IGBT (Insulated Gate Bipolar Translator) device aging state evaluation method, device and equipment based on neural network and medium

Abstract

The invention discloses an IGBT device aging state evaluation method, device and equipment based on a neural network and a medium, and belongs to the technical field of IGBT device detection.The method comprises the steps that training parameters of an IGBT device in a healthy state are obtained, and the training parameters comprise conduction voltage drops and collector currents of the IGBT device at different junction temperatures; performing model training on a preset PyTorch neural network model by adopting the training parameters to obtain a state evaluation model, the preset PyTorch neural network model being a fully connected neural network model; and calling the state evaluation model to evaluate the aging state of the IGBT device to be identified. According to the method, model training is carried out through different parameters of the IGBT device so as to fit the actual scene of the IGBT device, the evaluation error is reduced, the evaluation precision of the model is greatly improved, and the existing detection requirements are met.

Claims

1. A method for evaluating the aging status of an IGBT device based on a neural network, characterized in that: The method comprises: Acquire training parameters of an IGBT device in a healthy state, wherein the training parameters include a forward voltage drop and a collector current of the IGBT device at different junction temperatures; Performing model training on a preset PyTorch neural network model using the training parameters to obtain a state assessment model, wherein the preset PyTorch neural network model is a fully connected neural network model; The state assessment model is called to assess the aging state of the IGBT device to be identified.

2. The neural network-based IGBT device aging status assessment method according to claim 1, characterized in that: The obtaining of training parameters of the IGBT device in a healthy state includes: Determining a plurality of parameter curves of an IGBT device in a healthy state, each parameter curve being a characteristic curve of current and voltage of the IGBT device, and each parameter curve corresponding to a junction temperature; Taking points from the parameter curve according to a preset voltage interval to obtain interval parameters; Each value within the interval parameter is adjusted based on a preset adjustment ratio value, and the adjusted values ​​are manually labeled to obtain training parameters.

3. The neural network-based IGBT device aging status assessment method according to claim 1, characterized in that: The training parameters are used to perform model training on a preset PyTorch neural network model to obtain a state assessment model, including: Standardizing the on-state voltage drop, collector current, and junction temperature within the training parameters to obtain standardized parameters; After dividing the standardized parameters into a training set and a validation set, the training set is used to perform model training on a preset PyTorch neural network model, and the validation set is used to perform model validation to obtain a state assessment model.

4. The neural network-based IGBT device aging status assessment method according to claim 3, characterized in that: The preset PyTorch neural network model includes: an input layer, several fully connected layers, a hidden layer, and an output layer; The input layer is connected to the plurality of fully connected layers, the plurality of fully connected layers are connected to the hidden layer, and the hidden layer is connected to the output layer; The hidden layer includes three layers, consisting of 64 neurons, 128 neurons, and 64 neurons respectively; The output layer is a single neuron, and the mean square error loss function is used to calculate the prediction error; The mean square error loss function is shown as follows: In the above formula, n is the number of samples, y i is the true value, is the predicted value.

5. The neural network-based IGBT device aging status assessment method according to claim 3, characterized in that: After the step of obtaining the standardized parameters, the method further comprises: Calling the preset Adam optimizer to optimize the normalization parameters; Among them, the preset Adam optimizer is shown as follows: m t =β1m t-1 +(1-β1)g t In the above formula, α is the learning rate, β1 is the first-order moment decay rate, θ t-1 is the weight of the previous round, θ t is the updated weight.

6. The neural network-based IGBT device aging status assessment method according to claim 3, characterized in that: After the step of performing model validation using the validation set, the method further includes: Obtaining a validation set loss rate, where the validation set loss rate is calculated by using the validation set to perform model validation within a preset round; When the value of the validation set loss rate does not meet the validation requirements, the preset learning rate scheduler is called to adjust the learning rate.

7. The neural network-based IGBT device aging status assessment method according to any one of claims 1 to 6, characterized in that: The calling of the state assessment model to assess the aging state of the IGBT device to be identified includes: Inputting the input parameters of the IGBT device to be identified into the state assessment model to obtain an assessment prediction value; The aging state of the IGBT device to be identified is determined according to the magnitude of the evaluation prediction value.

8. A neural network-based IGBT device aging status assessment device, characterized in that: The device comprises: An acquisition module is used to acquire training parameters of an IGBT device in a healthy state, wherein the training parameters include a forward voltage drop and a collector current of the IGBT device at different junction temperatures; A training module, configured to perform model training on a preset PyTorch neural network model using the training parameters to obtain a state assessment model, wherein the preset PyTorch neural network model is a fully connected neural network model; The evaluation module is used to call the state evaluation model to evaluate the aging state of the IGBT device to be identified.

9. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the neural network-based IGBT device aging status assessment method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer-executable program, and the computer-executable program is used to enable a computer to execute the neural network-based IGBT device aging status assessment method according to any one of claims 1 to 7.

Images