Fault prediction method for front-end isolated switching power supply of aging table

Abstract

The invention discloses a fault prediction method for a front-end isolated switching power supply of an aging table, and the method comprises the steps: carrying out the dimension reduction of preset time sequence signals of a plurality of aging table switching power supplies, and extracting feature parameters; calculating the performance degradation path similarity of the switch power supply of each aging station according to the characteristic parameters of each preset time sequence signal; clustering the degradation paths according to the performance degradation path similarity of the switch power supplies of the aging stations to obtain N degradation path sets; establishing N switching power supply life prediction models, and training the N switching power supply life prediction models according to the N degradation path sets; and based on the trained N switching power supply service life prediction models, predicting the service life of the aging station switching power supply to obtain the residual service life of the switching power supply. According to the aging table front-end isolated switching power supply fault prediction method considering multiple degradation paths provided by the invention, the residual service life of the switching power supply can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of lossless reliability screening of integrated circuits, and in particular to a method and device for predicting faults of a front-end isolated switching power supply of a burn-in station taking multiple degradation paths into consideration. Background Art

[0002] Existing high-temperature aging products can achieve long-term monitoring of the test environment by adding long-term monitoring in the machine and overstress protection mechanisms, and can respond in time when the product fails, realizing post-maintenance based on fault data. However, due to the lack of active protection technology for high-temperature aging products, existing high-temperature aging products make it difficult for related products to achieve the integrity of the test process and the consistency of the test environment stress. It is very easy to cause the test process to be interrupted due to product failure, resulting in major property losses such as damage to millions of test devices (also known as test pieces), or the aging test is identified as a failure test due to adverse effects such as additional stress introduced during the test due to product performance degradation, resulting in ineffective waste of resources. At the same time, ensuring the quality of high-temperature aging tests can avoid over-aging and under-aging failures in integrated circuits, greatly reduce the failure rate of integrated circuits, and avoid large-scale electronic system failures and shutdowns such as new energy vehicles, civil airliners, and energy storage transmission stations due to integrated circuit failures.

[0003] Given that high-temperature aging products such as aging stations operate under harsh conditions such as high temperatures for a long time, the risk of failure of key equipment is relatively high. Therefore, the failure prediction and health management of key equipment in aging stations has become an important research field. Effective health assessment and failure prediction can promote the development of reliable maintenance plans to prevent potential failures. Key equipment in aging stations, such as primary switching power supplies, play a core role in the system. Once a failure occurs, it will have a serious impact on the entire system, which will not only cause test failures, but may also cause equipment damage and trigger safety accidents. Therefore, it is crucial to carry out accurate life prediction of the primary switching power supply of the aging station.

[0004] The existing technology mainly carries out life prediction based on the individual differences and degradation characteristics of equipment or systems, but there is little research on the impact of different degradation paths of primary switching power supplies on the aging bench on their life prediction results. However, since the primary switching power supply system generally has multiple degradation paths and other unstable situations, the data-driven life prediction method will have the following two problems: 1. Multiple degradation paths: There are multiple degradation paths for primary switching power supplies, which are reflected in the aging test bench that has not reached the end of its life as multiple performance degradation paths. Different performance degradation paths have different effects on the remaining life, which affects the life prediction; 2. Long time series: The time series of the current and voltage parameters of the primary switching power supply is long, and the degradation information in the series is difficult to capture and learn, which affects the life prediction. Due to the existence of the above two problems, it is currently impossible to carry out accurate life prediction for the primary switching power supply of the aging bench. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a method and device for predicting faults of a front-end isolated switching power supply of a burn-in station, and electronic equipment, which can solve the above-mentioned problems existing in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: An embodiment of the present invention provides a method for predicting faults of a front-end isolated switching power supply of a burn-in station, wherein the method comprises: Perform dimension reduction processing on the preset timing signals of multiple aging station switching power supplies and extract characteristic parameters; Calculating the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals; According to the similarity of the performance degradation paths of the switching power supplies of each aging station, each degradation path is clustered to obtain N degradation path sets; Establishing N switch power supply life prediction models, and training the N switch power supply life prediction models according to the N degradation path sets respectively; Based on the trained N switching power supply life prediction models, the life of the switching power supply on the aging station is predicted to obtain the remaining life of the switching power supply.

[0007] Optionally, the characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity and thermal diffusivity, etc.

[0008] Optionally, the step of calculating the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals includes: Construct parameter vectors for the characteristic parameters of each of the preset timing signals respectively; constructing a matrix grid according to each of the parameter vectors; Based on preset path boundary constraints, path continuity constraints and path monotonicity constraints, a regular minimum cost path passing through the matrix grid is found through multiple iterative adjustments, wherein the regular minimum cost path can characterize the similarity of the performance degradation paths of the switching power supplies of each aging station.

[0009] Optionally, according to the similarity of the performance degradation paths of the switch power supplies of each aging station, the step of clustering the degradation paths to obtain N degradation path sets includes: Randomly select K points from the regular minimum cost path as the center of the initial cluster; For each point in the regular minimum cost path, assign each point to the cluster closest to the center point; Update the center point of each cluster; For each point in the regular minimum cost path, assign each point to a cluster closest to the updated center point; Determine whether the clustering stop condition is met, and if not, return to execute the step of updating the center point of each cluster; If it has been reached, the clusters of the updated center points are clustered as the N degenerate path sets.

[0010] Optionally, the step of updating the center point of each cluster includes: For each cluster, calculate the average value of each point in the cluster; The center point of the cluster is updated to the location of the average value.

[0011] Optionally, the switch power supply life prediction model includes: an input layer, a Dense layer (i.e., a dense layer), a BiLSTM layer, i.e., a bidirectional long short-term memory network layer, for processing multi-sensor monitoring signals, a BiGRU layer, i.e., a bidirectional gated loop layer, for processing operating condition data, i.e., the transformation result after hidden representation), a Concatenate layer, i.e., a feature concatenation layer, for fusing the outputs of the BiLSTM and BiGRU layers, a multidimensional feature map, and an output layer for outputting RUL prediction values. Optionally, N switch power supply life prediction models are established, and the steps of training the N switch power supply life prediction models according to the N degradation path sets respectively include: Establish N switching power supply life prediction models; For each of the switching power supply life prediction models, the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set are used as model inputs; Performing a linear transformation on the input feature parameters in the fully connected layer, and performing a hidden representation on the transformation results; Constructing a high-order vector according to the transformation result after the hidden representation, and sending the high-order vector to the bidirectional long short-term memory network layer and the bidirectional gated recurrent layer to obtain hidden feature maps of different dimensions; The feature vectors generated by merging hidden features of different dimensions are input into two dense layers to generate RUL prediction values; The error between the RUL predicted value and the RUL actual value is used as a loss function to calculate the loss; Updating the parameters of the switching power supply life prediction model according to the loss back propagation; Determine whether the switching power supply life prediction model after updating the parameters meets the preset accuracy; If satisfied, determining that the training of the switching power supply life prediction model is completed; If not satisfied, return to the step of using the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set as model input.

[0012] The embodiment of the present invention further provides a device for predicting faults of a front-end isolated switching power supply of a burn-in station, wherein the device comprises: A feature extraction module is used to perform dimensionality reduction processing on preset timing signals of multiple aging station switching power supplies and extract feature parameters; A similarity determination module, used to calculate the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals; A clustering module, used to cluster the degradation paths according to the similarity of the performance degradation paths of the switching power supplies of each aging station to obtain N degradation path sets; A model training module, used to establish N switch power supply life prediction models, and train the N switch power supply life prediction models according to the N degradation path sets respectively; The model prediction module is used to predict the life of the switching power supply of the aging station based on the N trained switching power supply life prediction models to obtain the remaining life of the switching power supply.

[0013] Optionally, the characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity and thermal diffusivity, etc.

[0014] Optionally, the similarity determination module includes: The first submodule is used to construct parameter vectors for characteristic parameters of each of the preset timing signals; A second submodule is used to construct a matrix grid according to each of the parameter vectors; The third submodule is used to find the regular minimum cost path through the matrix grid through multiple iterative adjustments based on preset path boundary constraints, path continuity constraints and path monotonicity constraints, wherein the regular minimum cost path can characterize the similarity of the performance degradation paths of the switching power supplies of each aging station.

[0015] Optionally, the clustering module includes: A fourth submodule is used to randomly select K points from the regular minimum cost path as the center of the initial cluster; A fifth submodule is used to assign each point in the regular minimum cost path to a cluster closest to the center point; The sixth submodule is used to update the center point of each cluster; A seventh submodule is used to assign each point in the regular minimum cost path to a cluster closest to the updated center point; An eighth submodule is used to determine whether the clustering stop condition is met, and if not, return to execute the sixth submodule; If the ninth submodule has been reached, the clusters of the updated center points are clustered as N degenerate path sets.

[0016] Optionally, the sixth submodule is specifically used for: For each cluster, calculate the average value of each point in the cluster; The center point of the cluster is updated to the location of the average value.

[0017] Optionally, the switching power supply life prediction model includes: Input layer, Dense layer (i.e. dense layer), BiLSTM layer (i.e. bidirectional long short-term memory network layer), used to process multi-sensor monitoring signals, BiGRU layer (i.e. bidirectional gated recurrent layer), used to process operation condition data (i.e. transformation results after hidden representation), Concatenate layer (i.e. feature concatenation layer), used to fuse the outputs of BiLSTM and BiGRU layers, multi-dimensional feature maps, and output layer, used to output RUL prediction values.

[0018] Optionally, the model training module is specifically used for: Establish N switching power supply life prediction models; For each of the switching power supply life prediction models, the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set are used as model inputs; Performing a linear transformation on the input feature parameters in the fully connected layer, and performing a hidden representation on the transformation results; Constructing a high-order vector according to the transformation result after the hidden representation, and sending the high-order vector to the BLSTM layer and the BGRU layer to obtain hidden feature maps of different dimensions; The feature vectors generated by merging hidden features of different dimensions are input into two dense layers (also called linear regression dense layers) to generate RUL prediction values; The error between the RUL predicted value and the RUL actual value is used as a loss function to calculate the loss; Updating the parameters of the switching power supply life prediction model according to the loss back propagation; Determine whether the switching power supply life prediction model after updating the parameters meets the preset accuracy; If satisfied, determining that the training of the switching power supply life prediction model is completed; If not satisfied, return to execute the operation of using the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set as model input.

[0019] An embodiment of the present invention also provides an electronic device, characterized in that it includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; the memory is used to store computer programs; the processor is used to implement any of the above-mentioned methods for predicting the fault of the isolated switching power supply at the front end of the aging station when executing the program stored in the memory.

[0020] The embodiment of the present invention discloses a fault prediction scheme for isolated switching power supplies at the front end of a burn-in station. The scheme performs dimensionality reduction processing on preset timing signals of multiple switching power supplies at the burn-in station and extracts characteristic parameters. The scheme calculates the similarity of the performance degradation paths of the switching power supplies at each burn-in station based on the characteristic parameters of each preset timing signal. The scheme clusters the degradation paths based on the similarity of the performance degradation paths of the switching power supplies at each burn-in station to obtain N degradation path sets. The scheme establishes N switching power supply life prediction models and trains the N switching power supply life prediction models based on the N degradation path sets respectively. The scheme predicts the life of the switching power supply at the burn-in station based on the trained N switching power supply life prediction models to obtain the remaining life of the switching power supply. The embodiment of the present invention discloses a fault prediction scheme for the isolated switching power supply at the front end of the aging station. On the one hand, the life prediction model of the switching power supply of the aging station based on the deep neural network comprehensively considers the influence of the performance degradation path on the prediction of the remaining service life of the switching power supply, and can achieve high-precision life prediction; on the second hand, the failure mode is taken into consideration, the performance degradation path of the switching power supply of the aging station is identified, and the characteristic parameters of the switching power supply of the aging station with the same performance degradation path are brought into the same model for training and prediction, which is beneficial to improving the life prediction accuracy of the switching power supply of the aging station; on the third hand, since the fault of the isolated switching power supply at the front end of the aging station can be predicted, the damage to the inspected component caused by the fault of the isolated switching power supply during the aging detection process can be effectively avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart showing a method for predicting a fault of a front-end isolated switching power supply of a burn-in station according to an embodiment of the present application; Figure 2 1 is a schematic diagram showing the principle of a method for predicting a fault of a front-end isolated switching power supply of a test bench according to an embodiment of the present application; Figure 3 is a schematic diagram showing the DTW principle of an embodiment of the present application; Figure 4 is a flowchart showing the steps of a method for evaluating the health status of a secondary power supply of a high temperature aging test equipment according to an embodiment of the present application; Figure 5 It is a structural block diagram of a device for predicting faults of an isolated switching power supply at the front end of a burn-in station according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0023] In conjunction with the accompanying drawings, the method for predicting the fault of the isolated switching power supply at the front end of the aging station provided by the embodiment of the present application is described in detail through specific embodiments and application scenarios.

[0024] As attached Figure 1 As shown, the method for predicting the fault of the isolated switching power supply at the front end of the burn-in station of the embodiment of the present application includes the following steps: Step 101: performing dimensionality reduction processing on preset timing signals of multiple aging station switching power supplies, and extracting characteristic parameters.

[0025] The method for predicting the fault of the isolated switching power supply at the front end of the aging station provided by the implementation of the present application can be applied to electronic devices, and a computer program for predicting the life of the switching power supply of the aging station is provided in the electronic device. When the computer program is executed by the processor, the method for predicting the life of the switching power supply of the aging station of the embodiment of the present application is implemented to predict the remaining life of the switching power supply. The switching power supply of the aging station described in the embodiment of the present application is the isolated switching power supply at the front end of the aging station, and can also be a primary switching power supply, which can also be referred to as the switching power supply below, and the aging station can also be called a high-temperature aging station.

[0026] The aging station provided in the embodiment of the present application can perform non-destructive aging testing on the parts to be tested. In the embodiment of the present application, the feature extraction and life prediction of the switching power supply of the aging station are to ensure the safety of the parts to be tested.

[0027] Among them, characteristic parameters include but are not limited to: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity, thermal diffusivity, etc.

[0028] The preset timing signal can be a timing signal of the current, voltage, etc. of the switching power supply. The multiple aging station switching power supplies are multiple aging station switching power supplies with different decay rates and different fault initiation degrees.

[0029] Step 102: Calculate the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each preset timing signal.

[0030] In the actual implementation process, the DTW (Dynamic Time Warping) method can be used to evaluate the similarity of the paths between the switching power supplies of multiple aging stations with different decay rates and different fault initiation degrees (that is, the similarity of the performance degradation paths of the switching power supplies of each aging station), and on this basis, the K-Means performance degradation path clustering method is introduced to achieve fast and accurate classification of the performance degradation paths of the switching power supplies of the aging station (that is, the aging test station).

[0031] Optionally, a method for calculating the similarity of the performance degradation paths of the switch power supplies of each aging station based on the characteristic parameters of each preset timing signal may include the following sub-steps: Sub-step 1: construct parameter vectors for the characteristic parameters of each preset time series signal respectively; Sub-step 2: constructing a matrix grid according to each of the parameter vectors; For the performance parameter vector of the primary switching power supply sample of the aging station , (a, b are sample numbers, m, n are corresponding running cycle numbers), then construct a The matrix grid, matrix elements express and Distance between two points , each matrix element express and Alignment.

[0032] Sub-step 3: Based on the preset path boundary constraints, path continuity constraints, and path monotonicity constraints, find the regular minimum cost path through the matrix grid through multiple iterative adjustments.

[0033] Among them, the regularized minimum cost path can characterize the similarity of the performance degradation paths of the switching power supplies of each aging station.

[0034] When implementing sub-step 3, a regular path through the grid can be found based on the path boundary constraints, path continuity constraints, and path monotonicity constraints, and then used To express: 1. Then, based on these three constraints, the path is iterated multiple times until the path with the minimum cost is obtained. , in the denominator Used to compensate for regular paths of different lengths.

[0035] Path boundary condition constraints can be set as , , ensuring that the selected path must start from the common starting point of the sequence and end at the common ending point of the sequence.

[0036] The path continuity constraint can be set as follows: , then for the next point on the path Needs to be satisfied and ,ensure and Each coordinate in Appeared in.

[0037] The path monotonicity constraint can be set as follows: , then for the next point on the path Needs to be satisfied and ,limit The above points must be monotonically progressing over time.

[0038] The DTW method is used to calculate the similarity of characteristic time series trajectories such as ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity, thermal diffusivity, etc. between different aging station switching power supply samples (i.e., preset timing signals). Subsequently, the clustering method can be used to cluster the time series trajectories, so that the aging station switching power supplies can be classified according to the performance degradation trajectory.

[0039] Step 103: clustering each degradation path according to the similarity of the performance degradation paths of the switch power supplies of each aging station to obtain N degradation path sets.

[0040] N is a positive integer greater than or equal to 2. In the embodiment of the present application, N equals 2 as an example for illustration.

[0041] In an optional embodiment, clustering each degradation path according to the similarity of the performance degradation paths of the switch power supplies of each aging station to obtain N degradation path sets may include the following sub-steps: Sub-step 1: Randomly select K points from the regular minimum cost path as the centers of the initial cluster; Sub-step 2: For each point in the regularized minimum cost path, assign each point to the cluster closest to the center point; Sub-step 3: Update the center point of each cluster; An exemplary method of updating the center point of each cluster may be: for each cluster, calculating the average value of each point in the cluster; and updating the center point of the cluster to the location of the average value. The goal of updating the center point of the cluster is to minimize the squared error within the cluster.

[0042] Sub-step 4: For each point in the regularized minimum cost path, assign each point to the cluster closest to the updated center point; Sub-step 5: Determine whether the clustering stop condition is met. If not, return to sub-step 3 to update the center point of each cluster. If not reached, return to execute sub-step 3, sub-step 4 and sub-step 5, so as to update the center point again and cluster based on the updated center point.

[0043] The clustering stop condition can be flexibly set by those skilled in the art, and is not specifically limited in the embodiments of the present application. For example, the clustering stop condition can be set to: the change in the position of the cluster center point before and after the update is less than a preset value (the position of the cluster center point before and after the update no longer changes significantly), or it can be set to reach a set upper limit of the number of iterations.

[0044] This optional degradation path clustering method takes advantage of the time used for cluster centroid selection and the complexity of cluster overlapping space in K-Means, and takes the optimal point that actually exists in the data set as its centroid. Therefore, the characteristics of the trajectory corresponding to the centroid can be found, and the system samples are clustered through the performance degradation path similarity matrix of the aging station switching power supply, so as to realize the identification of the aging station switching power supply according to the performance degradation trajectory time series (i.e., the preset timing signal).

[0045] Sub-step 6: If it has been reached, the clusters of the updated center points are used as the N degenerate path sets.

[0046] Step 104: Establish N switch power supply life prediction models, and train the N switch power supply life prediction models according to the N degradation path sets respectively.

[0047] Among them, the switching power supply life prediction model includes: Input layer, Dense layer (i.e. dense layer), BiLSTM layer (i.e. bidirectional long short-term memory network layer), used to process multi-sensor monitoring signals, BiGRU layer (i.e. bidirectional gated recurrent layer), used to process operation condition data (i.e. transformation results after hidden representation), Concatenate layer (i.e. feature concatenation layer), used to fuse the outputs of BiLSTM and BiGRU layers, multi-dimensional feature maps, and output layer, used to output RUL prediction values.

[0048] In an optional embodiment, establishing N switch power supply life prediction models, and training the N switch power supply life prediction models according to N degradation path sets respectively may include the following sub-steps: Sub-step 1: Establish N switching power supply life prediction models.

[0049] When there are two performance degradation paths, two switching power supply life prediction models are established, and the switching power supply life prediction models are used to predict the remaining life of the switching power supply.

[0050] Sub-step 2: For each switching power supply life prediction model, the characteristic parameters corresponding to the degradation path in the corresponding degradation path set are used as model input.

[0051] Sub-step 3: Perform linear transformation on the input feature parameters in the fully connected layer and hide the transformation results; Among them, the input feature parameters can be regarded as multi-sensory monitoring data X i .

[0052] Sub-step 4: Construct a high-order vector based on the transformation result after the hidden representation, and send the high-order vector to the BiLSTM layer and BiGRU layer to obtain hidden feature maps of different dimensions; Among them, the number of stacked layers of BiLSTM layer and BiGRU layer is the same.

[0053] Sub-step 5: The feature vectors generated by merging the hidden features of different dimensions are input into two linear regression dense layers to generate RUL prediction values; Sub-step 6: Calculate the loss by using the error between the RUL prediction value and the RUL actual value as the loss function; Sub-step 7: updating the parameters of the switching power supply life prediction model according to the loss back propagation; Sub-step 8: Determine whether the switching power supply life prediction model after the updated parameters meets the preset accuracy; if so, determine that the training of the switching power supply life prediction model is completed; if not, return to the step of using the characteristic parameters corresponding to the degradation path in the corresponding degradation path set as the model input.

[0054] The prediction accuracy is used to evaluate the performance of the prediction model. The accuracy of the switching power supply life prediction model can be calculated using the following formula:

[0055] Among them, RUL is the actual RUL (remaining useful life); is the predicted RUL estimated by the prediction model; TUL is the actual total service life of the target circuit, which is defined as the total number of test cycles when the switching power supply of the aging station reaches the failure threshold.

[0056] The life prediction model of the switching power supply can be a multidimensional recurrent neural network (MDRNN) model. The multidimensional recurrent neural network uses momentum and adaptive learning rate to accelerate the convergence speed, and is proven to be superior to the classic stochastic gradient descent optimizer, thereby obtaining higher processing efficiency and prediction quality. The training process of the multidimensional recurrent neural network (MDRNN) model includes: S1: extracting the characteristic parameters of the switching power supply of the aging station; S2: establishing the MDRNN multidimensional recurrent neural network model; S3: initializing the parameters of the MDRNN multidimensional recurrent neural network model; S4: iteratively training the MDRNN multidimensional recurrent neural network model based on the characteristic parameters in S1 for multiple times, and calculating the loss function after each training is completed, and adjusting the model parameters based on the calculated loss function; after multiple rounds of forward and back propagation training, after reaching the stop state, the trained MDRNN multidimensional recurrent neural network model can be used to predict the remaining life of the switching power supply.

[0057] Step 105: Based on the trained N switch power supply life prediction models, the life of the switch power supply of the aging station is predicted to obtain the remaining life of the switch power supply.

[0058] When predicting the life of the switching power supply of the aging station based on N switching power supply life prediction models, the degradation path category can be determined based on the timing signal of the switching power supply to be predicted, and the switching power supply life prediction model matching the category can be called to make an accurate prediction. The specific prediction process may include: determining the characteristic parameters of the timing signal of the switching power supply of the aging station to be predicted; determining the degradation category to which the switching power supply of the aging station to be predicted belongs; inputting the characteristic parameters into the switching power supply life prediction model of the matching category to predict the remaining service life of the switching power supply. The prediction results of this time can also be used as basic data for the accuracy of the subsequent evaluation results of the switch unit prediction model, specifically: calculating the prediction error of the model based on the remaining service life obtained from this prediction and the actual data of the remaining service life of the switching power supply; and evaluating the accuracy of the model prediction results based on the prediction error.

[0059] The embodiment of the present application provides a method for predicting faults of isolated switching power supplies at the front end of a burn-in station, which performs dimensionality reduction processing on preset timing signals of multiple switching power supplies at the burn-in station, and extracts characteristic parameters; calculates the similarity of performance degradation paths of the switching power supplies at each burn-in station based on the characteristic parameters of each preset timing signal; clusters each degradation path based on the similarity of performance degradation paths of the switching power supplies at each burn-in station to obtain N degradation path sets; establishes N switching power supply life prediction models, and trains the N switching power supply life prediction models based on the N degradation path sets respectively; predicts the life of the switching power supply at the burn-in station based on the trained N switching power supply life prediction models to obtain the remaining life of the switching power supply. The method for predicting the fault of the isolated switching power supply at the front end of the aging station, on the one hand, the aging station switching power supply life prediction model based on the MDRNN multidimensional recurrent neural network model comprehensively considers the impact of the performance degradation path on the remaining service life prediction of the switching power supply, and can achieve high-precision life prediction; on the other hand, the failure mode is taken into consideration, the performance degradation path of the switching power supply of the aging station is identified, and the characteristic parameters of the switching power supply of the aging station with the same performance degradation path are brought into the same model for training and prediction, which is beneficial to improve the accuracy of the life prediction of the switching power supply of the aging station.

[0060] Combine the following Figure 2 Taking an example, the fault prediction scheme of the isolated switching power supply at the front end of the aging station provided by the present application is described, which specifically includes the following steps: The method for predicting the fault of the front-end isolated switching power supply of the aging station provided in this specific example uses the DTW method to evaluate the similarity of the paths between multiple front-end isolated switching power supplies of the aging station (also called primary switching power supplies of the aging station) with different decay rates and different fault initiation degrees, and introduces the K-Means performance degradation path clustering method on this basis, which can realize the rapid and accurate classification of the degradation path of the primary switching power supply of the aging test station; the remaining service life prediction model of the primary switching power supply of the aging station based on the MDRNN multi-dimensional recurrent neural network model realizes high-precision life prediction by comprehensively considering the impact of the performance degradation path on the remaining service life prediction of the primary switching power supply of the aging station; when training and predicting the multi-dimensional recurrent neural network model, the fault mode is taken into consideration to identify the performance degradation path of the primary switching power supply of the aging station, and the primary switching power supply of the aging station with the same performance degradation path is brought into the same model for training and prediction, which is beneficial to improve the remaining service life prediction accuracy of the primary switching power supply of the aging station.

[0061] like Figure 2As shown in the principle diagram of the front-end isolated switching power supply fault prediction method of the aging station, it mainly includes the following parts: identification and clustering of the performance degradation path of the primary switching power supply of the aging station, establishment of the MDRNN multidimensional recurrent neural network model, training of the category matching MDRNN multidimensional recurrent neural network model based on the clustered degradation path, and prediction of the remaining life of the primary switching power supply of the aging station based on the trained MDRNN multidimensional recurrent neural network model.

[0062] The clustering part of the primary switching power supply performance degradation path identification of the aging station includes the following steps: Sub-step 1: Dimensionality reduction and feature extraction.

[0063] In this specific example, the current, voltage and other timing signals of the primary switching power supply are reduced in dimension to extract characteristic parameters (also called performance parameters). The extracted characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flux rate, thermal conductivity, thermal diffusivity, etc.

[0064] Sub-step 2: DTW path similarity calculation.

[0065] In this specific example, based on the extracted feature parameters, the advantages of the dynamic time warping (DTW) algorithm are adopted. By scaling the time axis of the unknown sequence to the same length as the template sequence, the similarity between the sample performance degradation paths can be better grasped. The DTW principle is introduced as follows Figure 3 shown.

[0066] For the performance parameter vector of the primary switching power supply sample of the aging station , (a, b are sample numbers, m, n are the corresponding running cycle numbers), the main steps of the DTW method are as follows: 1) Construct a The matrix grid, matrix elements express and Distance between two points , each matrix element express and Alignment.

[0067] 2) Find a regular path through this grid and use To express: 1 3) Path boundary condition constraints, that is , , ensuring that the selected path must start from the common starting point of the sequence and end at the common ending point of the sequence.

[0068] 4) Path continuity constraint, if , then for the next point on the path Needs to be satisfied and ,ensure and Each coordinate in Appeared in.

[0069] 5) Path monotonicity constraint, if , then for the next point on the path Needs to be satisfied and ,limit The above points must be monotonically progressing over time.

[0070] 6) Repeat (2), (3), (4) to obtain the path with the minimum regularization cost. , in the denominator Used to compensate for regular paths of different lengths.

[0071] The DTW method is used to calculate the similarity of characteristic time series trajectories such as ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity, thermal diffusivity, etc. among the primary switching power supply samples of different aging stations. Subsequently, the clustering method can be used to cluster the time series trajectories, so that the primary switching power supplies of the aging stations can be classified according to the performance degradation path.

[0072] Sub-step 3: K-Means degenerate path clustering.

[0073] Taking advantage of the time used for cluster centroid selection and the spatial complexity of cluster overlap in K-Means, the optimal point that actually exists in the data set is used as its centroid, so the trajectory corresponding to the centroid can be found.

[0074] The steps of K-Means algorithm are as follows: 1) Initialize the center point: Randomly select K points as the centers of the initial clusters (Centroids).

[0075] 2) Allocate clusters: For each point in the data set, calculate its distance to each center point (usually using Euclidean distance) and assign it to the cluster with the closest distance.

[0076] 3) Update the center point: For each point in the cluster, calculate its mean and update the center point of the cluster to the mean position. The goal of K-Means is to minimize the square error within the cluster, that is:

[0077] 4) Repeat iteration: Repeat steps 2) and 3) until the cluster center no longer changes significantly or the set number of iterations is reached.

[0078] Using the K-Means method, the system samples are clustered through the performance degradation path similarity matrix of the primary switching power supply of the aging station, so as to realize the identification of the primary switching power supply of the aging station according to the performance degradation path time series (i.e., time series signal). Figure 2 As shown in the figure, the K-Means algorithm is used to cluster multiple switching power supply degradation paths into two categories: Category 1 and Category 2. Two MDRNN multidimensional recurrent neural network models will be constructed later, and the two models will be trained based on the two categories of switching power supply degradation paths. The MDRNN multidimensional recurrent neural network model is referred to as the MDRNN model below.

[0079] The MDRNN model is established, the MDRNN model for category matching is trained based on the degradation path after clustering, and the remaining life of the primary switch power supply of the aging station is predicted based on the trained MDRNN model as follows: The MDRNN multi-dimensional recurrent neural network uses the input of two switching power supply degradation path data, and uses the bidirectional LSTM (BiLSTM layer) and bidirectional GRU (BiGRU layer) layers to extract time series features, then splices the outputs of the two layers, and finally outputs the predicted remaining service life. Momentum and adaptive learning rate are used to speed up the convergence speed, and it has been proven to be superior to the classic stochastic gradient descent optimizer to achieve higher processing efficiency and prediction quality.

[0080] The main framework of the MDRNN model is divided into four parts: input layer, processing layer, fusion layer and output layer. The specific steps are as follows: 1) Establish two life prediction models corresponding to two performance degradation paths. According to the performance degradation path, all power supply components are divided into two categories, and all time series performance parameters of the pre-processed primary switching power supply (ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flux rate, thermal conductivity, thermal diffusivity) are used as model input to construct a training set. , where K is the number of training samples, Represents the input data matrix obtained from the monitoring data time series by time window, with a size of N, where are the sensor measurements of the s selected sensors at time t.

[0081] 2) The operating condition data Connect to In the above example, we construct a high-order vector Then, the vector Feed it into the stacked BiLSTM and BiGRU layers to obtain hidden feature maps of different dimensions.

[0082] 3) Set the number of BiLSTM layers and BiGRU layers to be the same, that is, the total number of stacked BiGRU layers is also M. For the last BiLSTM layer and BiGRU layer, the output of the last BiLSTM layer and BiGRU layer at the last time is used as the final output of the last BiLSTM layer and BiGRU layer.

[0083] 4) The hidden feature vectors are concatenated into a merged feature vector. Finally, the merged feature vector is fed into two more linear regression dense layers to generate the predicted RUL.

[0084] 5) The error between the predicted RUL value and the actual RUL value is used as the loss function to calculate the loss, and the relevant parameters in the network are updated through back propagation. The training is repeated until the loss reaches the requirement.

[0085] For the time series of performance parameters of the primary switching power supply of the aging station , , n is the length of the time series, each Contains m power supply parameters, that is, an m-dimensional vector, in this case m=11. Positional encoding , so that the input of the model is , where the PE calculation formula is as follows.

[0086]

[0087]

[0088] Among them, pos is the position of w in the sequence, and i is the position of the parameter in w.

[0089] The core of the MDRNN model is to use a multi-layer neural network structure, combined with multiple degradation path data, to use the bidirectional LSTM (BiLSTM) and bidirectional GRU (BiGRU) layers to extract long-term dependencies in the time series, and then fuse these features, and finally predict the remaining useful life (RUL) of the equipment through the fully connected layer (Dense layer). The key to the whole process is feature extraction and fusion to achieve accurate prediction of the health status of the aging station.

[0090] The prediction accuracy is used to evaluate the performance of the prediction model. In this patent, the prediction accuracy is defined as follows:

[0091] Among them, RUL is the actual RUL (remaining useful life); is the predicted RUL estimated by the prediction model, and TUL is the actual total service life of the target circuit, which is defined as the total number of test cycles when the primary switching power supply of the aging station reaches the failure threshold.

[0092] The embodiment of the present application provides a method for predicting the fault of the isolated switching power supply at the front end of the aging station. On the one hand, the life prediction model of the switching power supply of the aging station based on the deep neural network comprehensively considers the influence of the performance degradation path on the prediction of the remaining service life of the switching power supply, and can achieve high-precision life prediction; on the other hand, the failure mode is taken into consideration, the performance degradation path of the switching power supply of the aging station is identified, and the characteristic parameters of the switching power supply of the aging station with the same performance degradation path are brought into the same model for training and prediction, which is beneficial to improving the accuracy of the life prediction of the switching power supply of the aging station.

[0093] The present application also provides a method for evaluating the health status of a secondary power supply of a high-temperature aging test device. Figure 1 The aging test equipment shown in the figure is executed after the front-end isolated switching power supply fault prediction process is completed, or it can be executed before it, or it can be executed in parallel with it. The high temperature aging test equipment is the high temperature aging test equipment.

[0094] As attached Figure 4 As shown, the method for evaluating the health status of secondary power supply of high temperature aging test equipment in the embodiment of the present application includes the following steps: Step 401: Collect the secondary power supply monitoring signal of the high temperature aging test equipment to be evaluated.

[0095] The method for evaluating the health status of a secondary power supply of a high-temperature aging test equipment provided in the implementation of the present application can be applied to electronic devices. A computer program for evaluating the health status of a secondary power supply of a high-temperature aging test equipment is provided in the electronic device. When the computer program is executed by a processor, the method for evaluating the health status of a secondary power supply of a high-temperature aging test equipment in the embodiment of the present application is implemented.

[0096] The aging test equipment provided in the present application can perform non-destructive aging testing on the test piece. The secondary power supply health status assessment method made in the embodiment of the present application can assess the health status of the secondary power supply of the aging test equipment. The health assessment of the secondary power supply of the aging test equipment can ensure the safety of the test piece.

[0097] In the actual implementation process, the data volume of the secondary power supply monitoring signal of the high-temperature aging test equipment directly measured is large and contains a lot of redundant information. Therefore, preliminary feature extraction is performed on the directly measured data to obtain a feature sequence with high information density.

[0098] Step 402: Generate a multi-dimensional feature sequence according to the monitoring signal.

[0099] The monitoring signals include but are not limited to current signals and voltage output signals, power supply ripple fluctuation rate, MOSFET shell temperature to junction temperature, etc. The specific features extracted can be flexibly set by technicians in this field according to actual needs to ensure that the extracted features can characterize the state of the secondary power supply. Among them, MOSFET is abbreviated as MOS, which is a metal oxide semiconductor effect transistor and a voltage-controlled semiconductor device. In this embodiment, it refers to the voltage-controlled semiconductor device in the secondary power supply of the high temperature aging test equipment.

[0100] Specifically, the features of the current signal and the voltage output signal are extracted, and then the power supply ripple fluctuation rate and the degradation curve from the case temperature to the junction temperature of the MOSFET device are mixed to obtain a multi-dimensional feature sequence.

[0101] Step 403: input the multi-dimensional feature sequence into the pre-trained residual convolutional neural network model to obtain the degradation characteristic state of the secondary power supply of the high temperature aging test equipment.

[0102] Among them, the degradation characteristic state includes: healthy state and degraded state. If the degradation characteristic state of the secondary power supply of the high temperature aging test equipment is a healthy state, there is no need to execute step 404 and subsequent steps. If the degradation characteristic state of the secondary power supply of the high temperature aging test equipment is a degraded state, it means that the secondary power supply of the high temperature aging test equipment has been degraded. The degradation progress of the secondary power supply of the high temperature aging test equipment can be automatically identified by adaptively analyzing the multi-dimensional feature sequence data through the residual convolutional neural network model.

[0103] In an optional embodiment, the training process of the residual convolutional neural network model may include the following sub-steps: Sub-step 1: Obtaining a sample of the timing degradation characteristics of the secondary power supply training of the high temperature aging test equipment; Each training time series degradation feature sample is marked with a degradation feature state. The training time series degradation feature sample of the secondary power supply of the high temperature aging test equipment can be obtained by monitoring the temperature signal and degradation feature state of the secondary power supply of the high temperature aging test equipment throughout its life cycle.

[0104] In the actual implementation process, the secondary power supply training timing degradation feature sample of the high temperature aging test equipment can be expressed as , the maximum number of iterations max_iter. Mark the degradation characteristic state corresponding to the secondary power supply of the high temperature aging test equipment for training , including healthy and degraded states.

[0105] Sub-step 2: Input the training time series degradation feature samples into the pre-established residual convolutional neural network model to obtain the first probability space point after mapping; The residual convolutional neural network model includes: convolution layer, batch normalization layer, activation function (ReLU), maximum pooling layer, multiple residual blocks, global average pooling layer and fully connected layer.

[0106] In the actual implementation process, For input, Train the ResNet model (i.e., residual convolutional neural network model) for output; obtain iterative learning to update the ResNet model parameters. When the number of iterations reaches the maximum number of iterations max_iter, stop iterative learning.

[0107] In the actual implementation process, the maximum number of iterations max_iter can be flexibly set by those skilled in the art, and no specific limitation is imposed on this in the embodiments of the present application.

[0108] Sub-step 3: Determine the starting point of the secondary power degradation of the high temperature aging station according to the two-dimensional first probability space point, and generate a healthy baseline of the secondary power of the high temperature aging station.

[0109] In an optional embodiment, the training process of the residual convolutional neural network model may also include the following process: Obtain test timing degradation feature samples of the secondary power supply of the high-temperature aging test equipment; input the test timing degradation feature samples into the residual convolutional neural network model in training to obtain the mapped second probability space points; and calculate the distance between the second probability space and the healthy baseline based on a preset multi-distance metric algorithm.

[0110] The timing degradation characteristic sample of the secondary power supply test of high temperature aging test equipment can be expressed as: In the training and testing process of the residual convolutional neural network model, and Input the ResNet algorithm model to obtain the mapped probability space points and .

[0111] The degradation starting point is automatically identified based on the two-dimensional probability space results, and the weighted fusion multi-distance metric algorithm is used to calculate the distance between the healthy space and the degradation space of the secondary power supply of the training and testing high-temperature aging station. . Traverse all fault modes and obtain the health assessment results corresponding to each fault mode. The final trained residual convolutional neural network model outputs and Health assessment curve corresponding to the second power supply of each high-temperature aging station.

[0112] More specifically, based on the preset multi-distance measurement algorithm, the distance between the second probability space and the health baseline can be calculated as follows: the distance between the second probability space and the health baseline is calculated based on the Mahalanobis distance algorithm, the cosine similarity algorithm and the Manhattan distance algorithm respectively to obtain the first distance, the second distance and the third distance; the first distance, the second distance and the third distance are weightedly summed to obtain the corresponding health level.

[0113] In the actual implementation process, it is not limited to using the above three distance algorithms to calculate the distance. Any one of the above three distance algorithms can also be used, or any two of the above three distance algorithms can be used to obtain two distances and then perform weighted summation of the two distances.

[0114] In the training process of the residual convolutional neural network model, the test timing degradation feature samples of the secondary power supply of the high-temperature aging test equipment are introduced, and the prediction accuracy of the residual convolutional neural network model can be tested through the prediction results. In the actual implementation process, when the prediction accuracy of the residual convolutional neural network model reaches the accuracy threshold, it can be determined that the training of this residual convolutional neural network model is completed. After the subsequent training is completed, the residual convolutional neural network model can be used to predict the state degradation curve of the secondary power supply of the high-temperature aging station to be evaluated.

[0115] Step 404: When the secondary power supply of the high temperature aging test equipment is in a degraded state, the state degradation curve of the secondary power supply of the high temperature aging test equipment to be evaluated is determined based on the probability space points after multi-dimensional feature sequence mapping and the health baseline in the residual convolutional neural network model.

[0116] Among them, the state degradation curve can characterize the health status of the secondary power supply of the high temperature aging test equipment.

[0117] By analyzing the multi-dimensional characteristics of the secondary power supply of the high-temperature aging test equipment to be evaluated through the residual convolutional neural network model, the health of the secondary power supply at the current moment can be obtained, and the state degradation curve up to that moment (also called the health degradation curve) can be output, providing reliable data support for further remaining service life prediction research of the secondary power supply of the high-temperature aging test equipment.

[0118] The method for evaluating the health status of the secondary power supply of the high-temperature aging test equipment disclosed in the embodiment of the present invention collects the monitoring signal of the secondary power supply of the high-temperature aging test equipment to be evaluated, generates a multidimensional feature sequence according to the monitoring signal; inputs the multidimensional feature sequence into a pre-trained residual convolutional neural network model to obtain the degradation feature state of the secondary power supply of the high-temperature aging test equipment; when the secondary power supply of the high-temperature aging test equipment is in a degraded state, based on the probability space points mapped by the multidimensional feature sequence and the health baseline in the residual convolutional neural network model, the state degradation curve of the secondary power supply of the high-temperature aging platform to be evaluated is determined, thereby accurately determining the health status of the secondary power supply of the high-temperature aging platform. The health status evaluation scheme for the secondary power supply of the high-temperature aging test equipment provided in the embodiment of the present invention collects the monitoring signal of the secondary power supply and extracts the multidimensional feature sequence, and predicts the multidimensional feature sequence based on the trained model to obtain the state degradation curve of the secondary power supply of the high-temperature aging test equipment to be evaluated, thereby accurately determining the health status of the secondary power supply of the high-temperature aging test equipment.

[0119] Figure 5 A structural block diagram of a device for predicting faults of isolated switching power supplies at the front end of a burn-in station in an embodiment of the present application.

[0120] The aging station front-end isolated switching power supply fault prediction device provided in the embodiment of the present application includes the following functional modules: The feature extraction module 501 is used to perform dimensionality reduction processing on preset timing signals of multiple aging station switching power supplies and extract feature parameters; wherein the aging station switching power supply is a front-end isolated switching power supply of the aging station; A similarity determination module 502 is used to calculate the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals; The clustering module 503 is used to cluster the degradation paths according to the similarity of the performance degradation paths of the switch power supplies of each aging station to obtain N degradation path sets; A model training module 504 is used to establish N switch power supply life prediction models, and train the N switch power supply life prediction models according to the N degradation path sets respectively; The model prediction module 505 is used to predict the life of the switching power supply of the aging station based on the N trained switching power supply life prediction models to obtain the remaining life of the switching power supply.

[0121] Optionally, the characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flow rate, thermal conductivity, thermal diffusivity, etc.

[0122] Optionally, the similarity determination module includes: The first submodule is used to construct parameter vectors for characteristic parameters of each of the preset timing signals; A second submodule is used to construct a matrix grid according to each of the parameter vectors; The third submodule is used to find the regular minimum cost path through the matrix grid through multiple iterative adjustments based on preset path boundary constraints, path continuity constraints and path monotonicity constraints, wherein the regular minimum cost path can characterize the similarity of the performance degradation paths of the switching power supplies of each aging station.

[0123] Optionally, the clustering module includes: A fourth submodule is used to randomly select K points from the regular minimum cost path as the center of the initial cluster; A fifth submodule is used to assign each point in the regular minimum cost path to a cluster closest to the center point; The sixth submodule is used to update the center point of each cluster; A seventh submodule is used to assign each point in the regular minimum cost path to a cluster closest to the updated center point; An eighth submodule is used to determine whether the clustering stop condition is met, and if not, return to execute the sixth submodule; If the ninth submodule has been reached, the clusters of the updated center points are clustered as N degenerate path sets.

[0124] Optionally, the sixth submodule is specifically used for: For each cluster, calculate the average value of each point in the cluster; The center point of the cluster is updated to the location of the average value.

[0125] Optionally, the switching power supply life prediction model includes: an input layer, a Dense layer (i.e., a dense layer), a BiLSTM layer, i.e., a bidirectional long short-term memory network layer, for processing multi-sensor monitoring signals, a BiGRU layer, i.e., a bidirectional gated loop layer, for processing operating condition data, i.e., the transformation result after hidden representation), a Concatenate layer, i.e., a feature concatenation layer, for fusing the outputs of the BiLSTM and BiGRU layers, a multidimensional feature map, and an output layer for outputting a RUL prediction value. Optionally, the model training module is specifically used for: Establish N switching power supply life prediction models; For each of the switching power supply life prediction models, the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set are used as model inputs; Performing a linear transformation on the input feature parameters in the fully connected layer, and performing a hidden representation on the transformation results; Constructing a high-order vector according to the transformation result after the hidden representation, and sending the high-order vector to the BLSTM layer and the BGRU layer to obtain hidden feature maps of different dimensions; The feature vectors generated by merging the hidden features of different dimensions are input into two linear regression dense layers to generate RUL prediction values; The error between the RUL predicted value and the RUL actual value is used as a loss function to calculate the loss; Updating the parameters of the switching power supply life prediction model according to the loss back propagation; Determine whether the switching power supply life prediction model after updating the parameters meets the preset accuracy; If satisfied, determining that the training of the switching power supply life prediction model is completed; If not satisfied, return to execute the operation of using the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set as model input.

[0126] The embodiments of the present application provide Figure 5 The isolated switching power supply fault prediction device at the front end of the aging station shown can achieve Figure 1 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0127] The embodiment of the present application provides a front-end isolated switching power supply fault prediction device for a aging station. On the one hand, the aging station switching power supply life prediction model based on a deep neural network comprehensively considers the impact of the performance degradation path on the remaining service life prediction of the switching power supply, and can achieve high-precision life prediction; on the other hand, the failure mode is taken into consideration, the performance degradation path of the aging station switching power supply is identified, and the characteristic parameters of the aging station switching power supply with the same performance degradation path are brought into the same model for training and prediction, which is beneficial to improving the accuracy of the aging station switching power supply life prediction.

[0128] An embodiment of the present invention further provides an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus.

[0129] Memory, used to store computer programs; The processor is used to implement the method for predicting the fault of the isolated switching power supply at the front end of the aging station considering multiple degradation paths as shown in the above method embodiment when executing the program stored in the memory.

[0130] The communication bus mentioned in the above terminal may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc.

[0131] The communication interface is used for communication between the above terminal and other devices.

[0132] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0133] In another embodiment provided by the present invention, a computer-readable storage medium is also provided, in which instructions are stored. When the computer-readable storage medium is executed on an electronic device, the electronic device implements the method for predicting the fault of the isolated switching power supply at the front end of the aging station considering multiple degradation paths as described in any of the above embodiments.

[0134] In another embodiment provided by the present invention, a computer program product containing instructions is also provided. When the computer program product is run on an electronic device, the electronic device implements the method for predicting faults of the isolated switching power supply at the front end of the aging station considering multiple degradation paths as described in any of the above embodiments.

[0135] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0136] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting faults of isolated switching power supplies at the front end of a burn-in station, characterized in that: The method comprises: Perform dimension reduction processing on the preset timing signals of multiple aging station switching power supplies and extract characteristic parameters; Calculating the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals; According to the similarity of the performance degradation paths of the switching power supplies of each aging station, each degradation path is clustered to obtain N degradation path sets; Establishing N switch power supply life prediction models, and training the N switch power supply life prediction models according to the N degradation path sets respectively; Based on the trained N switching power supply life prediction models, the life of the switching power supply on the aging station is predicted to obtain the remaining life of the switching power supply.

2. The method according to claim 1, characterized in that The characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flux rate, thermal conductivity and thermal diffusivity.

3. The method according to claim 1, characterized in that: The step of calculating the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals comprises: Construct parameter vectors for the characteristic parameters of each of the preset timing signals respectively; constructing a matrix grid according to each of the parameter vectors; Based on preset path boundary constraints, path continuity constraints and path monotonicity constraints, a regular minimum cost path passing through the matrix grid is found through multiple iterative adjustments, wherein the regular minimum cost path can characterize the similarity of the performance degradation paths of the switching power supplies of each aging station.

4. The method according to claim 1, characterized in that: The step of clustering the degradation paths according to the similarity of the performance degradation paths of the switch power supplies of each aging station to obtain a set of N degradation paths includes: Randomly select K points from the regular minimum cost path as the center of the initial cluster; For each point in the regular minimum cost path, assign each point to the cluster closest to the center point; Update the center point of each cluster; For each point in the regular minimum cost path, assign each point to a cluster closest to the updated center point; Determine whether the clustering stop condition is met, and if not, return to execute the step of updating the center point of each cluster; If it has been reached, the clusters of the updated center points are clustered as the N degenerate path sets.

5. The method according to claim 4, characterized in that The step of updating the center point of each cluster comprises: For each cluster, calculate the average value of each point in the cluster; The center point of the cluster is updated to the location of the average value.

6. The method according to claim 1, characterized in that The switching power supply life prediction model includes: an input layer, a dense layer, a bidirectional long short-term memory network layer, a bidirectional gated loop layer, a feature concatenation layer, a multidimensional feature map and an output layer.

7. The method according to claim 6, characterized in that The steps of establishing N switching power supply life prediction models and training the N switching power supply life prediction models according to the N degradation path sets respectively include: Establish N switching power supply life prediction models; For each of the switching power supply life prediction models, the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set are used as model inputs; Performing a linear transformation on the input feature parameters in the fully connected layer, and performing a hidden representation on the transformation results; Constructing a high-order vector according to the transformation result after the hidden representation, and sending the high-order vector to the bidirectional long short-term memory network layer and the bidirectional gated recurrent layer to obtain hidden feature maps of different dimensions; The feature vectors generated by merging hidden features of different dimensions are input into two dense layers to generate RUL prediction values; Determine whether the switching power supply life prediction model after updating the parameters meets the preset accuracy; If satisfied, determining that the training of the switching power supply life prediction model is completed; If not satisfied, return to the step of using the characteristic parameters corresponding to the degradation paths in the corresponding degradation path set as model input.

8. A device for predicting faults of isolated switching power supplies at the front end of a burn-in station, characterized in that: The device comprises: A feature extraction module is used to perform dimensionality reduction processing on preset timing signals of multiple aging station switching power supplies and extract feature parameters; A similarity determination module, used to calculate the similarity of the performance degradation paths of the switch power supplies of each aging station according to the characteristic parameters of each of the preset timing signals; A clustering module, used to cluster the degradation paths according to the similarity of the performance degradation paths of the switching power supplies of each aging station to obtain N degradation path sets; A model training module, used to establish N switch power supply life prediction models, and train the N switch power supply life prediction models according to the N degradation path sets respectively; The model prediction module is used to predict the life of the switching power supply of the aging station based on the N trained switching power supply life prediction models to obtain the remaining life of the switching power supply.

9. The device according to claim 8, characterized in that The characteristic parameters include: ripple voltage, average output voltage, inductor current peak-to-peak value, inductor current peak value, capacitor current peak-to-peak value, capacitor current peak value, circuit board temperature, heat flux density, heat flux rate, thermal conductivity and thermal diffusivity.

10. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor is used to implement the method for predicting the fault of the isolated switching power supply at the front end of the aging station as described in any one of claims 1 to 7 when executing the program stored in the memory.

Images