Intelligent electronic equipment fault diagnosis method and system

Abstract

The invention relates to the technical field of fault diagnosis of intelligent electronic equipment, and particularly discloses a fault diagnosis method and system for intelligent electronic equipment, and the method comprises the steps: collecting temperature data and operation response time in real time through a high-precision temperature sensor disposed on the surface of the equipment and an external high-precision timer; a temperature anomaly feature value and an operation response time anomaly feature value are respectively calculated by adopting fast Fourier transform and an autoregression integral moving average model, and are constructed into comprehensive feature vectors, and based on the feature vectors, whether the intelligent electronic equipment has a fault state is judged by using a random forest model. And the probability that the equipment breaks down in a period of time in the future is predicted through the multiple linear regression model, and for non-fault intelligent electronic equipment, the system formulates a personalized dynamic maintenance plan according to a prediction result, so that the accuracy and high efficiency of maintenance work are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis of intelligent electronic devices, and particularly to a method and system for fault diagnosis of intelligent electronic devices. Background Art

[0002] Intelligent electronic devices play a crucial role in modern industry and daily life, and their reliability and stability directly affect the overall performance and safety of the system. To ensure the continuous and efficient operation of these devices, traditional maintenance methods mainly rely on regular inspections and threshold-based monitoring systems. These methods typically include manually checking the device status, recording key parameters (such as temperature, operation response time), and judging whether the device is working properly through simple statistical analysis or preset fixed thresholds. However, with the increase in device complexity and the improvement of the demand for high reliability, traditional methods gradually show limitations in dealing with complex fault modes and early fault prediction.

[0003] The existing technologies have the following deficiencies:

[0004] Although the existing fault diagnosis technologies have improved the maintainability and reliability of devices to a certain extent, there are still several significant deficiencies. First, the method based on fixed thresholds is difficult to adapt to the individual differences between different devices and environmental changes, easily leading to false alarms or missed alarms, which affects the accuracy of maintenance decisions. Second, traditional monitoring systems lack the ability to comprehensively analyze multi-dimensional data (such as temperature, operation response time) and cannot fully capture the changing trends of the device health status. In addition, most of the existing technologies focus on the response measures after a fault occurs and fail to achieve true preventive maintenance, that is, taking effective measures to intervene before a fault occurs. This not only increases the risk of downtime and maintenance costs caused by sudden faults, but also may have a serious impact on production continuity and safety. Therefore, there is an urgent need for a more intelligent, data-driven fault diagnosis and prediction solution to overcome the deficiencies of the existing technologies and improve the overall reliability and operation efficiency of devices. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for fault diagnosis of intelligent electronic devices to solve the problems in the above background.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A method for fault diagnosis of intelligent electronic devices, comprising the following steps:

[0008] S1: Real-time collect the temperature data and operation response time of the intelligent electronic device during operation;

[0009] S2: Analyze the temperature data and operation response time of the intelligent electronic device. According to the analysis results, construct a machine learning model to determine whether the intelligent electronic device is in a faulty state;

[0010] S3: According to the judgment result, divide the intelligent electronic devices into faulty intelligent electronic devices and non-faulty intelligent electronic devices, and give early warning to the faulty intelligent electronic devices;

[0011] S4: Based on the non-faulty intelligent electronic devices, predict the probability of the intelligent electronic device having a fault in a future period of time, and formulate and implement a dynamic maintenance plan according to the prediction result.

[0012] As a further solution of the present invention: The analysis of the temperature data and operation response time of the intelligent electronic device specifically includes:

[0013] During the monitoring period, obtain the temperature data and operation response time of the intelligent electronic device in real time according to the time series. According to the fluctuation amplitude of the temperature data of the intelligent electronic device, calculate the temperature anomaly characteristic value to evaluate the temperature stability of the intelligent electronic device. According to the delay degree of the operation response time of the intelligent electronic device, calculate the operation response time anomaly characteristic value to evaluate the stability of the operation response time of the intelligent electronic device.

[0014] As a further solution of the present invention: The process of obtaining the temperature anomaly characteristic value is as follows:

[0015] During the monitoring period, obtain the temperature data of the intelligent electronic device in real time according to the time series. Preprocess the collected temperature data, apply the fast Fourier transform to the preprocessed temperature data to convert it into a frequency domain representation. According to the frequency domain representation of the temperature data, calculate its power spectral density, and determine the key frequency interval. In the selected key frequency interval, calculate the sum of the power spectral densities of all frequency components to obtain the temperature anomaly characteristic value.

[0016] As a further solution of the present invention: The process of obtaining the operation response time anomaly characteristic value is as follows:

[0017] During the monitoring period, the operation response time of the intelligent electronic device is obtained in real time according to the time series; determining the ARIMA model parameters includes: selecting the number of autoregressive terms, the order of differencing, and the number of moving average terms. According to the determined ARIMA model parameters, an ARIMA model is established, and the determined ARIMA model parameters are fitted to the operation response time series. The fitted ARIMA model is used to predict the future operation response time, and the difference between the actual operation response time and the predicted operation response time is calculated to obtain the residual value. The standard deviation of all residuals is calculated, and it is calculated whether the absolute value of the residual at all time points exceeds a set threshold. Wherein, the set threshold is 3 times the standard deviation of the residual value. The ratio of the number of all residuals exceeding the set threshold to the total number of residuals is calculated to obtain the operation response time anomaly eigenvalue.

[0018] As a further solution of the present invention: The constructing of the machine learning model specifically includes:

[0019] Obtain the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue of the intelligent electronic device during the monitoring period, construct the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue into a comprehensive feature vector as the input of the machine learning model, the output of the model is the fault score, and the error of minimizing the predicted fault score of the intelligent electronic device and the actual fault score of the intelligent electronic device is used as the model training target to train the model. Based on the trained machine learning model, the fault score of the intelligent electronic device is output, and the machine learning model is a random forest model.

[0020] As a further solution of the present invention: The judging whether the intelligent electronic device is in a fault state specifically includes:

[0021] Judge whether the fault score of the intelligent electronic device is greater than or equal to a preset threshold. If so, it is recorded as a faulty intelligent electronic device. If not, it is recorded as a non-faulty intelligent electronic device.

[0022] As a further solution of the present invention: The predicting the probability that the intelligent electronic device will have a fault in a future period of time based on the non-faulty intelligent electronic device specifically includes:

[0023] Construct the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue of the non-faulty intelligent electronic device into a comprehensive feature vector as the input of the prediction model. The error of minimizing the probability that the intelligent electronic device has a fault and the actual probability that the intelligent electronic device has a fault is used as the training target of the prediction model to train the prediction model. According to the trained prediction model, the probability that the intelligent electronic device will have a fault in the future is output, and the prediction model is a multiple linear regression model.

[0024] As a further solution of the present invention: according to the probability of failure of the intelligent electronic device, hierarchical maintenance is carried out on the intelligent electronic device, specifically including:

[0025] According to the probability of failure of the intelligent electronic device, the probability of failure is compared with a preset first threshold. If the probability of failure is greater than or equal to the preset first threshold, it means that the probability of failure of the corresponding intelligent electronic device is high, and the intelligent electronic device needs to be shut down for overhaul. If the probability of failure is greater than the preset second threshold and less than the preset first threshold, it means that the probability of failure of the corresponding intelligent electronic device is medium, and the monitoring frequency needs to be increased. If the probability of failure continues to increase, the intelligent electronic device needs to be shut down for overhaul. If the probability of failure is less than or equal to the preset second threshold, it means that the probability of failure of the corresponding intelligent electronic device is low, and no additional intervention is required, and normal monitoring is maintained.

[0026] The intelligent electronic device fault diagnosis system includes:

[0027] A data acquisition module, which is used to collect the temperature data and operation response time of the intelligent electronic device during operation in real time;

[0028] A fault identification module, which analyzes the temperature data and operation response time of the intelligent electronic device, and constructs a machine learning model based on the analysis results to judge whether the intelligent electronic device is in a fault state;

[0029] A fault classification and warning module, which classifies the intelligent electronic device into a faulty intelligent electronic device and a non-faulty intelligent electronic device according to the judgment result, and performs warning processing on the faulty intelligent electronic device;

[0030] A fault prediction and maintenance module, which predicts the probability of failure of the intelligent electronic device in a future period of time based on the non-faulty intelligent electronic device, and formulates and implements a dynamic maintenance plan according to the prediction result.

[0031] The beneficial effects of the present invention:

[0032] (1) The present invention proposes an innovative fault diagnosis method for intelligent electronic devices. By collecting key performance indicators during device operation in real time, including temperature data and operation response time, and applying advanced data analysis techniques for in-depth processing, it achieves accurate assessment of the device's health status and effective prediction of future fault risks. Specifically, the present invention first uses high-precision temperature sensors deployed on the device surface and external high-precision timers to obtain temperature data and operation response time in real time. After preprocessing these raw data, the fast Fourier transform and autoregressive integrated moving average model are respectively used for analysis to calculate the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue. The calculation process of the temperature anomaly eigenvalue includes converting the temperature data into a frequency-domain representation, calculating its power spectral density, and then obtaining the total energy of all frequency components in the selected key frequency interval to reflect the temperature stability of the device. The operation response time anomaly eigenvalue is obtained by establishing an ARIMA model to predict the future operation response time, and calculating the residual and its standard deviation between the actual and predicted values to quantify the delay degree of the operation response time. These two anomaly eigenvalues together constitute a comprehensive feature vector, which is used as the input of the machine learning model. To determine whether the device is in a fault state, the present invention uses a random forest model, which is an ensemble learning method composed of multiple decision trees. Each tree independently trains on the input features and then aggregates the results to output the final fault score. Based on the set threshold, the system can accurately distinguish between faulty and non-faulty devices and issue a warning signal for faulty devices. In addition, for non-faulty devices, the present invention further uses a multiple linear regression model, trained based on a historical data set, to minimize the mean square error between the predicted fault probability and the actual fault probability, thereby predicting the probability of the device failing in the future for a period of time. This two-layer model architecture not only improves the accuracy of fault detection but also enhances the predictability of future fault possibilities. Compared with traditional methods based on fixed thresholds or simple statistical analysis, the present invention has significant advantages. Traditional methods often rely on the setting of thresholds for a single parameter and are easily affected by environmental changes or device individual differences, resulting in false alarms or missed alarms. The present invention, by comprehensively considering data features in multiple dimensions and using advanced machine learning algorithms for in-depth mining, can more comprehensively and accurately identify potential fault signals. At the same time, by learning from historical data and continuously optimizing model parameters, the system can dynamically adjust its prediction ability, discover and prevent potential fault risks in advance, and achieve true preventive maintenance.

[0033] (2)The present invention designs a comprehensive and highly intelligent fault diagnosis and prediction system, which covers four core modules: data acquisition, fault identification, fault classification and early warning, and fault prediction and maintenance. This system not only realizes the real-time monitoring and accurate classification of the current health status of intelligent electronic devices, but also can, based on advanced data analysis techniques and machine learning models, customize personalized dynamic maintenance plans for each non-fault device, so as to maximize the benefits of preventive maintenance strategies. The data acquisition module is the cornerstone of the entire system. Through high-precision temperature sensors and external high-precision timers deployed on the device surface, key performance indicators during device operation are obtained in real time, including temperature data and operation response time. After preprocessing these data, fast Fourier transform and autoregressive integrated moving average models are respectively used for in-depth analysis to calculate the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue. These eigenvectors not only reflect the current working state of the device, but also provide a solid data foundation for subsequent fault identification. The fault identification module uses a random forest model to analyze the comprehensive eigenvector, outputs the fault score of the device, and determines whether the device is in a fault state according to the set threshold. This integrated learning method significantly improves the accuracy and robustness of fault detection through the collaborative work of multiple decision trees. For the devices identified as faulty, the system will immediately trigger the fault classification and early warning module, generate a detailed fault report and issue an early warning signal to remind the maintenance personnel to take measures in time to prevent the fault from deteriorating further. At the same time, for non-fault devices, the fault prediction and maintenance module plays a crucial role. This module uses a multiple linear regression model, trained based on historical data sets, to minimize the mean square error between the predicted fault probability and the actual fault probability, so as to predict the possibility of the device failing in the future for a period of time. Based on these prediction results, the system can formulate personalized dynamic maintenance plans for each device to ensure the accuracy and efficiency of maintenance work. For example, for some key components or vulnerable parts, the system can arrange replacement or repair in advance according to their specific usage conditions and wear degrees, avoiding the situations of over-maintenance or neglecting potential problems. To ensure the long-term effectiveness and adaptability of the system, regularly retraining the model and adjusting the maintenance plan is an essential part. By learning new data and optimizing parameters, the system can continuously improve its prediction ability and maintenance efficiency to ensure that it is always in the best state. In addition, this data-driven intelligent maintenance solution not only greatly reduces the human and material costs of device maintenance, extends the service life of the device, but also significantly improves the overall operation efficiency. Since preventive measures can be taken before a fault occurs, the system effectively reduces production interruptions and economic losses caused by sudden faults, ensuring the continuity and stability of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention will be further described below with reference to the accompanying drawings.

[0035] Figure 1 is the specific step flow block diagram of the intelligent electronic device fault diagnosis method of the present invention;

[0036] Figure 2 is the flow block diagram of the intelligent electronic device fault diagnosis system in the present invention. Specific Embodiments

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0038] Please refer to Figure 1 as shown, the present invention is an intelligent electronic device fault diagnosis method, including the following steps:

[0039] S1: Real-time collect the temperature data and operation response time of the intelligent electronic device during operation;

[0040] S2: Analyze the temperature data and operation response time of the intelligent electronic device, and build a machine learning model based on the analysis results to determine whether the intelligent electronic device is in a fault state;

[0041] S3: According to the judgment result, divide the intelligent electronic device into a faulty intelligent electronic device and a non-faulty intelligent electronic device, and perform a warning process on the faulty intelligent electronic device;

[0042] S4: Based on the non-faulty intelligent electronic device, predict the probability of the intelligent electronic device having a fault in a future period of time, and formulate and implement a dynamic maintenance plan according to the prediction result.

[0043] In S1, the temperature data and operation response time of the intelligent electronic device during operation are collected in real time, specifically including:

[0044] During the monitoring period, the temperature data of the intelligent electronic device is obtained in real time according to the time series, and the temperature data is collected by a temperature sensor deployed on the surface of the intelligent electronic device. The operation response time of the intelligent electronic device is obtained in real time according to the time series, and the operation response time is collected by an external timer, specifically including:

[0045] During the monitoring period, temperature data is obtained in real time by high-precision temperature sensors deployed on the surface of the intelligent electronic device. These sensors are precisely placed near the key heat sources of the device (such as the processor and power module) to ensure that the internal operating temperature changes of the device can be accurately captured. Each sensor is connected to the central data acquisition unit through a standard communication interface (such as SPI), and the central data acquisition unit is responsible for synchronizing and recording the timestamp temperature data from each sensor. To ensure the stability and accuracy of long-term operation, the system also integrates a self-calibration function, and the sensors will perform self-calibration regularly to eliminate measurement deviations caused by environmental changes or long-term use. In addition, all the collected data will be transmitted to a central database for further processing and analysis.

[0046] The operation response time is measured in real time using an external high-precision timer. When the intelligent electronic device receives an instruction, the external timer immediately starts timing and stops timing when the device completes the corresponding operation and returns the result, thereby recording the time required for the entire operation process. This timer usually has a nanosecond-level resolution, which can provide extremely high time accuracy and avoid data inaccuracies caused by asynchronous or inaccurate internal clocks of the device. To improve the reliability of the data and reduce noise interference, the same set of operations will be measured simultaneously by multiple timers, and the average value will be taken as the actual response time of this operation. All the timing data will also be marked with accurate timestamps and stored in the central database together with the temperature data, providing high-quality basic data for subsequent fault diagnosis and prediction models.

[0047] In S2, the temperature data and operation response time of the intelligent electronic device are analyzed. According to the analysis results, a machine learning model is constructed to determine whether the intelligent electronic device is in a faulty state, specifically including:

[0048] During the monitoring period, the temperature data and operation response time of the intelligent electronic device are obtained in real time according to the time series. According to the fluctuation range of the temperature data of the intelligent electronic device, the temperature anomaly characteristic value is calculated to evaluate the temperature stability of the intelligent electronic device. According to the delay degree of the operation response time of the intelligent electronic device, the operation response time anomaly characteristic value is calculated to evaluate the stability of the operation response time of the intelligent electronic device;

[0049] The process of obtaining the temperature anomaly characteristic value is as follows:

[0050] During the monitoring period, the temperature data of the intelligent electronic device is obtained in real time according to the time series. The temperature data is collected by temperature sensors deployed on the surface of the intelligent electronic device. The collected temperature data is preprocessed, including removing noise and trend terms, to ensure the accuracy of subsequent analysis. The preprocessed temperature data is applied with the fast Fourier transform to be converted into a frequency-domain representation. According to the frequency-domain representation of the temperature data, its power spectral density is calculated. The calculation expression is: ; In the formula, represents the angular frequency, is the frequency-domain representation of the temperature data, represents the power spectral density; the power spectral density reflects the energy intensity of different frequency components;

[0051] According to the working characteristics of the intelligent electronic device and its possible failure modes, a key frequency interval is determined. Within the selected key frequency interval, the sum of the power spectral densities of all frequency components is calculated to obtain the temperature anomaly characteristic value.

[0052] It should be noted that: the temperature anomaly characteristic value reflects the temperature stability degree of the intelligent electronic device, and when the value of the temperature anomaly characteristic value is larger, the corresponding temperature stability of the intelligent electronic device is worse. The temperature anomaly characteristic value is a key indicator for subsequent evaluation of the failure state of the intelligent electronic device.

[0053] The process of obtaining the operation response time anomaly characteristic value is as follows:

[0054] During the monitoring period, the operation response time of the intelligent electronic device is obtained in real time according to the time series. The operation response time is collected by an external timer; determining the ARIMA model parameters includes: selecting the number of autoregressive terms, the order of differencing, and the number of moving average terms. According to the determined ARIMA model parameters, an ARIMA model is established, and the determined ARIMA model parameters are fitted to the operation response time series. The fitted ARIMA model is used to predict the future operation response time, and the difference between the actual operation response time and the predicted operation response time is calculated to obtain the residual value. The standard deviation of all residuals is calculated, and whether the absolute value of the residual at all time points exceeds the set threshold is calculated. Among them, the set threshold is 3 times the standard deviation of the residuals. The ratio of the number of all residuals exceeding the set threshold to the total number of residuals is calculated to obtain the operation response time anomaly characteristic value.

[0055] It should be noted that: the operation response time anomaly characteristic value reflects the degree of operation response time delay of the intelligent electronic device, and when the value of the operation response time anomaly characteristic value is larger, the corresponding operation response time delay of the intelligent electronic device is longer. The operation response time anomaly characteristic value is a key indicator for subsequent evaluation of the failure state of the intelligent electronic device.

[0056] The construction of the machine learning model specifically includes:

[0057] Obtain the temperature anomaly eigenvalue and operation response time anomaly eigenvalue of the intelligent electronic device during the monitoring period, construct the temperature anomaly eigenvalue and operation response time anomaly eigenvalue into a comprehensive feature vector as the input of the machine learning model, and the output of the model is the fault score. Minimize the error between the predicted fault score of the intelligent electronic device and the actual fault score of the intelligent electronic device as the model training objective, train the model, and based on the trained machine learning model, output the fault score of the intelligent electronic device. The machine learning model is a random forest model.

[0058] It should be noted that: During the model training stage, the random forest model is trained using a historical dataset, which includes temperature anomaly eigenvalues, operation response time anomaly eigenvalues, and their corresponding fault scores under known normal and fault states. The training objective of the model is to minimize the error between the predicted fault score and the actual fault score, that is, to make the prediction result close to the real situation by adjusting the model parameters. After sufficient training, the random forest model can accurately output the fault score of the intelligent electronic device according to the new comprehensive feature vector, thereby realizing an effective assessment of the device health status. This method not only improves the accuracy of fault detection but also enhances the predictability of future fault possibilities, providing strong support for equipment maintenance.

[0059] The judgment of whether the intelligent electronic device is in a fault state specifically includes:

[0060] Judge whether the fault score of the intelligent electronic device is greater than or equal to a preset threshold. If so, record it as a faulty intelligent electronic device; if not, record it as a non-faulty intelligent electronic device.

[0061] In S3, perform early warning processing on the faulty intelligent electronic device, specifically including:

[0062] Based on the faulty intelligent electronic device, the system immediately activates the early warning mechanism to generate a detailed fault report, which includes the fault type, severity, possible causes, and recommended emergency response measures. At the same time, the system sends real-time alerts to relevant maintenance personnel and management personnel through multiple channels (such as email, text message, in-app notification) to ensure that they can respond quickly and take necessary corrective measures. In addition, the system will automatically generate a maintenance work order and assign it to the appropriate repair team to arrange maintenance work in a timely manner, minimizing equipment downtime and potential production losses. This series of early warning processing operations aims to quickly locate problems and intervene in a timely manner to effectively prevent the further deterioration of faults.

[0063] In S4, based on non-faulty intelligent electronic devices, predict the probability of a fault occurring in intelligent electronic devices within a certain period in the future, and formulate and implement a dynamic maintenance plan according to the prediction results, specifically including:

[0064] Construct a comprehensive feature vector from the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue of non-faulty intelligent electronic devices as the input of the prediction model. Take minimizing the error between the probability of a fault occurring in intelligent electronic devices and the actual probability of a fault occurring in intelligent electronic devices as the training objective of the prediction model, train the prediction model, and according to the trained prediction model, output the probability of a fault occurring in future intelligent electronic devices. The prediction model is a multiple linear regression model;

[0065] It should be noted that: in the model training stage, use the historical data set to train the multiple linear regression model, which contains the comprehensive feature vectors and their corresponding fault probabilities in known normal and potential fault states. The training objective is to minimize the mean square error between the predicted fault probability and the actual fault probability, that is, to make the model output close to the real situation by adjusting the regression coefficients. Specifically, use the gradient descent optimization algorithm to solve the optimal regression coefficients to ensure that the model has high prediction accuracy. After sufficient training, the multiple linear regression model can accurately output the probability of a fault occurring in future intelligent electronic devices according to the new comprehensive feature vector. When the new temperature anomaly eigenvalue and operation response time anomaly eigenvalue are input into the model, the model will generate a fault probability prediction value based on the learned parameters. This prediction result not only improves the accuracy of the evaluation of the device health status, but also enhances the predictability of future fault possibilities, providing strong support for device maintenance. In addition, retraining the model regularly can further improve its performance and ensure long-term effective fault prediction ability.

[0066] Construct a comprehensive feature vector from the temperature anomaly eigenvalue and the operation response time anomaly eigenvalue of non-faulty intelligent electronic devices as the input of the prediction model. Take minimizing the error between the probability of a fault occurring in intelligent electronic devices and the actual probability of a fault occurring in intelligent electronic devices as the training objective of the prediction model, train the prediction model, and according to the trained prediction model, output the probability of a fault occurring in future intelligent electronic devices. The prediction model is a multiple linear regression model.

[0067] Please refer to Figure 2 as shown, the intelligent electronic device fault diagnosis system includes:

[0068] A data acquisition module, which is used to collect the temperature data and operation response time of the intelligent electronic device in real time during operation;

[0069] A fault identification module that analyzes the temperature data and operation response time of the intelligent electronic device, and constructs a machine learning model based on the analysis results to determine whether the intelligent electronic device is in a fault state;

[0070] A fault classification and warning module that classifies the intelligent electronic device into a faulty intelligent electronic device and a non-faulty intelligent electronic device according to the judgment result, and performs warning processing on the faulty intelligent electronic device;

[0071] A fault prediction and maintenance module that predicts the probability of a fault in the intelligent electronic device in a future period of time based on the non-faulty intelligent electronic device, and formulates and implements a dynamic maintenance plan according to the prediction result.

[0072] The working principle of the present invention: The present invention provides a method and system for fault diagnosis of intelligent electronic devices, aiming to evaluate the health status and predict potential faults by real-time monitoring and analysis of the temperature data and operation response time of the devices. First, during the monitoring period, the temperature data and operation response time of the device are respectively collected in real time through a high-precision temperature sensor deployed on the surface of the intelligent electronic device and an external high-precision timer. The temperature data is preprocessed and then converted into a frequency-domain representation using the fast Fourier transform, and the power spectral density is calculated to determine the temperature anomaly eigenvalue; the operation response time is predicted by establishing an ARIMA model, and the residual and its standard deviation between the actual and predicted values are calculated to obtain the operation response time anomaly eigenvalue. Subsequently, these two anomaly eigenvalues are constructed into a comprehensive feature vector, which is used as the input of the random forest model to output the fault score of the intelligent electronic device, and it is judged whether the device is in a fault state according to the set threshold. For non-faulty devices, a multiple linear regression model is further trained based on the historical data set to minimize the error between the predicted fault probability and the actual fault probability, so as to predict the probability of a fault in the device in a future period of time and formulate a dynamic maintenance plan. The entire system includes a data acquisition module, a fault identification module, a fault classification and warning module, and a fault prediction and maintenance module, ensuring the full-process automated management from data acquisition to fault prediction. This method not only improves the accuracy of fault detection, but also enhances the predictability of future fault possibilities, provides strong support for the efficient management and preventive maintenance of intelligent electronic devices, and significantly reduces the device failure rate and maintenance cost. In addition, regularly retraining the model can continuously optimize the performance and ensure long-term effective fault prediction ability. Through this comprehensive data-driven method, the present invention realizes the accurate assessment and timely warning of the health status of intelligent electronic devices, greatly improving the safety and reliability of device operation.

[0073] The above has described in detail an embodiment of the present invention, but the above content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention shall still fall within the scope covered by the patent of the present invention.

Claims

1. A method for fault diagnosis of intelligent electronic devices, characterized in that, It includes the following steps: S1: Real-time collect the temperature data and operation response time of the intelligent electronic device during operation; S2: Analyze the temperature data and operation response time of the intelligent electronic device, and build a machine learning model according to the analysis results to judge whether the intelligent electronic device is in a fault state; S3: According to the judgment result, divide the intelligent electronic device into a faulty intelligent electronic device and a non-faulty intelligent electronic device, and perform early warning processing on the faulty intelligent electronic device; S4: Based on the non-faulty intelligent electronic device, predict the probability of the intelligent electronic device having a fault in a future period of time, and formulate and implement a dynamic maintenance plan according to the prediction result.

2. The fault diagnosis method of the intelligent electronic device according to claim 1, characterized in that The analysis of the temperature data and operation response time of the intelligent electronic device specifically includes: During the monitoring period, obtain the temperature data and operation response time of the intelligent electronic device in real time according to the time series. According to the fluctuation range of the temperature data of the intelligent electronic device, calculate the temperature anomaly characteristic value to evaluate the temperature stability of the intelligent electronic device. According to the delay degree of the operation response time of the intelligent electronic device, calculate the operation response time anomaly characteristic value to evaluate the stability of the operation response time of the intelligent electronic device.

3. The intelligent electronic device fault diagnosis method according to claim 2, wherein The process of obtaining the temperature anomaly characteristic value is: During the monitoring period, obtain the temperature data of the intelligent electronic device in real time according to the time series. Preprocess the collected temperature data, apply the fast Fourier transform to the preprocessed temperature data to convert it into a frequency domain representation. According to the frequency domain representation of the temperature data, calculate its power spectral density, and determine the key frequency interval. In the selected key frequency interval, calculate the sum of the power spectral densities of all frequency components to obtain the temperature anomaly characteristic value.

4. The intelligent electronic device fault diagnosis method according to claim 2, characterized in that, The process of obtaining the operation response time anomaly characteristic value is: During the monitoring period, obtain the operation response time of the intelligent electronic device in real time according to the time series; Determine the ARIMA model parameters including: select the number of autoregressive terms, the order of differencing, and the number of moving average terms. According to the determined ARIMA model parameters, establish an ARIMA model, and fit the determined ARIMA model parameters to the operation response time series. Use the fitted ARIMA model to predict the future operation response time, and calculate the difference between the actual operation response time and the predicted operation response time to obtain the residual value. Calculate the standard deviation of all residuals, calculate whether the absolute value of the residual at all time points exceeds the set threshold. Among them, the set threshold is 3 times the standard deviation of the residuals. Calculate the ratio of the number of residuals exceeding the set threshold to the total number of residuals to obtain the operation response time anomaly characteristic value.

5. The fault diagnosis method for the intelligent electronic device according to claim 1, wherein, The construction of the machine learning model specifically includes: Obtain the temperature anomaly eigenvalue and operation response time anomaly eigenvalue of the intelligent electronic device within the monitoring period, construct the temperature anomaly eigenvalue and operation response time anomaly eigenvalue into a comprehensive feature vector as the input of the machine learning model, the output of the model is the fault score, and minimize the error between the predicted fault score of the intelligent electronic device and the actual fault score of the intelligent electronic device as the model training objective, train the model, and based on the trained machine learning model, output the fault score of the intelligent electronic device. The machine learning model is a random forest model.

6. The intelligent electronic device fault diagnosis method according to claim 1, wherein, The judgment of whether the intelligent electronic device is in a fault state specifically includes: Judge whether the fault score of the intelligent electronic device is greater than or equal to the preset threshold. If so, record it as a faulty intelligent electronic device; if not, record it as a non-faulty intelligent electronic device.

7. The intelligent electronic device fault diagnosis method according to claim 1, characterized in that Based on the non-faulty intelligent electronic device, predicting the probability of the intelligent electronic device having a fault in a future period of time specifically includes: Construct the temperature anomaly eigenvalue and operation response time anomaly eigenvalue of the non-faulty intelligent electronic device into a comprehensive feature vector as the input of the prediction model, and minimize the error between the probability of the intelligent electronic device having a fault and the actual probability of the intelligent electronic device having a fault as the training objective of the prediction model, train the prediction model, and according to the trained prediction model, output the probability of the intelligent electronic device having a fault in the future. The prediction model is a multiple linear regression model.

8. The intelligent electronic device fault diagnosis method according to claim 7, characterized in that, According to the probability of the intelligent electronic device having a fault, perform hierarchical maintenance on the intelligent electronic device, specifically including: Compare the probability of the fault with the preset first threshold according to the probability of the intelligent electronic device having a fault. If the probability of the fault is greater than or equal to the preset first threshold, the probability of the corresponding intelligent electronic device having a fault is high, and the intelligent electronic device needs to be shut down for maintenance. If the probability of the fault is greater than the preset second threshold and less than the preset first threshold, the probability of the corresponding intelligent electronic device having a fault is medium, and the monitoring frequency needs to be increased. If the probability of the fault continues to increase, the intelligent electronic device needs to be shut down for maintenance. If the probability of the fault is less than or equal to the preset second threshold, the probability of the corresponding intelligent electronic device having a fault is low, and no additional intervention is required, and normal monitoring is maintained.

9. An intelligent electronic device fault diagnosis system for the intelligent electronic device fault diagnosis method according to any one of claims 1-8, characterized in that, Including: A data acquisition module for real-time collecting the temperature data and operation response time of the intelligent electronic device during operation; A fault identification module for analyzing the temperature data and operation response time of the intelligent electronic device, and constructing a machine learning model according to the analysis result for judging whether the intelligent electronic device is in a fault state; A fault classification and early warning module for classifying the intelligent electronic device into a faulty intelligent electronic device and a non-faulty intelligent electronic device according to the judgment result, and performing early warning processing on the faulty intelligent electronic device; A fault prediction and maintenance module, which is based on non-fault intelligent electronic devices, predicts the probability of a fault in intelligent electronic devices within a future period of time, and formulates and implements a dynamic maintenance plan according to the prediction results.

Images