Summary of the Invention

[0006] This application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, one purpose of this application is to propose an industrial equipment fault prediction and health management method and system, which realizes adaptive equipment time anomaly detection and repair.

[0007] One aspect of the present application provides a method for industrial equipment fault prediction and health management, comprising:

[0008] Step S100: Send a time synchronization instruction to the device and record the time synchronization moment, set the power-off waiting time, calculate the RTC time error after the device is powered on again, and record the environmental parameters;

[0009] Step S200: obtaining the time drift between the device RTC time and the most recent time synchronization in real time, updating the abnormality judgment criteria based on the time drift and the RTC time error, and entering the time correction mode;

[0010] Step S300: Identify key events based on the machine learning model, extract their time anchor points, build an LSTM model based on environmental parameters and RTC time, predict the estimated time interval, and correct the estimated time intervals that include and exclude the time anchor points, and update the RTC time error.

[0011] The specific method of sending a time synchronization instruction to the device and recording the time synchronization moment, setting the power-off waiting time, calculating the RTC time error after the device is powered on again, and recording the environmental parameters is as follows:

[0012] Step S110: Send a time synchronization instruction to the device via the CAN bus interface, and set the current master time T m Write to the device RTC chip and record the time T s ;

[0013] Step S120: Set the power-off waiting time Δt according to the device type and working conditions p ;

[0014] Step S130: After the power-off waiting period is over, power on again and record the power-on time T r , and read the device RTC time T d , record the reading time of the device RTC time;

[0015] Step S140: Calculate the RTC time error ΔT between the power-on time and the device RTC time rtc , preset RTC error threshold, if the RTC time error is less than or equal to the RTC error threshold, the RTC is judged to be normal, otherwise it is marked as clock abnormal;

[0016] Step S150: Recording environmental parameters before and after power failure; the environmental parameters include current I, power supply voltage U, temperature t, and humidity h;

[0017] Step S160: saving the RTC time error as an initial error reference in the NVM memory;

[0018] The specific method of obtaining the time drift between the device RTC time and the most recent time synchronization in real time, and updating the abnormality criterion based on the time drift and the RTC time error to enter the time correction mode is as follows:

[0019] Step S210: Obtain device RTC time T in real time d , and calculate its difference with the most recent successful time T s The time drift between

[0020] Step S220: Read the RTC time error stored in the NVM memory Combined with the time drift, the time anomaly criterion is updated and the time anomaly is judged using the time anomaly criterion;

[0021] Step S230: After detecting m time anomalies, the system enters the time correction mode. However, if the time anomaly recovers before entering the time correction mode and the duration exceeds T n seconds, the abnormal mark is removed and the system is considered to have been automatically repaired;

[0022] The time anomaly criterion includes three conditions. If any one of the conditions is met, it is identified as abnormal data. Condition 1 is that if the time drift ΔT j Smaller than the RTC time error read from NVM memory The time is judged to be abnormal; the second condition is if the time drift ΔT j The absolute difference between the RTC time error read from the NVM memory is greater than the adaptive time jump threshold ε j , then the time is determined to be abnormal; condition three is to calculate the difference between the current time drift and the time drift change rate, if the absolute value of the difference between the difference and the RTC time error is greater than the adaptive time difference change threshold ε for M consecutive times g , then the time is judged to be abnormal;

[0023] The time drift change rate is the value obtained by dividing the difference between two adjacent time drifts by the difference between two adjacent reading times of the device RTC time;

[0024] The adaptive time jump threshold ε j and adaptive time difference change threshold ε gThe calculation method is as follows: obtain historical environmental parameters and corresponding timestamps, and record the time drift at the same time; divide each environmental parameter into H groups according to its value range; for each environmental parameter group, calculate the statistical characteristics of its time jump amplitude and time difference change rate; for each environmental parameter group, establish a mapping function of the adaptive time jump threshold and the adaptive time difference change threshold according to its corresponding statistical characteristics; collect the current environmental parameters in real time, and calculate the corresponding adaptive time jump threshold and adaptive time difference change threshold according to the statistical characteristics of the group to which it belongs;

[0025] The specific method of identifying key events based on the machine learning model and extracting their time anchor points is as follows:

[0026] Step S310: Using environmental parameters as input features and outputting them as event types to train a machine learning model; the environmental parameters include current I, power supply voltage U, temperature t, and humidity h;

[0027] Step S320: Collect environmental parameters during the operation of the equipment, according to a fixed time window Δt w Divide and obtain the environmental parameter sequence {S1, S2, ..., S N};

[0028] Step S330: For each environmental parameter segment S n , extract its current change rate ΔI n , power supply voltage change rate ΔU n and temperature change rate Δt n , to determine whether it is greater than the preset current mutation threshold ε I , voltage mutation threshold ε U , temperature mutation threshold ε t If the current change rate is greater than the current mutation threshold or the power supply voltage change rate is greater than the voltage mutation threshold or the temperature change rate is greater than the temperature mutation threshold, then the environmental parameter segment S n Input into the trained machine learning model to identify event types;

[0029] Step S340: For the environmental parameter segment S identified as a key event type n , extract the reading time of the device RTC time of the environmental parameter fragment as the time anchor point T n , and record its event type E n ;

[0030] The specific method of constructing an LSTM model based on environmental parameters and RTC time, predicting an estimated time interval, and respectively correcting the estimated time interval including and excluding the time anchor point is as follows:

[0031] Step S350: For the environmental parameter segment sequence {S a ,S a+1 ,…,S b}, extract its original device RTC time series Construct feature matrix X based on environmental parameters and device RTC time series;

[0032] Step S360: Input the feature matrix X into the pre-trained LSTM model to obtain the estimated time interval to which each environmental parameter segment belongs;

[0033] Step S370: If the end time of the i-th estimated time interval is and the start time of the i+1th estimated time interval If the absolute value of the difference is greater than the time tolerance ε, the start time of the i+1th estimated time interval is set to Adjust to the end time of the i-th estimated time interval Add the value of the time offset δ;

[0034] Step S380: Calibrate the estimated time interval based on the time anchor point and the original device RTC time to obtain a corrected time corresponding to the estimated time interval;

[0035] Step S390: Compare the corrected time with the device RTC time and calculate the difference between the two as the new RTC time error And stored in NVM memory.

[0036] The specific method of calibrating the estimated time interval based on the time anchor point and the original device RTC time to obtain the corrected time of the corresponding time interval and update the RTC time error is as follows:

[0037] Step S381: If the time anchor point T n If it falls within the estimated time interval, the start time of the corrected time of the estimated time interval is the difference between the time anchor point and half of the fixed time window, and the end time is the sum of the time anchor point and half of the fixed time window;

[0038] Step S382: For an estimated time interval without a time anchor point, for each environmental parameter segment, calculate a first difference between the original device RTC time and the start time of the estimated time interval, adjust the start time of the revised time of the estimated time interval to the weighted sum of the start time and the first difference, and adjust the end time of the revised time of the estimated time interval to the weighted sum of the end time and the first difference;

[0039] Step S383: Smoothing the corrected time, and mapping the smoothed corrected time back to the original device RTC time sequence to obtain the final corrected time sequence.

[0040] One aspect of the present application provides an industrial equipment fault prediction and health management system, comprising:

[0041] The periodic time test module is used to send time synchronization instructions to the device and record the synchronization time, set the power-off waiting time, calculate the RTC time error after the device is powered on again, and record environmental parameters;

[0042] The time anomaly judgment module is used to obtain the time drift between the device RTC time and the most recent time synchronization in real time, update the anomaly judgment criteria based on the time drift and RTC time error, and enter the time correction mode;

[0043] The time correction module is used to identify key events based on the machine learning model, extract their time anchor points, build an LSTM model based on environmental parameters and RTC time, predict the estimated time interval, and correct the estimated time interval containing and excluding time anchor points respectively, and update the RTC time error.

[0044] One aspect of the present application provides a readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor to execute steps in a method for fault prediction and health management of industrial equipment.

[0045] The industrial equipment fault prediction and health management method and system proposed in this application have the following advantages over existing technologies:

[0046] This application actively obtains the device's RTC initial error benchmark and the current environmental parameters through periodic time synchronization and power-off tests to evaluate the device's RTC timing accuracy. These parameters are used to optimize the adaptive threshold in subsequent steps to improve the robustness of anomaly recognition.

[0047] During the normal operation of the device, the present application monitors the drift of the device time in real time, and combines the RTC error benchmark obtained from the first step and the real-time environmental parameters to dynamically update the abnormality judgment criteria to improve the accuracy of abnormality identification. When a persistent abnormality is detected, it triggers the entry into the time repair mode, realizing real-time monitoring of device time abnormalities. By dynamically reading the RTC time error stored in the NVM, and combining the adaptive time jump threshold and the adaptive time difference change threshold, abnormal problems such as time jump and time difference change rate can be discovered in time. The introduction of the adaptive threshold is an important innovation. It dynamically adjusts the sensitivity of the abnormality judgment based on the environmental parameters, significantly improves the accuracy and robustness of abnormality identification, and enables the solution to better adapt to different environmental conditions and working conditions. In addition, considering the automatic recovery of the abnormal state avoids frequent false alarms, forming an adaptive time anomaly detection mechanism.

[0048] This application introduces a machine learning model to intelligently identify key events that may cause time anomalies and extract time anchor points. An LSTM model is constructed based on environmental parameters and RTC time to dynamically estimate the true time interval. Time anchor points are used to adaptively calibrate the LSTM model's estimated results, significantly improving the accuracy and reliability of time repair. By mapping the corrected time to the device's RTC time, a new RTC time error is calculated and stored back in NVM, providing a more accurate baseline for the next time correction, forming a self-learning and continuously optimizing correction mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A method flow chart of the industrial equipment fault prediction and health management method provided in this application;

[0050] Figure 2 Flowchart of the time anomaly judgment method provided in this application;

[0051] Figure 3 Flowchart of the time anchor point extraction method provided by this application;

[0052] Figure 4 This is a functional module diagram of the industrial equipment fault prediction and health management system provided in this application. DETAILED DESCRIPTION

[0053] To better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely descriptions of exemplary embodiments of the present application and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0054] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The accompanying drawings are for illustration only and are not drawn strictly to scale. As used herein, the terms "substantially," "approximately," and similar terms are used to indicate approximate values, not degrees, and are intended to illustrate inherent deviations in measurements or calculations that would be recognized by a person of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these steps would occur in actual operation, unless otherwise specified or inferred from the context.

[0055] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0056] Unless otherwise defined, all words used herein (including engineering terms and scientific and technological terms) have the same meaning as commonly understood by those skilled in the art to which this application belongs. It should also be understood that, unless otherwise specified in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as they do in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0057] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0058] Example 1

[0059] like Figure 1 As shown, the industrial equipment failure prediction and health management method provided by this application includes:

[0060] Step S100: Send a time synchronization instruction to the device and record the time synchronization moment, set the power-off waiting time, calculate the RTC time error after the device is powered on again, and record the environmental parameters;

[0061] The specific method of sending a time synchronization instruction to the device and recording the time synchronization moment, setting the power-off waiting time, calculating the RTC time error after the device is powered on again, and recording the environmental parameters is as follows:

[0062] Step S110: Send a time synchronization instruction to the device via the CAN bus interface, and set the current master time T m Write to the device RTC chip and record the time T s ;

[0063] The master control time T m Refers to the current time of the main control system, which is used to write the device RTC to synchronize it with the main control system; the time T s The system time point at which the master time is written into the device RTC. At this moment, the device RTC time T dShould be consistent with the master time T m synchronous.

[0064] Step S120: Set the power-off waiting time Δt according to the device type and working conditions p ;

[0065] The power-off waiting time is used to test the timekeeping accuracy of the RTC in the case of power failure, and its value range is [30,120] seconds;

[0066] Step S130: After the power-off waiting period is over, power on again and record the power-on time T r , and read the device RTC time T d , record the reading time of the device RTC time;

[0067] The power-on time is used to calculate the time error during the RTC power-off period, and the device RTC time is the RTC time read after power-on.

[0068] Step S140: Calculate the RTC time error ΔT between the power-on time and the device RTC time rtc , preset RTC error threshold, if the RTC time error is less than or equal to the RTC error threshold, the RTC is judged to be normal, otherwise it is marked as clock abnormal;

[0069] The RTC time error is the time error of the RTC during the power-off period, and its calculation formula is:

[0070] ΔT rtc =|T d -T r |

[0071] The preset RTC error threshold is used to determine whether the RTC is normal. Its value is set by those skilled in the art based on experience, and is generally set to 1 to 5 seconds:

[0072] Step S150: Recording environmental parameters before and after power failure; the environmental parameters include current I, power supply voltage U, temperature t, and humidity h;

[0073] The current before and after power failure is expressed as I s , I e , the power supply voltage before and after power failure is expressed as U s 、U e , the temperature before and after power failure is expressed as t s , t e , the humidity before and after power failure is expressed as h s 、h e The environmental parameters before and after the power outage are used to optimize the adaptive threshold in subsequent steps to improve the robustness of anomaly recognition.

[0074] The environmental parameters before and after the power outage are used as a reference for the adaptive threshold in step S200;

[0075] Step S160: saving the RTC time error as an initial error reference in the NVM memory;

[0076] The NVM memory is a non-volatile memory;

[0077] The above steps obtain the device's RTC initial error benchmark and the environmental parameters at the time through periodic time synchronization and power-off testing, evaluate the device's RTC timing accuracy, and obtain the initial value of the RTC time error, providing basic information for continuous anomaly identification in step S200.

[0078] Step S200: obtaining the time drift between the device RTC time and the most recent time synchronization in real time, updating the abnormality judgment criteria based on the time drift and the RTC time error, and entering the time correction mode;

[0079] like Figure 2 As shown, the specific method of obtaining the time drift between the device RTC time and the most recent time synchronization moment in real time and updating the abnormality criterion based on the time drift and the RTC time error to enter the time correction mode is as follows:

[0080] Step S210: Obtain device RTC time T in real time d , and calculate its difference with the most recent successful time T s The time drift between

[0081] The calculation formula of the time drift is:

[0082] ΔT j =T d -T s

[0083] Where, ΔT j The device RTC time T d Time T of the most recent successful match s The time drift between

[0084] Step S220: Read the RTC time error stored in the NVM memory Combined with the time drift, the time anomaly criterion is updated and the time anomaly is judged using the time anomaly criterion;

[0085] The time anomaly criterion includes three conditions. If any one of the conditions is met, it is identified as abnormal data. Condition 1 is that if the time drift ΔT j Smaller than the RTC time error read from NVM memory The time is judged to be abnormal; the second condition is if the time drift ΔTj The absolute difference between the RTC time error read from the NVM memory is greater than the adaptive time jump threshold ε j , then the time is determined to be abnormal; condition three is to calculate the difference between the current time drift and the time drift change rate, if the absolute value of the difference between the difference and the RTC time error is greater than the adaptive time difference change threshold ε for M consecutive times g , then the time is judged to be abnormal;

[0086] The time anomaly criterion is expressed as follows:

[0087] Condition one:

[0088]

[0089] Condition two:

[0090]

[0091] Condition 3: satisfied M times consecutively

[0092] Where, ΔT j ' is the rate of change of the device time drift; the rate of change of the time drift is the value of the difference between two adjacent time drifts divided by the difference between two adjacent reading times of the device RTC time;

[0093] The time drift change rate reflects the change in the speed of device time drift and is an important indicator of time anomalies. When calculating the device time drift change rate, the read time refers to the moment when the device RTC time is read, that is, the starting timestamp of the device RTC time is obtained.

[0094] Specifically, let the time of reading the RTC time of the i-th device be T i , the corresponding device RTC time is The last time was The calculation formula for the i-th device time drift is:

[0095]

[0096] The reading times of two consecutive device RTC times are T i and T i+1 , and the corresponding time drifts are and The calculation formula for the time drift change rate is:

[0097]

[0098] Where, ΔT j ' (i)Indicates the rate of change of drift between the i-th and i+1-th reading times, The time drift rate calculation formula reflects the change in the device's time drift rate and is an important indicator in time anomaly detection.

[0099] To understand this formula more intuitively, we can expand it into:

[0100]

[0101] In the above formula, Indicates the difference between two consecutive reads of the device RTC time. Represents the difference between two consecutive time intervals, (T i+1 -T i ) represents the time interval between two consecutive readings of the device RTC time.

[0102] The time drift rate reflects the change of the device's time drift speed. If ΔT j ' (i) >0, indicating that the device time drift speed is accelerating; if ΔT j ' (i) <0, indicating that the device time drift speed is slowing down; if ΔT j ' (i) =0, indicating that the device time drift speed remains unchanged.

[0103] In practical applications, the absolute value of the rate of change of the device time drift |ΔT j ' (i) | Perform monitoring and abnormality judgment, because an increase or decrease in drift speed may indicate an abnormality in the device time.

[0104] The adaptive time jump threshold ε j and adaptive time difference change threshold ε g The calculation method is as follows: obtain historical environmental parameters and corresponding timestamps, and record the time drift at the same time; divide each environmental parameter into H groups according to its value range; for each environmental parameter group, calculate the statistical characteristics of its time jump amplitude and time difference change rate; for each environmental parameter group, establish a mapping function of the adaptive time jump threshold and the adaptive time difference change threshold according to its corresponding statistical characteristics; collect the current environmental parameters in real time, and calculate the corresponding adaptive time jump threshold and adaptive time difference change threshold according to the statistical characteristics of the group to which it belongs;

[0105] Optionally, the adaptive time jump threshold ε j and adaptive time difference change threshold ε gThe calculation method further includes: during the actual operation of the device, continuously collecting environmental parameters and periodically updating the mapping functions of the adaptive time jump threshold and the adaptive time difference change threshold.

[0106] The time jump amplitude is expressed as |ΔT j |, the time difference change rate is expressed as |ΔT j ′|, and the statistical characteristics of the time jump amplitude and the time difference change rate include the mean and the standard deviation.

[0107] The expression of the mapping function of the adaptive time jump threshold is:

[0108] ε j =f j (t, h, U)=μ j +k j ×σ j

[0109] The expression of the mapping function of the adaptive time difference change threshold is:

[0110] ε g =f g (t, h, U)=μ g +k g ×σ g

[0111] Where, μ j and σ j are respectively the mean and the standard deviation of the time jump amplitude, μ g and σ g are respectively the mean and the standard deviation of the time difference change rate, k j and k g are adjustable coefficients used to control the sensitivity of the adaptive time jump threshold and the adaptive time difference change threshold; for different groups of environmental parameters, the calculated statistical characteristics are different, so the adaptive time jump threshold and the adaptive time difference change threshold are also different. In actual use, according to the group where the environmental parameters corresponding to the current time are located, calculate the corresponding adaptive time jump threshold and adaptive time difference change threshold.

[0112] Exemplarily, according to the value range of each environmental parameter, it is divided into H groups. According to the temperature range, the temperature is divided into a low temperature group (t≤10°C), a normal temperature group (10°C<t≤30°C), and a high temperature group (t>30°C). According to the humidity range, it is divided into a dry group (h≤30%), a moderate group (30%<h≤70%), and a humid group (h>70%). According to the power supply voltage, it is divided into an undervoltage group (U≤22V), a normal group (22V<U≤26V), and an overvoltage group (U>26V);

[0113] Step S230: After detecting m time anomalies, the system enters the time correction mode. However, if the time anomaly recovers before entering the time correction mode and the duration exceeds T n seconds, the abnormal mark is removed and the system is considered to have been automatically repaired;

[0114] Preferably, m is equal to 13, T n The value is 60300 seconds;

[0115] The purpose of the above steps is to monitor device time anomalies in real time. By dynamically reading the RTC time error stored in NVM and combining it with the adaptive time jump threshold and adaptive time difference change threshold, problems such as time jumps and abnormal time difference change rates can be detected promptly, triggering the time correction process in subsequent steps. At the same time, the above steps also take into account the automatic recovery of abnormal states to avoid frequent false alarms.

[0116] Steps S100 and S200 are tightly integrated to form an adaptive time anomaly detection mechanism, providing reliable triggering conditions for subsequent time corrections. The introduction of adaptive time jump thresholds and adaptive time difference change thresholds further improves the accuracy and robustness of anomaly identification, making the entire solution more adaptable to diverse environmental conditions and operating conditions.

[0117] Step S300: Identify key events based on the machine learning model, extract their time anchor points, build an LSTM model based on environmental parameters and RTC time, predict the estimated time interval, and correct the estimated time intervals that include and exclude the time anchor points, and update the RTC time error.

[0118] like Figure 3 As shown, the specific method of identifying key events based on the machine learning model and extracting their time anchor points is:

[0119] Step S310: Using environmental parameters as input features and outputting them as event types to train a machine learning model; the environmental parameters include current I, power supply voltage U, temperature t, and humidity h;

[0120] The machine learning model analyzes environmental parameters such as the device's current, power supply voltage, and temperature to identify key events that may cause time anomalies, such as power-on and charging start, and uses them as reference time anchor points for time correction.

[0121] The training method of the machine learning model is as follows: obtaining historical environmental parameters and corresponding event type labels, constructing a training data set, randomly dividing the training data set into a training set and a test set in a ratio of 8:2, selecting a machine learning algorithm, which may be a decision tree algorithm; inputting the training set into the selected machine learning algorithm for training, using the test set to evaluate the trained model, and adjusting the model hyperparameters based on the evaluation results until satisfactory performance is obtained.

[0122] For the decision tree algorithm, use information gain or Gini index to select the optimal split attribute;

[0123] Step S320: Collect environmental parameters during the operation of the equipment, according to a fixed time window Δt w Divide and obtain the environmental parameter sequence {S1, S2, ..., S N};

[0124] The value of the fixed time window is the inverse of the sampling frequency of the device;

[0125] Step S330: For each environmental parameter segment S n , extract its current change rate ΔI n , power supply voltage change rate ΔU n and temperature change rate Δt n , to determine whether it is greater than the preset current mutation threshold ε I , voltage mutation threshold ε U , temperature mutation threshold ε t If the current change rate is greater than the current mutation threshold or the power supply voltage change rate is greater than the voltage mutation threshold or the temperature change rate is greater than the temperature mutation threshold, then the environmental parameter segment S n Input into the trained machine learning model to identify event types;

[0126] Step S340: For the environmental parameter segment S identified as a key event type n , extract the reading time of the device RTC time of the environmental parameter fragment as the time anchor point T n , and record its event type E n ;

[0127] The key event types include power-on events and charging start events.

[0128] The time anchor point T n It reflects the actual moment of the device RTC time corresponding to the reading of the environmental parameter segment, rather than the segment start RTC time itself.

[0129] The specific method of constructing an LSTM model based on environmental parameters and RTC time, predicting an estimated time interval, and respectively correcting the estimated time interval including and excluding the time anchor point is as follows:

[0130] Step S350: For the environmental parameter segment sequence {S a ,S a+1 ,…,S b}, extract its original device RTC time series Construct feature matrix X based on environmental parameters and device RTC time series;

[0131] The feature matrix can be expressed as Where p = b-a + 1, p is the number of time points in the device RTC time series, and q is the dimension of the environmental parameter in the environmental parameter segment obtained at each time point. Added as an additional feature to the last column of the feature matrix X, that is, the q+1th column.

[0132] Step S360: Input the feature matrix X into the pre-trained LSTM model to obtain the estimated time interval to which each environmental parameter segment belongs;

[0133] The estimated time interval sequence to which each environmental parameter segment belongs can be expressed as:

[0134] The estimated time interval It represents the time interval of the i-th environmental parameter segment estimated by the LTSM model, which is a correction to the device RTC time;

[0135] The training method of the LSTM model is as follows: obtaining the accurate values ​​of the environmental parameters and the corresponding RTC time of the historical time, taking the RTC time as the feature label of the corresponding environmental parameter segment, constructing a feature matrix based on the environmental parameters and RTC time, taking the feature matrix corresponding to the environmental parameter segment after the time anomaly as input, normalizing the feature matrix, scaling each feature to the interval [0,1], and outputting the feature vector X for each environmental parameter segment. i and the corresponding RTC time; randomly divide the data into training and test sets according to a certain ratio; design the number of layers and hidden units of the LSTM model based on the feature dimension and time steps, input the training set into the LSTM model, perform end-to-end training, use mean square error as the loss function, adopt the stochastic gradient descent optimization algorithm to update the model parameters, use the test set to evaluate the trained LSTM model and adjust the hyperparameters, and complete the training when the loss function converges to the minimum value.

[0136] Step S370: If the end time of the i-th estimated time interval is and the start time of the i+1th estimated time interval If the absolute value of the difference is greater than the time tolerance ε, the start time of the i+1th estimated time interval is set to Adjust to the end time of the i-th estimated time interval Add the value of the time offset δ;

[0137] The time tolerance represents the tolerance for the continuity of adjacent estimated time intervals, which is determined according to the sampling frequency, clock accuracy and real-time requirements of the device. The sampling frequency is the reciprocal of the theoretical time interval between two adjacent environmental parameter segments, that is, the fixed time window Δt w The clock accuracy takes into account the maximum drift rate v of the device clock max The maximum drift rate is obtained based on the measured data of the device clock. The maximum tolerance offset is obtained by multiplying the maximum drift rate and the theoretical time interval. The amplification factor k is selected according to the real-time requirements of the application scenario. f , amplify the maximum tolerance offset by k f times the value as the time tolerance;

[0138] The calculation formula of the time tolerance is: ε = k f ×v max ×Δt w ; Preferably, the amplification factor ranges from 2 to 5; for example, the sampling frequency of the device is 100 Hz, the maximum clock drift rate is 0.01, and the amplification factor is 3, then the time tolerance value is That is 0.0003 seconds;

[0139] The time offset is used to adjust the discontinuous estimated time interval to make it continuous, so the determination of its value takes into account both the continuity and authenticity of the estimated time; the method for determining the time offset is: calculating the time interval between adjacent estimated time intervals, that is, the end time of the i-th estimated time interval and the start time of the i+1th estimated time interval If the time interval is greater than the time tolerance, it means that the estimated time intervals of the i-th and i+1-th intervals are discontinuous and need to be adjusted. The difference between the start time of the i+1-th interval and the sum of the end time of the i-th interval and the time tolerance is calculated as the time offset.

[0140] The calculation formula of the time offset is:

[0141] Step S380: Calibrate the estimated time interval based on the time anchor point and the original device RTC time to obtain a corrected time corresponding to the estimated time interval;

[0142] The specific method of calibrating the estimated time interval based on the time anchor point and the original device RTC time to obtain the corrected time of the corresponding time interval and update the RTC time error is as follows:

[0143] Step S381: If the time anchor point T n If it falls within the estimated time interval, the start time of the corrected time of the estimated time interval is the difference between the time anchor point and half of the fixed time window, and the end time is the sum of the time anchor point and half of the fixed time window;

[0144] The modified time of the time interval includes the start time and the end time of the modified time;

[0145] The calculation formulas for the start time and end time of the corrected time of the time interval are:

[0146]

[0147] Step S382: For an estimated time interval without a time anchor point, for each environmental parameter segment, calculate a first difference between the original device RTC time and the start time of the estimated time interval, adjust the start time of the revised time of the estimated time interval to the weighted sum of the start time and the first difference, and adjust the end time of the revised time of the estimated time interval to the weighted sum of the end time and the first difference;

[0148] Step S383: Smoothing the corrected time, and mapping the smoothed corrected time back to the original device RTC time sequence to obtain the final corrected time sequence;

[0149] Step S390: Compare the corrected time with the device RTC time and calculate the difference between the two as the new RTC time error And stored in NVM memory.

[0150] The above steps utilize machine learning and LSTM models to intelligently identify key events based on environmental parameters and dynamically estimate the true time interval. Time anchors are used to adaptively calibrate the estimated results, improving the accuracy and reliability of time correction. A mapping is established between the corrected time and the device's RTC time, and a new RTC time error is calculated for the next time correction.

[0151] Example 2

[0152] like Figure 4 As shown, the industrial equipment failure prediction and health management system provided by this application includes:

[0153] The periodic time test module is used to send time synchronization instructions to the device and record the synchronization time, set the power-off waiting time, calculate the RTC time error after the device is powered on again, and record environmental parameters;

[0154] The time anomaly judgment module is used to obtain the time drift between the device RTC time and the most recent time synchronization in real time, update the anomaly judgment criteria based on the time drift and RTC time error, and enter the time correction mode;

[0155] The time correction module is used to identify key events based on the machine learning model, extract their time anchor points, build an LSTM model based on environmental parameters and RTC time, predict the estimated time interval, and correct the estimated time interval containing and excluding time anchor points respectively, and update the RTC time error.

[0156] Example 3

[0157] According to one embodiment of the present application, a readable storage medium is also provided. A computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, the industrial equipment fault prediction and health management method according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0158] In addition, according to the embodiment of the present application, the process described with reference to the flowchart above can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application, such as: sending a time synchronization instruction to the device and recording the time synchronization moment, setting a power-off waiting time, calculating the RTC time error after the device is powered on again, and recording environmental parameters; obtaining the time drift between the device RTC time and the most recent time synchronization moment in real time, updating the abnormality judgment based on the time drift and the RTC time error, and entering the time correction mode; identifying key events based on a machine learning model, extracting their time anchor points, building an LSTM model based on environmental parameters and RTC time, predicting an estimated time interval, and correcting the estimated time intervals containing time anchor points and not containing time anchor points respectively, and updating the RTC time error. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0159] The methods, apparatus, and devices of the present application may be implemented in many ways. For example, the methods, apparatus, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the methods according to the present application. Therefore, the present application also covers recording media that store programs for executing the methods according to the present application.

[0160] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0161] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Industrial equipment failure prediction and health management method, characterized in that: include: Send a time synchronization command to the device and record the synchronization time, set the power-off wait time, calculate the RTC time error after the device is powered on again, and record the environmental parameters; Obtain the time drift between the device RTC time and the most recent time synchronization in real time, update the abnormality judgment criteria based on the time drift and RTC time error, and enter the time correction mode; Based on the machine learning model, key events are identified and their time anchor points are extracted. An LSTM model is built based on environmental parameters and RTC time to predict the estimated time interval. The estimated time intervals containing and excluding time anchor points are corrected separately, and the RTC time error is updated.

2. The method for industrial equipment failure prediction and health management according to claim 1, wherein: The specific method of sending a time synchronization instruction to the device and recording the time synchronization moment, setting the power-off waiting time, calculating the RTC time error after the device is powered on again, and recording the environmental parameters is as follows: Send the time synchronization command to the device through the CAN bus interface, and set the current master time T m Write to the device RTC chip and record the time T s ; Set the power-off waiting time Δt according to the equipment type and working conditions p ; After the power-off wait is over, power on again and record the power-on time T r , and read the device RTC time T d , record the reading time of the device RTC time; Calculate the RTC time error ΔT between the power-on time and the device RTC time rtc , preset RTC error threshold, if the RTC time error is less than or equal to the RTC error threshold, the RTC is judged to be normal, otherwise it is marked as clock abnormal; Record environmental parameters before and after power failure; the environmental parameters include current I, power supply voltage U, temperature t and humidity h; The RTC time error is stored in the NVM memory as the initial error reference.

3. The method for industrial equipment failure prediction and health management according to claim 2, wherein: The specific method of obtaining the time drift between the device RTC time and the most recent time synchronization in real time, and updating the abnormality criterion based on the time drift and the RTC time error to enter the time correction mode is as follows: Get the device RTC time T in real time d , and calculate its difference with the most recent successful time T s The time drift between Read the RTC time error stored in NVM memory Combined with the time drift, the time anomaly criterion is updated and the time anomaly is judged using the time anomaly criterion; When m time anomalies are detected, the system enters the time correction mode. However, if the time anomaly recovers before entering the time correction mode and the duration exceeds T n If the error is less than 1 second, the abnormal mark is removed and the system is considered to have been automatically repaired.

4. The method for industrial equipment failure prediction and health management according to claim 3, wherein: The time anomaly criterion includes three conditions. If any one of the conditions is met, it is identified as abnormal data. Condition 1 is that if the time drift ΔT j Smaller than the RTC time error read from NVM memory The time is judged to be abnormal; the second condition is if the time drift ΔT j The absolute difference between the RTC time error read from the NVM memory is greater than the adaptive time jump threshold ε j , then the time is determined to be abnormal; condition three is to calculate the difference between the current time drift and the time drift change rate, if the absolute value of the difference between the difference and the RTC time error is greater than the adaptive time difference change threshold ε for M consecutive times g , then the time is determined to be abnormal.

5. The method for industrial equipment failure prediction and health management according to claim 4, wherein: The adaptive time jump threshold ε j and adaptive time difference change threshold ε g The calculation method is as follows: obtain historical environmental parameters and corresponding timestamps, and record the time drift; divide each environmental parameter into H groups according to its value range; for each environmental parameter group, calculate the statistical characteristics of its time jump amplitude and time difference change rate; for each environmental parameter group, establish a mapping function between the adaptive time jump threshold and the adaptive time difference change threshold based on its corresponding statistical characteristics; The current environmental parameters are collected in real time, and the corresponding adaptive time jump threshold and adaptive time difference change threshold are calculated according to the statistical characteristics of the group to which they belong.

6. The method for industrial equipment failure prediction and health management according to claim 5, wherein: The specific method of identifying key events based on the machine learning model and extracting their time anchor points is as follows: Environmental parameters are used as input features and output as event types to train a machine learning model; the environmental parameters include current I, power supply voltage U, temperature t, and humidity h; Collect environmental parameters during equipment operation, according to a fixed time window Δt w Divide and obtain the environmental parameter sequence {S1, S2, ..., S N }; For each environmental parameter segment S n , extract its current change rate ΔI n , power supply voltage change rate ΔU n and the temperature change rate Δt n , to determine whether it is greater than the preset current mutation threshold ε I , voltage mutation threshold ε U , temperature mutation threshold ε t If the current change rate is greater than the current mutation threshold or the power supply voltage change rate is greater than the voltage mutation threshold or the temperature change rate is greater than the temperature mutation threshold, then the environmental parameter segment S n Input into the trained machine learning model to identify event types; For the environmental parameter segment S identified as a key event type n , extract the reading time of the device RTC time of the environmental parameter fragment as the time anchor point T n , and record its event type E n .

7. The method for industrial equipment failure prediction and health management according to claim 6, wherein: The specific method of constructing an LSTM model based on environmental parameters and RTC time, predicting an estimated time interval, and respectively correcting the estimated time interval including and excluding the time anchor point is as follows: For the environmental parameter fragment sequence {S a ,S a+1 ,…,S b }, extract its original device RTC time series Construct feature matrix X based on environmental parameters and device RTC time series; Input the feature matrix X into the pre-trained LSTM model to obtain the estimated time interval to which each environmental parameter segment belongs; If the end time of the i-th estimated time interval is and the start time of the i+1th estimated time interval If the absolute value of the difference is greater than the time tolerance ε, the start time of the i+1th estimated time interval is set to Adjust to the end time of the i-th estimated time interval Add the value of the time offset δ; Based on the time anchor point and the original device RTC time, the estimated time interval is calibrated to obtain the corrected time corresponding to the estimated time interval; Compare the corrected time with the device RTC time and calculate the difference between the two as the new RTC time error And stored in NVM memory.

8. The method for industrial equipment failure prediction and health management according to claim 7, wherein: The specific method of calibrating the estimated time interval based on the time anchor point and the original device RTC time to obtain the corrected time of the corresponding time interval and update the RTC time error is as follows: If the time anchor point T n If it falls within the estimated time interval, the start time of the corrected time of the estimated time interval is the difference between the time anchor point and half of the fixed time window, and the end time is the sum of the time anchor point and half of the fixed time window; For an estimated time interval without a time anchor point, for each environmental parameter segment, calculate the first difference between the original device RTC time and the start time of the estimated time interval, adjust the start time of the revised time of the estimated time interval to the weighted sum of the start time and the first difference, and adjust the end time of the revised time of the estimated time interval to the weighted sum of the end time and the first difference; The correction time is smoothed and mapped back to the original device RTC time series to obtain the final correction time series.

9. An industrial equipment fault prediction and health management system, which is implemented based on the industrial equipment fault prediction and health management method according to any one of claims 1 to 8, characterized in that: include: The periodic time test module is used to send time synchronization instructions to the device and record the synchronization time, set the power-off waiting time, calculate the RTC time error after the device is powered on again, and record environmental parameters; The time anomaly judgment module is used to obtain the time drift between the device RTC time and the most recent time synchronization in real time, update the anomaly judgment criteria based on the time drift and RTC time error, and enter the time correction mode; The time correction module is used to identify key events based on the machine learning model, extract their time anchor points, build an LSTM model based on environmental parameters and RTC time, predict the estimated time interval, and correct the estimated time interval containing and excluding time anchor points respectively, and update the RTC time error.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps in the industrial equipment fault prediction and health management method according to any one of claims 1 to 8.

Images