SiC MOSFET life prediction method based on WOA-BP

Abstract

The invention discloses a WOA-BP-based SiC MOSFET life prediction method and device and a storage medium, and relates to the technical field of reliability prediction of power electronic equipment. The method comprises the following steps: selecting drain-source electrode voltage as an aging characteristic parameter of the SiC MOSFET, and acquiring voltage data in real time in an aging test process by using a sensor; the collected voltage data are preprocessed, abnormal values are eliminated, missing values are supplemented, and the data are smoothed; parameters of a back propagation neural network BP are optimized through a whale optimization algorithm WOA, and a WOA-BP prediction model is constructed; and initializing a WOA-BP model and inputting voltage data for training, and establishing a mapping relation between the service life of the SiC MOSFET and the voltage to realize prediction of the residual service life of the SiC MOSFET. The SiC MOSFET service life prediction model has the advantages of being high in adaptability, low in calculation complexity, high in prediction precision and the like, can be widely applied to the fields of the semiconductor industry, new energy automobiles, industrial automation and the like, and provides powerful support for health management and maintenance of equipment.

Description

Technical Field

[0001] The present application relates to the technical field of reliability prediction of power electronic equipment, and in particular to a SiC MOSFET life prediction method, device and storage medium based on WOA-BP. Background Art

[0002] With the continuous progress in the field of power electronics technology, SiC MOSFET, as a core component in power electronic equipment, plays a vital role in the performance of the entire system. However, the aging of SiC MOSFET is a complex process, which is affected by multiple factors such as electrical, thermal, and mechanical stress. Traditional life prediction methods often find it difficult to accurately predict its remaining service life. Therefore, developing a SiC MOSFET life prediction method based on WOA-BP is of great significance for improving the reliability of power electronic equipment and reducing maintenance costs. At present, some life prediction methods based on machine learning algorithms have been proposed, such as neural networks and support vector machines. These methods have improved the accuracy and efficiency of prediction to a certain extent, but they still have certain limitations in some aspects, such as the prediction accuracy needs to be further improved, and the generalization ability needs to be enhanced. At the same time, due to the complex and changeable aging process of SiC MOSFET, traditional prediction methods often find it difficult to adapt to the life prediction needs under different operating conditions.

[0003] It can be seen that when traditional data-driven models are used to predict the life of SiC MOSFET, there is often a strong dependence on data quality and it is easy to fall into the local optimal solution, resulting in low prediction accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a SiC MOSFET lifetime prediction method, device and storage medium based on WOA-BP to address the above technical problems.

[0005] This manual adopts the following technical solutions:

[0006] This specification provides a SiC MOSFET lifetime prediction method based on WOA-BP, including:

[0007] The drain-source voltage is selected as the aging characteristic parameter of SiC MOSFET, and the voltage data is obtained in real time by using a sensor during the aging test;

[0008] Preprocess the collected voltage data to eliminate outliers, supplement missing values, and smooth the data; optimize the parameters of the back propagation neural network BP through the whale optimization algorithm WOA to build a WOA-BP prediction model;

[0009] Initialize the WOA-BP model and input voltage data for training, establish the mapping relationship between SiC MOSFET life and voltage, and realize the prediction of the remaining life of SiC MOSFET.

[0010] Preferably, the sensor is used to obtain voltage data in real time during the aging test, specifically including: based on the accelerated aging test of SiC MOSFET, according to the working principle and failure mechanism of SiC MOSFET, key voltage parameters closely related to its life are selected and collected through voltage sensors.

[0011] Preferably, the preprocessing of the collected voltage data specifically includes: cleaning the collected raw voltage data, using statistical methods to identify and eliminate outliers in the data, using filters to remove high-frequency noise, and filling by interpolation, or deleting records containing missing values ​​according to specific circumstances.

[0012] Preferably, the optimization of the parameters of the back propagation neural network BP by the whale optimization algorithm WOA specifically includes:

[0013] The optimal parameter configuration searched by the WOA algorithm is applied to the BP model to build a WOA-BP prediction model. These parameters include learning rate, number of hidden units, number of iterations, etc.

[0014] Preferably, the initializing the WOA-BP model specifically includes:

[0015] The parameters of the initialization BP include the number of neurons, weights, biases, the number of layers of the BP model, and the selection of the internal optimizer;

[0016] The initialization WOA algorithm parameters include the size of the whale population, the location of the whale population and the number of iterations of the algorithm.

[0017] Preferably, the input voltage parameter is trained, specifically including:

[0018] The preprocessed voltage data is divided into a training set and a validation set. The constructed WOA-BP prediction model is trained using the training set data. According to the performance indicators on the validation set, the parameters of the WOA algorithm are continuously adjusted to improve the optimization process.

[0019] Preferably, the prediction of the remaining life of the SiC MOSFET is achieved by:

[0020] In each iteration, the WOA algorithm is used to optimize the parameters of the BP network. In each iteration, the WOA algorithm calculates the fitness value based on the position of each whale in the current population (i.e., the parameter combination of the BP model), and updates the position of the whale based on the fitness value to find a better BP model parameter combination. Repeat the above steps until the preset number of iterations is reached or the convergence condition is met, and the result of the remaining life of the SiC MOSFET is obtained.

[0021] The present specification provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned SiC MOSFET life prediction method based on WOA-BP is implemented.

[0022] This specification provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned SiCMOSFET life prediction method based on WOA-BP is implemented.

[0023] A SiC MOSFET life prediction method based on WOA-BP provided in this specification has the following beneficial effects compared with the prior art:

[0024] The present invention combines the advantages of the WOA algorithm and the BP network, optimizes the BP network parameters through the WOA algorithm, and constructs a WOA-BP prediction model, which can effectively search for the optimal solution or a parameter combination close to the optimal solution in a complex parameter space, and find a better solution in a shorter time, thereby improving the efficiency of model training and prediction. At the same time, it can be widely used in the semiconductor industry, new energy vehicles, industrial automation and other fields, providing strong support for the health management and maintenance of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0026] Figure 1 A schematic diagram of a SiC MOSFET life prediction method provided in this specification;

[0027] Figure 2 It is a simple circuit platform design for the DC power cycle test in this embodiment;

[0028] Figure 3 The corresponding SiC MOSFET drain-source voltage is collected when the constant current is 15A and the temperature is 150°C in this embodiment;

[0029] Figure 4 is the drain-source voltage of the SiC MOSFET after data processing in this embodiment;

[0030] Figure 5 This is the BP network topology in this embodiment;

[0031] Figure 6 Graph showing the life prediction result of SiC MOSFET based on the traditional BP model in this embodiment;

[0032] Figure 7 This is a graph showing the life prediction results of SiC MOSFET based on the WOA-BP model in this embodiment. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this application will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in the specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0034] SiC MOSFET lifetime prediction based on traditional data-driven models usually trains and verifies models on specific data sets, which results in limited generalization capabilities of the models to other data sets or operating conditions. When the operating environment or conditions of the SiC MOSFET change, these models may not provide accurate prediction results.

[0035] Therefore, this specification provides a SiC MOSFET life prediction method based on WOA-BP, which can effectively search for the optimal solution or a parameter combination close to the optimal solution in a complex parameter space, and find a better solution in a shorter time, thereby improving the efficiency of model training and prediction, and solving the problems of strong dependence on data quality when predicting SiC MOSFET life under traditional data-driven models, and the traditional data-driven models are prone to fall into local optimal solutions, resulting in low prediction accuracy.

[0036] In a SiC MOSFET life prediction method based on WOA-BP proposed in this specification, a WOA-BP prediction model is constructed to predict the life of SiC MOSFET. This model is an optimization and prediction method that combines the Whale Optimization Algorithm (WOA) and the Back Propagation Neural Network (BPNN). The Whale Optimization Algorithm is a new type of heuristic optimization algorithm. It simulates the social behavior of humpback whales and introduces bubble nets for hunting. As a gregarious mammal, humpback whales will cooperate with each other to hunt prey. Whales have two behaviors in group hunting: encirclement and expulsion. Whales in the group move toward other whales to encircle prey, and swim in a circle and spray bubbles to form a bubble net to drive prey. This unique hunting method is called bubble net foraging. The WOA algorithm is based on this hunting behavior. It simulates the search behavior of whales to find the optimal solution and has the ability of global search and local search. BP is a multi-layer feedforward neural network, whose main features are the forward propagation of signals and the back propagation of errors. During the training process, the input signal passes through the input layer, hidden layer and output layer to generate the output signal. Then, the error between the output signal and the target signal is calculated and the error is back-propagated to the hidden layer to reduce the error by adjusting the weights and thresholds of the hidden layer.

[0037] Normally, the traditional BP network is often used as a life prediction method for IGBTs. However, the training and inference process of this method may be slow in large-scale data sets or complex models, requiring more computing resources. The high complexity may also cause the model to be more prone to overfitting problems during training, and it is highly dependent on data quality. In addition, the training process may fall into a local optimal solution, resulting in an inability to provide accurate prediction results.

[0038] Compared with traditional data-driven prediction models such as BP, the WOA-BP model provided in this manual combines the advantages of the whale optimization algorithm (WOA) and the back propagation neural network (BP). WOA optimizes the key parameters of the BP network to improve the prediction accuracy and generalization ability of the BP model. WOA simulates the hunting behavior of whale groups, finds the optimal parameter combination through iterative optimization, and then these optimized parameters are used to construct the BP network, so that BP can process time series data more effectively and capture long-term dependencies in the data, thereby achieving more accurate predictions.

[0039] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0040] Figure 1The following is a flow chart of a SiC MOSFET lifetime prediction method based on WOA-BP in this specification, which specifically includes the following steps:

[0041] S101: Selecting drain-source voltage as the aging characteristic parameter of SiC MOSFET, and using a sensor to obtain voltage data in real time during the aging test; specifically including:

[0042] Based on the accelerated aging test of SiC MOSFET, according to the working principle and failure mechanism of SiC MOSFET, key voltage parameters closely related to its life are selected and collected through voltage sensors.

[0043] S102: Preprocessing the collected voltage data to eliminate abnormal values, supplement missing values, and smooth the data, specifically including:

[0044] The collected raw voltage data is cleaned, and outliers in the data are identified and eliminated using statistical methods. High-frequency noise is removed using filters, and interpolation is used to fill in the data, or records with missing values ​​are deleted according to specific circumstances.

[0045] S103: Optimizing the parameters of the back propagation neural network BP through the whale optimization algorithm WOA, and constructing a WOA-BP prediction model, specifically including:

[0046] The optimal parameter configuration searched by the WOA algorithm is applied to the BP model to build a WOA-BP prediction model. These parameters include learning rate, number of hidden units, number of iterations, etc.

[0047] S104: Initialize the WOA-BP model and input voltage data for training, establish a mapping relationship between SiC MOSFET life and voltage, and predict the remaining life of SiC MOSFET, including:

[0048] Initialize BP parameters, including the number of neurons, weights, biases, the number of layers of the BP model, and the selection of the internal optimizer;

[0049] Initialize WOA algorithm parameters, including whale population size, whale population location, and number of algorithm iterations.

[0050] Input voltage parameters for training, including dividing the preprocessed voltage data into a training set and a validation set, using the training set data to train the constructed WOA-BP prediction model, and continuously adjusting the parameters of the WOA algorithm according to the performance indicators on the validation set to improve the optimization process.

[0051] In each iteration, the WOA algorithm is used to optimize the parameters of the BP network. In each iteration, the WOA algorithm calculates the fitness value based on the position of each whale in the current population (i.e., the parameter combination of the BP model), and updates the position of the whale based on the fitness value to find a better BP model parameter combination. Repeat the above steps until the preset number of iterations is reached or the convergence condition is met, and the result of the remaining life of the SiC MOSFET is obtained.

[0052] It should be noted that the SiC MOSFET accelerated aging test is a test method that simulates the power cycle conditions of the SiC MOSFET in actual work to accelerate its aging process. The experiment continuously applies power to the SiC MOSFET to make it switch periodically between the on and off states, thereby using the thermal stress generated inside the device to impact the SiC MOSFET and accelerate its aging failure. The power cycle aging experiment can effectively evaluate the reliability and life of the SiC MOSFET under long-term use, and provide a basis for product design and optimization. The simple circuit platform design of the power cycle test is as follows: Figure 2 shown.

[0053] Example

[0054] The SiC MOSFET life prediction method proposed in this embodiment adds the WOA optimization algorithm to the traditional SiC MOSFET life prediction method. Under this premise, the above scheme uses the position of each whale to calculate the fitness, and updates the position of the whale according to the fitness value to find the optimal solution to deal with the problem that the data quality is highly dependent on the life prediction of SiC MOSFET under the traditional data-driven model, and it is easy to fall into the local optimal solution; during the prediction process, the WOA-BP algorithm will continuously update the parameters of the BP network to adapt to the changes in the input data and improve the accuracy of the prediction.

[0055] The flow chart of SiC MOSFET lifetime prediction method based on WOA-BP model is as follows: Figure 1 As shown, this embodiment specifically includes the following steps:

[0056] Step 1: Select the drain-source voltage as the aging characteristic parameter of SiC MOSFET, and use the sensor to obtain voltage data in real time during the aging test.

[0057] The aging test process of SiC MOSFET in this embodiment specifically includes:

[0058] The device supplies power to the aging board or test board, and applies a periodically changing load current to the SiC MOSFET to simulate the junction temperature changes in actual use. In the test, the device is repeatedly heated and cooled by controlling the opening and closing of the current to accelerate its aging process, and the changes in the drain-source voltage are monitored in real time to evaluate the packaging reliability and aging characteristics of the device.

[0059] In addition, the key parameters of SiC MOSFET collected are the corresponding SiC MOSFET drain-source voltage when the constant current is 15A and the temperature is 150℃, such as Figure 3 shown.

[0060] Step 2: Preprocess the collected voltage data to eliminate outliers, supplement missing values, and smooth the data, including:

[0061] The collected voltage outliers are eliminated. This step aims to identify and eliminate values ​​that are significantly deviated from the normal data range. These outliers may be caused by measurement errors, equipment failures, or external interference. A reasonable threshold range can be set, and data points outside this range are regarded as outliers and eliminated. Alternatively, statistical methods such as box plots can be used to automatically identify and eliminate outliers.

[0062] To supplement the missing values ​​that may exist in the data, the missing data will affect the accuracy and reliability of subsequent analysis. The method to deal with missing values ​​is to directly delete the records containing missing values ​​(if the missing ratio is not high) and use interpolation methods (such as linear interpolation, polynomial interpolation, etc.) to estimate the missing values ​​based on adjacent data points.

[0063] Smoothing the data helps reduce random noise in the data and make the data change trend clearer. The mean filtering method is commonly used. The basic principle of mean filtering is to replace the pixel values ​​in the original image with the mean, that is, for the current pixel point (x, y) to be processed, select a template composed of several pixels of its neighbors, calculate the mean of all pixels in the template, and then assign the mean to the current pixel point (x, y) as the grayscale g(x, y) of the processed image at this point. Its expression is:

[0064]

[0065] In the above expression, m is the total number of pixels in the template including the current pixel. After data processing, the drain-source voltage of the SiC MOSFET is as follows: Figure 4 shown.

[0066] Step 3: Optimize the parameters of the back propagation neural network BP through the whale optimization algorithm WOA and build a WOA-BP prediction model.

[0067] It should be noted that the advantage of combining WOA and BP is that the global search capability and high efficiency of the whale optimization algorithm WOA can make up for the shortcomings of the BP network in parameter optimization; and the complex structure and powerful learning ability of the BP network can make full use of the optimization results of the whale optimization algorithm WOA to further improve the performance of the model.

[0068] The specific steps of optimizing the parameters of the BP network through the WOA algorithm and constructing the WOA-BP prediction model are as follows: applying the optimal parameter configuration searched by the WOA algorithm to the BP model to construct the WOA-BP prediction model. These parameters include learning rate, number of hidden units, number of iterations, etc.

[0069] It should be noted that the BP neural network is a three-layer neural network, including an input layer, a hidden layer, and an output layer. The transformation from the input space to the hidden layer space is nonlinear, while the transformation from the hidden layer space to the output layer space is linear. Its topological structure is as follows: Figure 5 shown.

[0070] In this embodiment, the S-shaped activation function is expressed as

[0071]

[0072] Where x is the input of the neuron and σ(x) is the output of the neuron.

[0073] It should be noted that WOA simulates the strategy of whales in hunting and predation, and has global search capabilities and good stability and convergence accuracy. The algorithm mainly includes three parts: surrounding prey, bubble net predation, and searching for prey.

[0074] In the stage of surrounding prey, the size of the whale population is set to N, and the position of the i-th whale in space is d represents the spatial dimension of the search. The optimal position of the prey in the group corresponds to the global optimal solution. Other whales will gradually move closer to this position and surround it, using equations (3) and (4) to update the position:

[0075] D=|CX * (t)-X(t) (3)

[0076] X(t+1)=X * (t)-AD (4)

[0077] Where t is the current number of iterations; X * (t) is the current prey position; X(t) is the current individual position; D is the distance between the whale and the prey; A and C are position coefficients, defined as follows:

[0078] A=2αr1-α (5)

[0079] C=2r2 (6)

[0080] Where r1 and r2 are random values ​​in [0,1]; α is the convergence factor, which can be defined by the following expression:

[0081] α=2-2t / T (7)

[0082] Where T is the maximum number of iterations.

[0083] In the bubble net predation stage, by calculating the position (X, Y) to reach the position (X * ,Y * ), and uses the spiral position update method to update the position distance between the whale and the prey, simulating the way the whale spirally attacks the prey. The expression is as follows:

[0084]

[0085] Where D t is the distance between the whale and its prey; b is a constant that limits the spiral; l is a random number between [-1,1]; during the continuous optimization process, in order to simulate the whale approaching its prey along the path surrounded by the spiral, the update probability p is usually set to 0.5.

[0086] In the prey search stage, in order to improve the global search ability of the algorithm and avoid falling into the local optimum, that is, when A≥1, the algorithm randomly searches for the location of the prey. The expression of this stage is as follows:

[0087]

[0088] In the formula, is the position of a randomly selected whale in the whale group.

[0089] Finally, by initializing the WOA-BP model and inputting voltage data for training, the mapping relationship between SiC MOSFET lifetime and voltage is established, and the remaining lifetime of SiC MOSFET can be predicted.

[0090] It should be noted that the parameters of BP are initialized, including the number of neurons, weights, biases, the number of layers of the BP model, and the selection of the internal optimizer; the parameters of the WOA algorithm are initialized, including the size of the whale population, the location of the whale population, and the number of iterations of the algorithm.

[0091] It should be noted that the preprocessed voltage data is divided into a training set and a validation set. The constructed WOA-BP prediction model is trained using the training set data. According to the performance indicators on the validation set, the parameters of the WOA algorithm are continuously adjusted to improve the optimization process.

[0092] In each iteration, the WOA algorithm is used to optimize the parameters of the BP network. In each iteration, the WOA algorithm calculates the fitness value based on the position of each whale in the current population (i.e., the parameter combination of the BP model), and updates the position of the whale based on the fitness value to find a better BP model parameter combination. Repeat the above steps until the preset number of iterations is reached or the convergence condition is met, and the result of the remaining life of the SiC MOSFET is obtained. The WOA-BP prediction results are compared with the traditional BP prediction results, such as Figure 6 and Figure 7 shown.

[0093] according to Figure 6 and Figure 7 From the prediction results, it can be seen that the WOA-BP prediction method of this embodiment effectively improves the prediction accuracy in SiC MOSFET life prediction compared with the traditional BP prediction method.

[0094] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A SiC MOSFET lifetime prediction method based on WOA-BP, characterized in that: include: The drain-source voltage is selected as the aging characteristic parameter of SiC MOSFET, and the voltage data is obtained in real time by using a sensor during the aging test; Preprocess the collected voltage data to eliminate outliers, supplement missing values, and smooth the data; The parameters of the back propagation neural network BP are optimized through the whale optimization algorithm WOA, and the WOA-BP prediction model is constructed; Initialize the WOA-BP model and input voltage data for training, establish the mapping relationship between SiC MOSFET life and voltage, and realize the prediction of the remaining life of SiC MOSFET.

2. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The method of using a sensor to obtain voltage data in real time during the aging test specifically includes: Based on the accelerated aging test of SiC MOSFET, according to the working principle and failure mechanism of SiC MOSFET, key voltage parameters closely related to its life are selected and collected through voltage sensors.

3. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The preprocessing of the collected voltage data specifically includes: The collected raw voltage data is cleaned, and outliers in the data are identified and eliminated using statistical methods. High-frequency noise is removed using filters, and interpolation is used to fill in the data, or records with missing values ​​are deleted according to specific circumstances.

4. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The Whale Optimization Algorithm (WOA) is used to optimize the parameters of the back propagation neural network BP, specifically including: The optimal parameter configuration searched by the WOA algorithm is applied to the BP model to build a WOA-BP prediction model. These parameters include learning rate, number of hidden units, number of iterations, etc.

5. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The initialization of the WOA-BP model specifically includes: The parameters of the initialization BP include the number of neurons, weights, biases, the number of layers of the BP model, and the selection of the internal optimizer; The initialization WOA algorithm parameters include the size of the whale population, the location of the whale population and the number of iterations of the algorithm.

6. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The input voltage parameters are trained, specifically including: The preprocessed voltage data is divided into a training set and a validation set. The constructed WOA-BP prediction model is trained using the training set data. According to the performance indicators on the validation set, the parameters of the WOA algorithm are continuously adjusted to improve the optimization process.

7. The SiC MOSFET lifetime prediction method based on WOA-BP according to claim 1, characterized in that: The prediction of the remaining life of WOA-BP specifically includes: In each iteration, the WOA algorithm is used to optimize the parameters of the BP network. In each iteration, the WOA algorithm calculates the fitness value based on the position of each whale in the current population (i.e., the parameter combination of the BP model), and updates the position of the whale based on the fitness value to find a better BP model parameter combination. Repeat the above steps until the preset number of iterations is reached or the convergence condition is met, and the result of the remaining life of the SiC MOSFET is obtained.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in any one of claims 1 to 7 is implemented.

Images