Electronic component performance data monitoring method and system

Abstract

The present invention discloses a method and system for monitoring electronic component performance data, comprising: inputting an LSTM neural network model for training based on the electronic component's operating history data to construct an electronic component performance prediction model; inputting current electronic component surface temperature field distribution data, partial discharge characteristics, and leakage current dynamic parameters into the electronic component performance prediction model to obtain multiple predicted change trend functions of the electronic component performance in a future time period; comparing the predicted trend functions with the actual electronic component performance data collected in real time, calculating the error value, and obtaining the predicted data with the minimum error; performing dynamic error analysis to generate model correction parameters; and performing online parameter updates on the electronic component performance prediction model to form a closed-loop optimization system. The present invention has the advantages of accurately predicting the performance changes of electronic components by comparing the prediction model with multi-dimensional real-time data, and ensuring the long-term stability and reliability of the electronic components through a closed-loop optimization mechanism.

Description

Technical Field

[0001] The present invention relates to data monitoring technology, and in particular to a method and system for monitoring electronic component performance data. Background Art

[0002] With the rapid development of electronic technology, electronic components are increasingly used in various electronic products, particularly in the fields of communications, automotive, aerospace, and medical. Electronic components play a vital role in these fields. For example, microprocessors in communications systems, sensors in automobiles, power management systems in spacecraft, and precision sensors in medical devices all require continuous and stable operation in high-intensity and complex environments. Therefore, ensuring the efficient and stable operation of these components is crucial to the safety and reliability of the entire system.

[0003] Most of the electronic component performance data monitoring methods currently available on the market rely on traditional static models or single-parameter monitoring methods, and lack comprehensive analysis of multi-dimensional data. These methods are usually unable to perform real-time performance prediction and dynamic adjustment, and the accuracy and adaptability of the prediction results are poor, and are easily affected by factors such as environmental changes, working conditions, and aging. In addition, many existing methods lack adaptive capabilities and are unable to optimize and correct based on real-time data, resulting in prediction errors that are difficult to effectively control. Traditional methods usually rely on fixed models and cannot fully consider multiple influencing factors such as temperature, partial discharge, and leakage current, resulting in limited monitoring accuracy and insufficient ability to respond to complex situations. More importantly, most of these methods lack dynamic error analysis and online update mechanisms, and are unable to adjust according to errors and deviations in actual operation, resulting in poor long-term model stability and an inability to continuously optimize and improve prediction accuracy. Summary of the Invention

[0004] To improve existing electronic component performance data monitoring methods, a method and system for monitoring electronic component performance data are provided. This method combines an LSTM neural network with multi-dimensional real-time data to accurately predict performance changes of electronic components, dynamically optimizes the prediction model to improve accuracy, and ensures the long-term stability and reliability of electronic components under different operating conditions through online correction and closed-loop optimization mechanisms.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] The electronic component performance data monitoring method includes:

[0007] Based on the working history data of electronic components, the LSTM neural network model is input for training to build an electronic component performance prediction model and obtain the function of the electronic component performance changing with working time;

[0008] Obtain current electronic component surface temperature field distribution data, partial discharge characteristics, and leakage current dynamic parameters, input them into the electronic component performance prediction model, and obtain multiple predicted change trend functions of the electronic component performance in the future time period based on the changes in various factors;

[0009] Based on the obtained multiple predicted change trend function data, it is compared with the actual performance data of the electronic components collected in real time, the error value is calculated, and the predicted data with the smallest error is obtained;

[0010] Based on the prediction data with the minimum error, dynamic error analysis is performed to generate model correction parameters;

[0011] The electronic component performance prediction model is updated online based on the model correction parameters to form a closed-loop optimization system.

[0012] Preferably, the electronic component's working history data is input into an LSTM neural network model for training, an electronic component performance prediction model is constructed, and a function of how the electronic component's performance changes with working time is obtained specifically includes:

[0013] The working history data of electronic components is transferred to the LSTM neural network model for training to build an electronic component performance prediction model;

[0014] Based on the dual-branch structure of the output layer in the electronic component performance prediction model, the main branch predicts the performance parameters at the current moment, and the auxiliary branch outputs the performance degradation rate. The functional relationship between the performance parameters and the working time is constructed through integral operation.

[0015] Preferably, the obtaining of current electronic component surface temperature field distribution data, partial discharge characteristic quantities, and leakage current dynamic parameters, inputting these into an electronic component performance prediction model, and obtaining multiple predicted change trend functions of electronic component performance in a future time period based on changes in various factors specifically includes:

[0016] Scan the surface of electronic components with a non-contact infrared thermal imager array to obtain a two-dimensional temperature distribution matrix and extract characteristic parameters;

[0017] The pulse signal is collected by a high-frequency current sensor to obtain the characteristic value of partial discharge;

[0018] Acquire current characteristic parameters through dual-channel synchronous acquisition;

[0019] The preprocessed collected data is input into the electronic component performance prediction model, and based on the possible change trends of the electronic component parameter data, a multi-scenario electronic component performance prediction change trend function is generated.

[0020] Preferably, inputting the pre-processed collected data into the electronic component performance prediction model and generating a multi-scenario electronic component performance prediction change trend function based on the possible change trend of the electronic component parameter data specifically includes:

[0021] Based on the characteristic parameters that affect the performance of electronic components, the variation range of each disturbance dimension is obtained;

[0022] Through orthogonal experimental design, various parameter change scenarios are combined to generate multiple scenarios of electronic component performance changes;

[0023] Based on the characteristic parameter data under various scenarios, they are input into the electronic component performance prediction model to generate a variety of electronic component performance prediction trend functions. The characteristic data obtained include: performance parameter time series, performance degradation acceleration, and key threshold arrival time prediction.

[0024] Preferably, the method of comparing the obtained multiple predicted change trend function data with the actual electronic component performance data collected in real time, calculating the error value, and obtaining the predicted data with the smallest error specifically includes:

[0025] Based on the error contribution of each characteristic parameter, dynamic error weight allocation is performed;

[0026] Compare the real-time collected data with the data of the corresponding time nodes in the predicted change trend function, and obtain the error size between each predicted change trend function data and the actual collected data through multi-dimensional error joint calculation;

[0027] A set of prediction change trend functions with the smallest error value in the error data is selected as the best electronic component performance prediction model;

[0028] Based on the actual performance data of electronic components collected at each moment, the prediction results of an optimal electronic component performance prediction model are screened and obtained.

[0029] Preferably, the dynamic error analysis based on the prediction data with the minimum error to generate the model correction parameters specifically includes:

[0030] Based on the forecast data with the minimum error, the error fluctuation is obtained through the error distribution diagram;

[0031] Through the relationship between error and working conditions, the cause of error under specific conditions can be obtained;

[0032] By comparing the error distribution graph with the model prediction output, the deviation in data, model structure or parameter setting can be obtained;

[0033] Based on the results of dynamic error analysis, the model parameters are adjusted to generate model correction parameters, including: adjustment of learning rate, addition of regularization terms, and adjustment of training data volume.

[0034] Preferably, the online parameter updating of the electronic component performance prediction model based on the model correction parameters to form a closed-loop optimization system specifically includes:

[0035] Based on the obtained model correction parameters, the model is trained online through incremental learning, and the model is continuously optimized using the real-time electronic component performance data and error correction information;

[0036] Through the automated parameter adjustment mechanism, when the error exceeds the set threshold, the model parameter update is automatically triggered;

[0037] The model prediction results after each update are compared with the real-time data again to form a continuous feedback process.

[0038] Furthermore, an electronic component performance data monitoring system is proposed, comprising:

[0039] Model training module: The model training module is mainly used to input the working history data of electronic components into the LSTM neural network for training, build a performance prediction model and obtain a function of performance changes over time;

[0040] Performance prediction model module: The performance prediction model module is mainly used to build a performance prediction model for electronic components based on the output structure of the LSTM network, and predict various change trend functions in the future time period;

[0041] Multi-scenario prediction and analysis module: The multi-scenario prediction and analysis module generates multiple prediction scenarios based on different working conditions and disturbance factors combined with orthogonal experimental design;

[0042] Error calculation module: The error calculation module is mainly used to compare the real-time collected performance data with the prediction model results, calculate the error and dynamically assign weights based on the error contribution to obtain the prediction model with the minimum error value;

[0043] Online update and closed-loop optimization module: The online update and closed-loop optimization module is mainly used to perform incremental learning based on model correction parameters, update the prediction model online, automatically adjust parameters and continuously optimize performance prediction results;

[0044] Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

[0045] Compared with the prior art, the advantages of the present invention are:

[0046] By training a performance prediction model based on historical data, it is possible to accurately capture how electronic component performance changes over time and promptly identify potential failure risks. Furthermore, by integrating multi-dimensional data such as surface temperature, partial discharge, and leakage current, this method provides a more comprehensive reference for performance prediction, ensuring both accuracy and diversity. Through dynamic error analysis and an online correction mechanism, the system can continuously optimize the prediction model during actual operation, ensuring its efficiency and stability under different operating conditions. This method not only possesses adaptive capabilities, capable of adjusting prediction results based on real-time data, but also implements a closed-loop optimization system to ensure the long-term reliability and performance stability of electronic components. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the method proposed in the present invention;

[0048] Figure 2 A schematic diagram of the electronic component performance prediction model proposed in the present invention;

[0049] Figure 3 This is a schematic diagram of the calculation of the electronic component performance prediction model proposed in the present invention;

[0050] Figure 4 Schematic diagram of obtaining various prediction change trend functions proposed by the present invention;

[0051] Figure 5 This is a schematic diagram of obtaining the minimum error prediction data proposed by the present invention;

[0052] Figure 6 A schematic diagram of generating the model correction parameters proposed by the present invention;

[0053] Figure 7 This is a schematic diagram of online parameter update proposed by the present invention;

[0054] Figure 8 This is a diagram of the architecture of the electronic equipment in this solution;

[0055] Figure 9 This is a schematic diagram of the computer-readable storage medium structure in this solution. DETAILED DESCRIPTION

[0056] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0057] Electronic component performance data monitoring system, including:

[0058] Model training module: The model training module is mainly used to input the working history data of electronic components into the LSTM neural network for training, build a performance prediction model and obtain a function of performance changes over time;

[0059] Performance prediction model module: The performance prediction model module is mainly used to build a performance prediction model for electronic components based on the output structure of the LSTM network, and predict various change trend functions in the future time period;

[0060] Multi-scenario prediction and analysis module: The multi-scenario prediction and analysis module generates multiple prediction scenarios based on different working conditions and disturbance factors combined with orthogonal experimental design;

[0061] Error calculation module: The error calculation module is mainly used to compare the real-time collected performance data with the prediction model results, calculate the error and dynamically assign weights based on the error contribution to obtain the prediction model with the minimum error value;

[0062] Online update and closed-loop optimization module: The online update and closed-loop optimization module is mainly used to perform incremental learning based on model correction parameters, update the prediction model online, automatically adjust parameters and continuously optimize performance prediction results;

[0063] Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

[0064] See Figure 1 As shown, the electronic component performance data monitoring method includes:

[0065] Step 1: Based on the working history data of the electronic component, input the LSTM neural network model for training, build the electronic component performance prediction model, and obtain the function of the electronic component performance changing with working time;

[0066] Step 2: Obtain current electronic component surface temperature field distribution data, partial discharge characteristics, and leakage current dynamic parameters, input them into the electronic component performance prediction model, and obtain multiple predicted change trend functions of the electronic component performance in the future time period based on the changes in various factors;

[0067] Step 3: Based on the obtained multiple predicted change trend function data, compare it with the actual electronic component performance data collected in real time, calculate the error value, and obtain the predicted data with the smallest error;

[0068] Step 4: Based on the predicted data with the minimum error, perform dynamic error analysis and generate model correction parameters;

[0069] Step 5: Update the parameters of the electronic component performance prediction model online based on the model correction parameters to form a closed-loop optimization system.

[0070] See Figure 2 As shown in the figure, based on the working history data of electronic components, the LSTM neural network model is input for training to build an electronic component performance prediction model. The function of how the electronic component's performance changes with working time is obtained. Specifically, the following are included:

[0071] The working history data of electronic components is transferred to the LSTM neural network model for training to build an electronic component performance prediction model;

[0072] Based on the dual-branch structure of the output layer in the electronic component performance prediction model, the main branch predicts the performance parameters at the current moment, and the auxiliary branch outputs the performance degradation rate. The functional relationship between the performance parameters and the working time is constructed through integral operation.

[0073] Specifically, a dual-branch structure is designed in the output layer of the LSTM network. The main branch is used to predict the performance parameters at the current moment, and the auxiliary branch is used to output the performance degradation rate.

[0074] The output of LSTM in the main branch is , the output of the main branch is the performance parameter at the current moment , predicted through a fully connected layer, the formula is:

[0075] ;

[0076] in, is the predicted value of the performance parameter at the current moment, is the output of LSTM, is a trainable weight matrix used to linearly map high-dimensional hidden states to target parameter dimensions and determine feature weight distribution. is a trainable bias term;

[0077] The output of the auxiliary branch is the performance degradation rate Prediction is performed through a fully connected layer, the formula is:

[0078] ;

[0079] in, The output of the auxiliary branch is the performance degradation rate, is the weight matrix of the fully connected layer, is the bias term of the fully connected layer;

[0080] The functional relationship between performance parameters and working time is established through integral operation. Specifically, performance parameters Performance degradation rate The updated formula is:

[0081] ;

[0082] The degradation rate By integrating the time, the performance parameters can be obtained The relationship with time is:

[0083] ;

[0084] in, is the performance parameter value at time t, is the integral variable, indicating that Any instant in the time period from t to t;

[0085] See Figure 3 As shown, the current electronic component surface temperature field distribution data, partial discharge characteristics and leakage current dynamic parameters are obtained and input into the electronic component performance prediction model. Based on the changes in various factors, multiple prediction change trend functions of the electronic component performance in the future time period are obtained, including:

[0086] Scan the surface of electronic components with a non-contact infrared thermal imager array to obtain a two-dimensional temperature distribution matrix and extract characteristic parameters;

[0087] The pulse signal is collected by a high-frequency current sensor to obtain the characteristic value of partial discharge;

[0088] Acquire current characteristic parameters through dual-channel synchronous acquisition;

[0089] The preprocessed collected data is input into the electronic component performance prediction model, and based on the possible change trends of the electronic component parameter data, a multi-scenario electronic component performance prediction change trend function is generated.

[0090] Specifically, a non-contact infrared thermal imager array is used to scan the surface of electronic components to obtain a two-dimensional temperature distribution matrix, and characteristic data including average temperature, temperature resistance, and maximum and minimum temperatures are extracted from the temperature matrix; a high-frequency current sensor is used to collect pulse signals to obtain partial discharge characteristics, and the characteristic parameters of partial discharge are extracted by analyzing the amplitude, frequency, and duration of the current signal; and current characteristic parameters, including the maximum current value and the average value of the waveform, are obtained by synchronously collecting current signals through two channels.

[0091] The features extracted from different sensors are fused and fed into the electronic component performance prediction model as input data.

[0092] See Figure 4 As shown, the pre-processed collected data is input into the electronic component performance prediction model. Based on the possible change trend of the electronic component parameter data, a multi-scenario electronic component performance prediction change trend function is generated, specifically including:

[0093] Based on the characteristic parameters that affect the performance of electronic components, the variation range of each disturbance dimension is obtained;

[0094] Through orthogonal experimental design, various parameter change scenarios are combined to generate multiple scenarios of electronic component performance changes;

[0095] Based on the characteristic parameter data under various scenarios, they are input into the electronic component performance prediction model to generate a variety of electronic component performance prediction trend functions. The characteristic data obtained include: performance parameter time series, performance degradation acceleration, and key threshold arrival time prediction.

[0096] Specifically, the performance of electronic components is affected by multiple factors, such as temperature, current, voltage, and operating frequency. Based on historical data and changes in the operating status of the components, different disturbance dimensions are determined to obtain the minimum, maximum, and variation range of each feature.

[0097] According to the perturbation dimension and its variation range, select an appropriate orthogonal table to carry out experimental design, ensuring that each experimental combination can effectively evaluate the impact of different perturbation dimensions and obtain a variety of different experimental scenarios (combinations), that is, the specific values of characteristic parameters under each experimental condition;

[0098] Based on the characteristic parameter data under different scenarios, it is input into the electronic component performance prediction model to generate different performance prediction change trend functions;

[0099] By predicting the time series data under each scenario, we can obtain the trend of performance parameters changing over time, and the formula is:

[0100] ;

[0101] in, is the performance parameter at the initial moment, It's in the situation degradation rate under ;

[0102] Performance degradation acceleration refers to the change in performance degradation rate over time. For each scenario, the LSTM model can output the degradation rate , and further deduce its acceleration, the formula is:

[0103] ;

[0104] in, is the performance degradation acceleration;

[0105] By predicting the changes in performance parameters over time, it is possible to predict when a critical threshold (e.g., when performance drops to a critical value) will be reached.

[0106] See Figure 5 As shown, based on the obtained multiple predicted change trend function data, it is compared with the actual data of electronic component performance acquired in real time, the error value is calculated, and the predicted data with the smallest error is obtained, which specifically includes:

[0107] Based on the error contribution of each characteristic parameter, dynamic error weight allocation is performed;

[0108] Compare the real-time collected data with the data of the corresponding time nodes in the predicted change trend function, and obtain the error size between each predicted change trend function data and the actual collected data through multi-dimensional error joint calculation;

[0109] A set of prediction change trend functions with the smallest error value in the error data is selected as the best electronic component performance prediction model;

[0110] Based on the actual performance data of electronic components collected at each moment, the prediction results of an optimal electronic component performance prediction model are screened and obtained.

[0111] Specifically, in the error contribution calculation of the feature parameters, the contribution is calculated based on the square of each feature error. , the formula is:

[0112] ;

[0113] in, is the error contribution, is the error contribution of the i-th characteristic parameter at time t, is the sum of square errors of all n feature parameters at time t;

[0114] According to the error contribution , dynamically adjust the weight of each feature;

[0115] After real-time data is collected, it needs to be compared with the change trend function predicted by the model to calculate the error between the real-time data and the predicted data. Through the joint calculation of multi-dimensional errors, the error size at each moment is obtained, including absolute error, relative error, and trend error.

[0116] By comparing the errors of different prediction models, the prediction model with the smallest error is selected as the best prediction model. Once the best model is selected, the prediction results of the model can be used to make subsequent performance predictions based on the actual data at each moment.

[0117] In actual applications, as real-time data is continuously updated, the error contribution and dynamic weights can be recalculated based on the new error data, and the model parameters can be adjusted. In this way, the model is gradually optimized and can more accurately predict the performance of electronic components.

[0118] See Figure 6 As shown in the figure, based on the prediction data with the minimum error, dynamic error analysis is performed to generate model correction parameters, including:

[0119] Based on the forecast data with the minimum error, the error fluctuation is obtained through the error distribution diagram;

[0120] Through the relationship between error and working conditions, the cause of error under specific conditions can be obtained;

[0121] By comparing the error distribution graph with the model prediction output, the deviation in data, model structure or parameter setting can be obtained;

[0122] Based on the results of dynamic error analysis, the model parameters are adjusted to generate model correction parameters, including: adjustment of learning rate, addition of regularization terms, and adjustment of training data volume.

[0123] Specifically, by calculating the statistics of the error sequence and obtaining the distribution of the errors, and by analyzing the relationship between the errors and the working conditions, we can gain a deeper understanding of the causes of the errors under specific conditions.

[0124] By comparing the error distribution graph with the model prediction output, analyze the deviations in data, model structure or parameter settings, including:

[0125] Error bias: If the errors are mostly concentrated on the positive or negative side, it indicates that the prediction model may be biased. For example, the model may be more optimistic in low temperature conditions and more pessimistic in high temperature conditions.

[0126] Error fluctuation: If the error fluctuates significantly over time, it may mean that the model's predictions are unstable at certain points in time;

[0127] The degree of match between model output and data: If the distribution of model output is inconsistent with the distribution of actual data, there may be a problem with the model structure or parameters not being suitable for the current data;

[0128] Based on the error analysis results, the model parameters are dynamically adjusted. The amplitude of parameter adjustment during each model update is adjusted by adjusting the learning rate. Regularization terms are added to prevent model overfitting. The generalization ability of the model is improved by increasing the amount of training data.

[0129] Based on the results of the above steps, a set of model correction parameters can be generated to improve the predictive ability of the model.

[0130] See Figure 7 As shown in FIG, the online parameter update of the electronic component performance prediction model based on the model correction parameters is performed to form a closed-loop optimization system, which specifically includes:

[0131] Based on the obtained model correction parameters, the model is trained online through incremental learning, and the model is continuously optimized using the real-time electronic component performance data and error correction information;

[0132] Through the automated parameter adjustment mechanism, when the error exceeds the set threshold, the model parameter update is automatically triggered;

[0133] The model prediction results after each update are compared with the real-time data again to form a continuous feedback process.

[0134] Specifically, gradient descent is used to update the model parameters online. The formula is:

[0135] ;

[0136] in, is the current parameter of the model, is the gradient of the loss function with respect to the parameters, is the current learning rate;

[0137] After updating the model, based on Generate new prediction results, compare the error between the updated prediction value and the actual value, and add the error history for the next step of dynamic weight or model correction judgment;

[0138] Based on the error trend, the learning rate parameters are automatically adjusted to avoid continuous increase or decrease in the error.

[0139] Furthermore, the method according to the embodiment of the present application can also be used with the aid of Figure 8 The electronic device architecture shown in FIG. Figure 8 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the electronic component performance data monitoring method and system provided in this application. The electronic device 500 may also include a terminal interface 508. Of course, Figure 8 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 8 One or more components of an electronic device are shown.

[0140] Figure 9 This is a schematic diagram of the computer-readable storage medium structure provided by an embodiment of the present application. Figure 9, a computer-readable storage medium 600 according to one embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by a processor, the electronic component performance data monitoring method and system according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 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), a hard disk, flash memory, etc.

[0141] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0142] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0143] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring electronic component performance data, characterized in that: include: Based on the working history data of electronic components, the LSTM neural network model is input for training to build an electronic component performance prediction model and obtain the function of the electronic component performance changing with working time; Obtain current electronic component surface temperature field distribution data, partial discharge characteristics, and leakage current dynamic parameters, input them into the electronic component performance prediction model, and obtain multiple predicted change trend functions of the electronic component performance in the future time period based on the changes in various factors; Based on the obtained multiple predicted change trend function data, it is compared with the actual performance data of the electronic components collected in real time, the error value is calculated, and the predicted data with the smallest error is obtained; Based on the prediction data with the minimum error, dynamic error analysis is performed to generate model correction parameters; The electronic component performance prediction model is updated online based on the model correction parameters to form a closed-loop optimization system.

2. The electronic component performance data monitoring method according to claim 1, characterized in that: The method of inputting the LSTM neural network model into the working history data of the electronic component for training, constructing the electronic component performance prediction model, and obtaining the function of the electronic component performance changing with working time specifically includes: The working history data of electronic components is transferred to the LSTM neural network model for training to build an electronic component performance prediction model; Based on the dual-branch structure of the output layer in the electronic component performance prediction model, the main branch predicts the performance parameters at the current moment, and the auxiliary branch outputs the performance degradation rate. The functional relationship between the performance parameters and the working time is constructed through integral operation.

3. The electronic component performance data monitoring method according to claim 1, characterized in that: The method of obtaining current electronic component surface temperature field distribution data, partial discharge characteristic quantities, and leakage current dynamic parameters, and inputting them into the electronic component performance prediction model, and obtaining multiple prediction change trend functions of the electronic component performance in the future time period based on the changes of various factors, specifically includes: Scan the surface of electronic components with a non-contact infrared thermal imager array to obtain a two-dimensional temperature distribution matrix and extract characteristic parameters; The pulse signal is collected by a high-frequency current sensor to obtain the characteristic value of partial discharge; Acquire current characteristic parameters through dual-channel synchronous acquisition; The preprocessed collected data is input into the electronic component performance prediction model, and based on the possible change trends of the electronic component parameter data, a multi-scenario electronic component performance prediction change trend function is generated.

4. The electronic component performance data monitoring method according to claim 3, characterized in that: Inputting the pre-processed collected data into the electronic component performance prediction model and generating a multi-scenario electronic component performance prediction trend function based on the possible change trend of the electronic component parameter data specifically includes: Based on the characteristic parameters that affect the performance of electronic components, the variation range of each disturbance dimension is obtained; Through orthogonal experimental design, various parameter change scenarios are combined to generate multiple scenarios of electronic component performance changes; Based on the characteristic parameter data under various scenarios, they are input into the electronic component performance prediction model to generate a variety of electronic component performance prediction trend functions. The characteristic data obtained include: performance parameter time series, performance degradation acceleration, and key threshold arrival time prediction.

5. The electronic component performance data monitoring method according to claim 1, characterized in that: The method of comparing the obtained multiple predicted change trend function data with the actual electronic component performance data collected in real time, calculating the error value, and obtaining the predicted data with the smallest error specifically includes: Based on the error contribution of each characteristic parameter, dynamic error weight allocation is performed; Compare the real-time collected data with the data of the corresponding time nodes in the predicted change trend function, and obtain the error size between each predicted change trend function data and the actual collected data through multi-dimensional error joint calculation; A set of prediction change trend functions with the smallest error value in the error data is selected as the best electronic component performance prediction model; Based on the actual performance data of electronic components collected at each moment, the prediction results of an optimal electronic component performance prediction model are screened and obtained.

6. The electronic component performance data monitoring method according to claim 1, characterized in that: The dynamic error analysis based on the prediction data with the minimum error and the generation of model correction parameters specifically include: Based on the forecast data with the minimum error, the error fluctuation is obtained through the error distribution diagram; Through the relationship between error and working conditions, the cause of error under specific conditions can be obtained; By comparing the error distribution graph with the model prediction output, the deviation in data, model structure or parameter setting can be obtained; Based on the results of dynamic error analysis, the model parameters are adjusted to generate model correction parameters, including: adjustment of learning rate, addition of regularization terms, and adjustment of training data volume.

7. The electronic component performance data monitoring method according to claim 1, characterized in that: The online parameter updating of the electronic component performance prediction model based on the model correction parameters to form a closed-loop optimization system specifically includes: Based on the obtained model correction parameters, the model is trained online through incremental learning, and the model is continuously optimized using the real-time electronic component performance data and error correction information; Through the automated parameter adjustment mechanism, when the error exceeds the set threshold, the model parameter update is automatically triggered; The model prediction results after each update are compared with the real-time data again to form a continuous feedback process.

8. In combination with an electronic component performance data monitoring system, the method for monitoring electronic component performance data according to any one of claims 1 to 7 is implemented, characterized in that: include: Model training module: The model training module is mainly used to input the working history data of electronic components into the LSTM neural network for training, build a performance prediction model and obtain a function of performance changes over time; Performance prediction model module: The performance prediction model module is mainly used to build a performance prediction model for electronic components based on the output structure of the LSTM network, and predict various change trend functions in the future time period; Multi-scenario prediction and analysis module: The multi-scenario prediction and analysis module generates multiple prediction scenarios based on different working conditions and disturbance factors combined with orthogonal experimental design; Error calculation module: The error calculation module is mainly used to compare the real-time collected performance data with the prediction model results, calculate the error and dynamically assign weights based on the error contribution to obtain the prediction model with the minimum error value; Online update and closed-loop optimization module: The online update and closed-loop optimization module is mainly used to perform incremental learning based on model correction parameters, update the prediction model online, automatically adjust parameters and continuously optimize performance prediction results; Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

9. An electronic device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the electronic component performance data monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by a processor, the electronic component performance data monitoring method according to any one of claims 1 to 7 is implemented.

Images