Method and equipment for constructing component degradation prediction model and component degradation prediction

Abstract

The invention discloses a method and equipment for constructing a component degradation prediction model and component degradation prediction, and relates to the field of component simulation, and the method comprises the steps: obtaining an original degradation data sequence of a sample feature component when the sample feature component is subjected to a long-term power-up test under a target constant stress level, n sequence values in the original degradation data sequence are performance parameter values of the sample feature component at N time points respectively; sequentially deleting the sequence values in the original degradation data sequence according to the sequence of the time points, and taking the remaining part of the original degradation data sequence as a new degradation data sequence after each deletion to obtain N degradation data sequences; a corresponding feature sequence group is constructed according to each degradation data sequence, the feature sequence group comprises a plurality of feature sequences representing different degradation influence factors, and a sequence value in each feature sequence is used for representing the influence degree of the degradation influence factor at the corresponding time point; and training the target deep learning model by taking the feature sequence group as a sample feature and taking the degradation data sequence as a sample label to obtain a component degradation prediction model.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of component simulation, in particular to a method and device for constructing a component degradation prediction model and predicting component degradation. BACKGROUND

[0002] Electrical and electronic products are usually composed of various components, and as integrated products, their function implementation highly depends on the reliability of underlying components. In combination with the structure, function and task profile characteristics of electrical and electronic products, high-quality reliability modeling of components therein is a key link for effectively predicting the reliability of electrical and electronic products and supporting availability analysis and maintenance strategy development. Due to the influence of factors such as multiple stages of working conditions and fluctuations in working environment, various components inside electrical and electronic products work in a complex dynamic stress environment for a long time, and thus the degradation rate and reliability of components under such conditions are directly related to the dynamic stress spectrum corresponding to the environment, and in addition, the performance degradation process of various components has universal nonlinear characteristics. Therefore, the reliability model constructed for components of electrical and electronic products should not only be able to accurately predict the degradation and reliability of components under constant stress environment, but also be able to accurately predict the degradation and reliability of components under dynamic stress environment.

[0003] Current technologies can better realize the reliability prediction of components under constant stress level, but the accuracy of reliability prediction of components under dynamic stress level is low. SUMMARY

[0004] The purpose of the present application is to provide a method and device for constructing a component degradation prediction model and predicting component degradation, which can more accurately predict the reliability of components and solve the problem of low accuracy of current technologies in predicting the reliability of components under dynamic stress level.

[0005] To achieve the above purpose, the present application provides the following solutions: In a first aspect, the present application provides a method for constructing a component degradation prediction model, comprising: obtaining an original degradation data sequence of a sample characteristic component when the sample characteristic component is subjected to long-term power-on test under a target constant stress level, wherein N sequence values in the original degradation data sequence are performance parameter values of the sample characteristic component at N time points; sequentially deleting sequence values in the original degradation data sequence in the order of the time points and taking the remaining part of the original degradation data sequence as a new degradation data sequence after each deletion, thereby obtaining N degradation data sequences; construct a corresponding feature sequence group according to each of the degradation data sequences, the feature sequence group comprising a plurality of feature sequences respectively representing different degradation influencing factors, and a sequence value in each feature sequence being used to represent an influence degree of a degradation influencing factor at a corresponding time point; train a target deep learning model by taking the feature sequence group as sample features and the degradation data sequences as sample labels to obtain a component degradation prediction model.

[0006] In a second aspect, the present application provides a component degradation prediction method, comprising: a component degradation prediction model obtained by the method for constructing a component degradation prediction model according to the first aspect is used to predict the degradation of a target component In a third aspect, the present application provides a device for constructing a component degradation prediction model, comprising: a data acquisition module configured to acquire an original degradation data sequence of a sample feature component when the sample feature component is subjected to a long-term power-on test under a target constant stress level, wherein N sequence values in the original degradation data sequence are performance parameter values of the sample feature component at N time points; a data processing module configured to sequentially delete sequence values in the original degradation data sequence in the order of the time points and take the remaining part of the original degradation data sequence as a new degradation data sequence after each deletion, thereby obtaining N degradation data sequences; a data construction module configured to construct a corresponding feature sequence group according to each of the degradation data sequences, the feature sequence group comprising a plurality of feature sequences respectively representing different degradation influencing factors, and a sequence value in each feature sequence being used to represent an influence degree of a degradation influencing factor at a corresponding time point; a model training module configured to train a target deep learning model by taking the feature sequence group as sample features and the degradation data sequences as sample labels to obtain a component degradation prediction model.

[0007] In a fourth aspect, the present application provides a computer device, comprising a memory, a processor, a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the method for constructing a component degradation prediction model according to any one of the above aspects.

[0008] In a fifth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the steps of the method for constructing a component degradation prediction model according to any one of the above aspects.

[0009] In a sixth aspect, the present application provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the method for constructing a degradation prediction model of a component according to any one of the above.

[0010] According to the specific embodiments provided in the present application, the present application discloses the following technical effects: The present application provides a method and device for constructing a degradation prediction model of a component and predicting degradation of a component, by allowing a deep learning model to learn the mapping relationship between degradation influencing factors and device stability, the deep learning model can predict the degradation of a component to be tested under both constant stress level and dynamic stress level, thereby accurately predicting the reliability of the component, and solving the problem of low accuracy of current technology in predicting the reliability of the component under dynamic stress level. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0012] Figure 1 An application environment diagram of a method for constructing a degradation prediction model of a component according to an embodiment of the present application; Figure 2 A flowchart of a method for constructing a degradation prediction model of a component according to an embodiment of the present application; Figure 3 A constant stress condition diagram; Figure 4 A schematic diagram of nonlinear performance degradation and degradation rate of a component; Figure 5 A degradation trend diagram of a component under different temperature stresses; Figure 6 A degradation rate diagram of a component at different degradation increments; Figure 7 A diagram showing that individual differences of batch products affect degradation trend; Figure 8 A diagram showing parameters that need to be focused on for degradation prediction of a single stress stage; Figure 9 A diagram showing construction of a training set and a validation set based on test data; Figure 10 A diagram showing input-output relationship of an LSTM model; Figure 11Performance degradation data graph for metal film resistor RJ25-316kΩ at 80 degrees Celsius; Figure 12 Performance degradation data graph for metal film resistor RJ25-316kΩ at 100 degrees Celsius; Figure 13 Performance degradation data graph for metal film resistor RJ25-316kΩ at 120 degrees Celsius; Figure 14 Performance degradation data graph for metal film resistor RJ25-316kΩ at 140 degrees Celsius; Figure 15 Instance graph of prediction results of LSTM model for training set; Figure 16 Instance graph of prediction results of LSTM model for test set; Figure 17 Lognormal distribution probability graph of life distribution prediction results of metal film resistor RJ25-316kΩ; Figure 18 Schematic graph of life distribution prediction results of metal film resistor RJ25-316kΩ; Figure 19 Degradation curve graph of step stress accelerated test of metal film resistor RJ25-316kΩ; Figure 20 Step temperature curve graph of metal film resistor RJ25-316kΩ; Figure 21 Dynamic stress degradation prediction results (sample 1) of metal film resistor RJ25-316kΩ; Figure 22 Dynamic stress degradation prediction results (sample 2) of metal film resistor RJ25-316kΩ. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0014] The above purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be further described in detail below with reference to the drawings and specific embodiments.

[0015] The method for constructing a degradation prediction model of a component provided in the embodiments of the present application can be applied to, for example, Figure 1The terminal 101 communicates with the server 102 through a network in the application environment shown. The data storage system can store data required to be processed by the server 102. The data storage system can be separately arranged, integrated on the server 102, placed on the cloud or other servers. The terminal 101 can send N degraded data sequences and corresponding feature sequence groups to the server 102. The server 102 receives N degraded data sequences and corresponding feature sequence groups. The server 102 trains a target deep learning model by taking the feature sequence group as sample features and the degraded data sequence as sample labels to obtain a component degradation prediction model. The server 102 can feed back the obtained component degradation prediction model to the terminal 101. In addition, in some embodiments, the method of constructing the component degradation prediction model can also be implemented by the server 102 or the terminal 101 alone.

[0016] The terminal 101 can be, but is not limited to, various desktop computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 102 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0017] In an exemplary embodiment, as shown in Figure 2 , a method for constructing a component degradation prediction model is provided. The method is executed by a computer device, specifically by a terminal or a server, or by both a terminal and a server. In the embodiments of the present application, the method is applied to the server 102 in Figure 1 , which includes the following steps 210 to 240.

[0018] Step 210: Obtain the original degraded data sequence of the sample feature component when the sample feature component is subjected to long-term power-on test under a target constant stress level. The N sequence values in the original degraded data sequence are the performance parameter values of the sample feature component at N time points.

[0019] Wherein, the sample feature component is any given component, and the target constant stress level is any given constant stress level. The long-term power-on test under the constant stress level is a mature experiment at present, and its experimental process is not described herein. Generally, the prior art is to directly perform mathematical statistics on the degradation data sequence generated by the long-term power-on test of the component under a specific constant stress level, thereby establishing a degradation model of the component under the specific constant stress level, but this method can only predict the degradation of the specific component under the specific constant stress level, and does not have generalization. The key of the present application is to specially process the original degradation data sequence in the subsequent steps to construct the model training sample feature, thereby training a deep learning model which can more accurately predict the degradation of different components under dynamic stress levels.

[0020] For the degradation data sequence, it can be represented as X(t), X is a performance parameter value, t is a time variable, and there are N time points in total.

[0021] In step 220, the sequence values in the original degradation data sequence are sequentially deleted in the order of time points, and the remaining part of the original degradation data sequence is taken as a new degradation data sequence after each deletion, and N degradation data sequences are obtained in total.

[0022] In step 230, a corresponding feature sequence group is constructed according to each degradation data sequence, and the feature sequence group includes a plurality of feature sequences respectively representing different degradation influencing factors. The sequence values in each feature sequence are respectively used to represent the influence degree of the degradation influencing factor at the corresponding time point.

[0023] In step 240, the feature sequence group is taken as the sample feature, and the degradation data sequence is taken as the sample label to train the target deep learning model to obtain a component degradation prediction model.

[0024] Firstly, in step 220, a plurality of new degradation data sequences are obtained based on the original degradation data sequence. For example, assuming that the original degradation data sequence is {X(0), X(1), X(2), X(3), X(4)}, then the new degradation data sequences are {X(1), X(2), X(3), X(4)}, {X(2), X(3), X(4)}, {X(3), X(4)} and {X(4)} in turn.

[0025] Next, in step S230, a corresponding feature sequence group is constructed for each degradation data sequence. The feature sequence group corresponds to the degradation data sequence and can characterize different degradation influencing factors. By using the feature sequence group as sample features and the degradation data sequence as sample labels to train the target deep learning model, the target neural network model can predict the degradation data sequence based on different degradation influencing factors of the component under test, thereby obtaining the performance parameter values ​​of the component under test at different time points.

[0026] It should be noted that different degradation factors are the essential factors affecting the stability of components. By having the deep learning model learn the mapping relationship between degradation factors and component stability, the deep learning model can predict the degradation of the component under test under both constant stress level and dynamic stress level. This allows for a more accurate prediction of component reliability, solving the problem of low accuracy in predicting component reliability under dynamic stress level in current technologies.

[0027] In this embodiment, different degradation influencing factors include initial values ​​of performance parameters, cumulative degradation increment, stress application time, and stress level; multiple feature sequences are the initial value sequence of performance parameters, the cumulative degradation increment sequence, the stress application time sequence, and the stress level sequence.

[0028] The reasons for selecting the initial value of performance parameters, cumulative degradation increment, stress application time, and stress level as degradation influencing factors in this embodiment are explained below.

[0029] for Figure 3 The linear degradation performance parameters shown indicate that, when the stress level is fixed, the degradation rate of the component over its entire life cycle can be considered a constant (independent of the degree of degradation already occurred). That is, the key factors affecting the life cycle degradation rate of such components under dynamic stress levels are mainly the stress level, stress magnitude, and individual differences. However, for... Figure 4 Regarding the nonlinear degradation performance parameters of the components shown, which are widely present during long-term power-on, the degradation rate exhibits a time-varying state throughout their entire lifespan. Therefore, further analysis is needed to determine the key influencing factors that determine the dynamic changes in this degradation rate.

[0030] Assuming the same sample feature exhibits the following degradation trends under high and low temperature stress levels: Figure 6 As shown in Figure 7, the relationship between the degradation rate and the current degradation increment is illustrated in the curve. It can be seen that, in addition to the generally understood effect of temperature stress on the dynamic degradation rate of components throughout their entire life cycle, the accumulated degradation increment (i.e., the degree of degradation) at each moment of a nonlinear degradation component is also a key factor affecting its corresponding degradation rate and subsequent degradation process.

[0031] In addition, as shown in Figure 7 Due to factors such as product design tolerance, manufacturing process, etc., the same batch of components also have individual differences, which further leads to different degradation trajectories of the batch components. Therefore, individual differences also need to be considered as a key factor affecting the dynamic degradation rate of the component life cycle.

[0032] Through the above analysis, when modeling the degradation of the component based on the multi-constant stress level degradation data, the key factors affecting the dynamic degradation rate of the component life cycle are determined as stress level (representing the influence of stress), cumulative degradation increment (representing the influence of the degradation stage), and initial value of performance parameter (representing the influence of individual difference), which are used as the input feature parameter source for training the deep learning model. At the same time, considering the direct influence of stress action time on the cumulative degree of degradation, the input feature parameters of the deep learning model are set to the initial value of the performance parameter, the cumulative degradation increment, the stress action time, and the stress level, which are four dimensions in total.

[0033] As can be seen from the above description, the initial value of the performance parameter, the cumulative degradation increment, the stress action time, and the stress level are the key factors affecting the degradation of the component.

[0034] For the four-dimensional input feature parameters of the deep learning model determined in this embodiment, i.e., the initial value of the performance parameter, the cumulative degradation increment, the stress action time, and the stress level. Since the cumulative degradation increment of the component at any time will jointly determine its current degradation rate and subsequent degradation trend with the initial value of the performance parameter, the subsequent stress level, and the stress action time. Therefore, as shown in Figure 8 , only the degradation process of the component under a single stress stage (stress level S) under dynamic stress level (cumulative degradation increment ΔX at the beginning of the stage, stress action time Δt) is predicted, the essence is to find the time point corresponding to the cumulative degradation increment ΔX on the degradation curve of the component corresponding to the stress level S, and then cut the curve of the corresponding length Δt from this time point according to the stress action duration, which is the required stage degradation process prediction result.

[0035] Therefore, since the degradation data of the component in the constant stress long-term power-on test is from the initial state, the original degradation data sequence is split to construct the training set and validation set sample features that can quantitatively reflect the influence of the cumulative degradation increment ΔX. That is, step 220 is performed, and the execution result of step 220 can be understood with reference to Figure 9 Figure 9 In n ​For the training sample label 1, it is the original degradation data sequence, containing the performance parameter values of all time points; for the training sample label 2, it is obtained by removing the performance parameter value of time point t0 from the training sample label 1; for the training sample label 3, it is obtained by removing the performance parameter values of time point t0 and time point t1 from the training sample label 1; and so on, the training sample label n is obtained by removing the performance parameter values of all time points before time point t n from the training sample label 1.

[0036] And, for any one degradation data sequence A: each sequence value in the initial value sequence of the performance parameter of the degradation data sequence A is the target performance parameter value, and the target performance parameter is the initial performance parameter value of the sample characteristic component; each sequence value in the cumulative degradation increment sequence of the degradation data sequence A is the difference between the first sequence value of the degradation data sequence A and the target performance parameter; each sequence value in the stress action time sequence of the degradation data sequence A is the time difference between each time point in the degradation data sequence A and the first time point in the degradation data sequence A; and each sequence value in the stress level sequence of the degradation data sequence A is the target constant stress level.

[0037] In combination with Figure 9 , the specific composition of the data sample refers to Table 1.

[0038] Table 1: Data sample table

[0039] In the table, ΔX(t) = X(t) - X(0). t t

[0040] In constructing the degradation prediction model of the component, the data sequence obtained from multiple test pieces and multiple constant stress long-term power-on test data is taken as the benchmark according to the data sample table shown in Table 1 to construct the corresponding training set and validation set.

[0041] In terms of the selection and setting of the deep learning model and parameters, since the modeling of the degradation of the component in the embodiment belongs to a typical time series modeling and prediction problem. At the same time, the types of modeling objects are more, and the performance degradation data generally has nonlinear characteristics. Therefore, in view of the superiority of the long short-term memory network model LSTM (including BiLSTM) in time series data processing, capturing long-term dependencies, nonlinear fitting capability, computational complexity and generalization ability compared with RNN, CNN, Transformer and other deep learning models, the long short-term memory network model is selected as the target deep learning model. The long short-term memory network model includes a long short-term memory layer, an input full connection layer, a linear rectification layer, an output full connection layer and a regression layer stacked in turn. The determined input-output relationship of the target deep learning model is as follows:​​Figure 10

[0042] Based on the above target deep learning model, combined with the failure threshold of the component, the failure distribution and reliability prediction results can be given.

[0043] The construction process and performance of the component degradation prediction model are verified by an example as follows.

[0044] Component reliability verification under constant stress conditions.

[0045] Take the metal film resistor RJ25-316kΩ as an example, which was subjected to accelerated testing under the action of four constant temperature stresses. Some performance degradation data obtained are shown in Figures 11 to 14 Based on the performance degradation data, a total of 6519 model training and verification samples were constructed. The model training process is as follows: 1. Input data: sequence data in the shape of [f_, T], where f_ is the number of features at each time point (including the initial value of the performance parameter, the cumulative degradation increment, the stress action time, and the stress level); T is the time step (including each time point, the sequence length).

[0046] 2. LSTM layer: contains 64 hidden units, outputs the complete sequence (OutputMode:'sequence'), and the output shape is [64, T].

[0047] 3. Fully connected layer: maps the LSTM output from 64 dimensions to 32 dimensions, uses the He weight initialization method, and the output shape is [32, T].

[0048] 4. ReLU activation layer: applies a rectified linear unit activation function, and the output shape remains unchanged [32, T].

[0049] 5. Output fully connected layer: maps the features from 32 dimensions to outdim dimensions, and the output shape is [outdim, T].

[0050] 6. Regression layer: calculates the mean square error between the predicted value and the true value for model training and optimization.

[0051] According to the statistics, the error of the model corresponding to the training set and the test set is shown in Table 2, and the comparison of the predicted value and the true value of the training set and the test set is shown in Figure 15 and Figure 16

[0052] Table 2 Model error summary

[0053] ​​As can be seen, the model constructed in this embodiment can effectively and accurately predict the degradation process of the training set and the test set, with average relative error percentages (MAPE) of 0.0121% and 0.0111%, respectively, which are much smaller than the industry standard of 10%.

[0054] Based on the resistance degradation prediction model of the RJ25-316kΩ metal film resistor obtained from the above training, using 25℃ as the input stress condition and a given failure threshold, its lifetime distribution and reliability are predicted. The results are as follows: Figure 17 and Figure 18 As shown.

[0055] It can be seen that the median life of this resistor is 1890 days (approximately 5.17 years), and the B10 life is 1216 days (approximately 3.33 years).

[0056] Reliability verification of components under dynamic stress conditions.

[0057] Accelerated stress testing is a typical dynamic stress condition. Taking a metal film resistor RJ25-316kΩ as an example, such as... Figure 19 As shown, its degradation trend during the step stress test exhibits obvious stage-specific characteristics, with temperatures at each stage being 80℃, 90℃, 100℃, 110℃, 120℃, 130℃, 140℃, and 150℃, respectively. Therefore, this project uses... Figure 19 The accelerated stress test performance degradation dataset shown is used as a benchmark to verify the accuracy of the dynamic stress degradation prediction model built based on the constant stress accelerated test data of this metal film resistor, and the predicted values ​​are expressed as... Figure 19 The relative average error of the measured values ​​shown is ≤30%, which is used as an indicator to measure the effectiveness of the dynamic stress degradation prediction model.

[0058] For example Figure 20 The figure corresponds to Figure 19 The temperature curves of each stage of the accelerated stress degradation test are used as inputs for the "stress duration" and "stress level" shown in Table 1. Based on the initial resistance values ​​of each accelerated stress metal film resistor test sample, the corresponding model input characteristic quantity "initial value of performance parameter" is constructed as the input sequence. The "cumulative degradation increment" at the beginning of each stage is given by the model prediction result of the previous stage.

[0059] Based on the aforementioned setup, degradation prediction was conducted for the RJ25-316kΩ metal film resistor under dynamic stress conditions during accelerated step stress testing. Some predicted results are as follows: Figure 21 As shown, the comparison curve between the predicted and measured values ​​is as follows: Figure 22 As shown.

[0060] The error of the predicted results and the measured results of the dynamic stress degradation process of the metal film resistor RJ25-316kΩ step acceleration test is shown in Table 3.

[0061] Table 3 Error summary of model prediction results

[0062] It can be seen that the average relative error percentage MAPE of the prediction of the dynamic stress degradation process of the metal film resistor RJ25-316kΩ step acceleration test is 0.0150%, which is much smaller than the industry standard of 10%, which can prove the effectiveness of the migration application of the model and the related reliability modeling method to the dynamic stress condition.

[0063] From the above description, it can be seen that the method for constructing the degradation prediction model of the component in the embodiment can construct a model that can simultaneously predict the reliability of the component under constant stress conditions and dynamic stress conditions.

[0064] Therefore, in one exemplary embodiment, a method for degradation prediction of a component is also provided, which comprises: performing degradation prediction on a target component by the component degradation prediction model obtained by the method for constructing the component degradation prediction model in the embodiment.

[0065] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the method for constructing the component degradation prediction model as described above. The implementation scheme of the device for solving the problem is similar to the implementation scheme described in the above method, and therefore the specific limitations in one or more device embodiments for constructing the component degradation prediction model provided below can refer to the limitations of the method for constructing the component degradation prediction model described above, which will not be described here again.

[0066] In one exemplary embodiment, a device for constructing a component degradation prediction model is provided, comprising: a data acquisition module configured to acquire an original degradation data sequence of a sample characteristic component when the sample characteristic component is subjected to a long-term power-on test under a target constant stress level, wherein N sequence values in the original degradation data sequence are performance parameter values of the sample characteristic component at N time points; a data processing module configured to sequentially delete the sequence values in the original degradation data sequence in the order of the time points and take the remaining part of the original degradation data sequence as a new degradation data sequence after each deletion, thereby obtaining N degradation data sequences; a data construction module configured to construct a corresponding feature sequence group according to each degradation data sequence, wherein the feature sequence group comprises a plurality of feature sequences respectively representing different degradation influencing factors, and the sequence values in each feature sequence are respectively used to represent the influence degree of the degradation influencing factor at the corresponding time point; a model training module configured to train a target deep learning model by taking the feature sequence group as sample features and the degradation data sequence as sample labels to obtain the component degradation prediction model.

[0067] In an exemplary embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0068] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program implements the steps in the above method embodiments when executed by a processor.

[0069] In an exemplary embodiment, a computer program product is provided, including a computer program, and the computer program implements the steps in the above method embodiments when executed by a processor.

[0070] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0071] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to a memory, a database or other medium used in the embodiments provided in the present application can include at least one of a non-volatile and a volatile memory. The non-volatile memory can include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical storage, a high-density embedded non-volatile memory, a resistive random access memory (ReRAM), a magnetoresistive random access memory (MRAM), a ferroelectric random access memory (FRAM), a phase change memory (PCM), a graphene memory, etc. The volatile memory can include a random access memory (RAM) or an external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc.

[0072] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0073] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0074] The principles and implementation modes of the present application are described by applying specific examples herein. The above-mentioned embodiments are only used to help understand the method and its core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range can be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A method for constructing a component degradation prediction model, characterized in that, include: The original degradation data sequence of the sample characteristic component is obtained when it is subjected to long-term power-on test under the target constant stress level. The N sequence values ​​in the original degradation data sequence are the performance parameter values ​​of the sample characteristic component at N time points. The sequence values ​​in the original degraded data sequence are deleted sequentially according to the time points mentioned above, and the remaining part of the original degraded data sequence is used as a new degraded data sequence after each deletion, resulting in a total of N degraded data sequences. Based on each of the degradation data sequences, a corresponding feature sequence group is constructed. The feature sequence group includes multiple feature sequences that represent different degradation influencing factors. The sequence value in each feature sequence is used to represent the degree of influence of the degradation influencing factor at the corresponding time point. The component degradation prediction model is obtained by training the target deep learning model with the feature sequence group as sample features and the degradation data sequence as sample labels.

2. The method for constructing a component degradation prediction model according to claim 1, characterized in that, The different degradation influencing factors include initial values ​​of performance parameters, cumulative degradation increment, stress application time, and stress level; The multiple characteristic sequences are the initial value sequence of performance parameters, the cumulative degradation increment sequence, the stress application time sequence, and the stress level sequence.

3. The method for constructing a component degradation prediction model according to claim 2, characterized in that, For any one of the degraded data sequences A: Each sequence value in the initial performance parameter value sequence of the degraded data sequence A is a target performance parameter value, and the target performance parameter is the initial performance parameter value of the sample feature component; Each sequence value in the cumulative degradation increment sequence of the degradation data sequence A is the difference between the first sequence value of the degradation data sequence A and the target performance parameter; The stress action time series values ​​in the degraded data sequence A are the time differences between each time point in the degraded data sequence A and the first time point in the degraded data sequence A. Each sequence value in the stress level sequence of the degradation data sequence A is the target constant stress level.

4. The method for constructing a component degradation prediction model according to claim 1, characterized in that, The target deep learning model is a long short-term memory network model.

5. The method for constructing a component degradation prediction model according to claim 4, characterized in that, The Long Short-Term Memory (LSTM) network model comprises a LSM layer, an input fully connected layer, a linear rectifier layer, an output fully connected layer, and a regression layer stacked sequentially.

6. A method for predicting component degradation, characterized in that, include: The component degradation prediction model obtained by the method of constructing a component degradation prediction model according to any one of claims 1-5 is used to predict the degradation of the target component.

7. An apparatus for constructing a component degradation prediction model, characterized in that, include: The data acquisition module is used to acquire the original degradation data sequence of the sample characteristic components during long-term power-on tests under a target constant stress level. The N sequence values ​​in the original degradation data sequence are the performance parameter values ​​of the sample characteristic components at N time points. The data processing module is used to sequentially delete the sequence values ​​in the original degraded data sequence according to the time points, and after each deletion, take the remaining part of the original degraded data sequence as a new degraded data sequence, so as to obtain a total of N degraded data sequences. The data construction module is used to construct corresponding feature sequence groups according to each of the degradation data sequences. The feature sequence group includes multiple feature sequences that represent different degradation influencing factors. The sequence value in each feature sequence is used to represent the degree of influence of the degradation influencing factor at the corresponding time point. The model training module is used to train the target deep learning model using the feature sequence group as sample features and the degradation data sequence as sample labels to obtain the component degradation prediction model.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for constructing a component degradation prediction model according to any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for constructing a component degradation prediction model as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for constructing a component degradation prediction model as described in any one of claims 1-5.

Images