MOSFET fault diagnosis method and system based on driving current data

Abstract

The invention discloses an MOSFET (Metal-Oxide-Semiconductor Field Effect Transistor) fault diagnosis method and system based on driving current data, and relates to the technical field of device fault diagnosis, and the method comprises the steps: firstly obtaining the driving current data when an MOSFET power driving circuit works; a gray wolf optimization algorithm is adopted in advance to decompose a driving current signal through a variational mode decomposition algorithm after the decomposition layer number and penalty factor optimization, a plurality of intrinsic mode components are obtained, then the energy of each intrinsic mode component is calculated, and a normalized energy value of driving current data is obtained through calculation by selecting a number of intrinsic mode components with the maximum energy. And finally, outputting the judged fault type of the MOSFET power driving circuit by using a pre-trained fault diagnosis model. According to the technical scheme of the invention, the fault and performance degradation mode of the MOSFET power driving circuit can be judged and identified efficiently and accurately.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis of electric devices, and in particular to a method and system for diagnosing MOSFET faults based on drive current data. Background Art

[0002] MOSFET (Metal-Oxide-Semiconductor Field-Effect Transistor, MOSFET) is widely used in various high-power electronic drive systems due to its good characteristics such as high switching speed, small input current, low on-resistance, and strong overcurrent capability, such as motor H-bridge drive circuit, dual voltage time-sharing drive circuit of fuel system, etc. In the above application scenarios, the operating frequency of electronic components is extremely high and the working environment is harsh (voltage and current are large and sudden). There is a possibility of failure and performance degradation under long-term operation. Failure and performance degradation will affect the performance and safety of motors and fuel systems. Therefore, it is necessary to effectively diagnose MOSFET failures and performance degradation. Summary of the invention

[0003] The purpose of this application is to provide a MOSFET fault diagnosis method and system based on drive current data, which can efficiently and accurately judge and identify the fault and performance degradation mode of the MOSFET power drive circuit.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a MOSFET fault diagnosis method based on drive current data, comprising the following steps:

[0006] Get the driving current data when the MOSFET power driving circuit is working.

[0007] The driving current signal is decomposed by using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after the decomposition layers and the penalty factor are optimized by using a grey wolf optimization algorithm.

[0008] The energy of each of the eigenmode components is calculated, and several eigenmode components with the largest energy are selected to calculate and obtain the normalized energy value of the driving current data.

[0009] The normalized energy value is input into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

[0010] Optionally, before using the optimized variational mode decomposition algorithm to decompose the drive current signal to obtain a plurality of intrinsic mode components, the following steps are also included:

[0011] The gray wolf optimization algorithm is used to optimize the decomposition layer number and penalty factor of the variational mode decomposition algorithm to obtain an optimized variational mode decomposition algorithm; specifically, the following steps are included:

[0012] Initialize a gray wolf population and a coefficient vector; the gray wolf population includes a number of gray wolves, each gray wolf corresponds to a set of decomposition layers and penalty factors.

[0013] Calculate the fitness value of each gray wolf, and take the three gray wolves with the highest fitness values ​​as target gray wolves.

[0014] For any gray wolf in the gray wolf population except the target gray wolf, the distance of the gray wolf relative to the three target gray wolves is calculated.

[0015] According to the distance of the gray wolf relative to the three target gray wolves and the coefficient vector, the gray wolf is controlled to move in the direction of the three target gray wolves.

[0016] The coefficient vector is updated, and the fitness value of each gray wolf is calculated.

[0017] Determine whether the maximum number of iterations has been reached.

[0018] If so, the decomposition level and penalty factor of the gray wolf with the highest fitness value are taken as the optimal decomposition level and optimal penalty factor.

[0019] If not, the three gray wolves with the highest fitness values ​​are taken as target gray wolves, and the process jumps to step "for any gray wolf in the gray wolf population except the target gray wolf, calculating the distance of the gray wolf relative to the three target gray wolves".

[0020] Optionally, the fitness value of each gray wolf is calculated according to the following formula:

[0021]

[0022] Where a(j) is the envelope signal obtained after Hilbert transform of the original driving current data, p j is the result of normalizing the envelope signal, N is the length of the time series of the original driving current data, E p is the envelope entropy, i.e., the fitness value.

[0023] Optionally, the distance of the gray wolf relative to the three target gray wolves is calculated according to the following formula:

[0024]

[0025] Among them, α, β, and δ represent the labels of the three target gray wolves, and are the distances of the gray wolf relative to the target gray wolf α, β, δ, respectively. is the coefficient vector used in this round of optimization, and They are the locations of the three target gray wolves. For the location of the Gray Wolf.

[0026] Optionally, the position of the gray wolf after moving towards the three target gray wolves is calculated according to the following formula:

[0027]

[0028] in, is the direction and distance that the gray wolf steps toward the target gray wolf α, is the direction and distance that the gray wolf steps toward the target gray wolf β, is the direction and distance that the gray wolf steps toward the target gray wolf δ, and are the coefficient vectors used in this round of optimization, Indicates the position of the gray wolf after it moves toward the three target gray wolves.

[0029] Optionally, the coefficient vector is updated according to the following formula:

[0030]

[0031] in, and are the coefficient vectors used in the Grey Wolf Optimization Algorithm. is the control vector, It decreases linearly from 2 to 0. and is a random vector in [0,1].

[0032] Optionally, before inputting the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit, the following steps are also included:

[0033] The hard fault of the MOSFET power drive circuit is simulated through experiments and simulations, and the drive current data under the hard fault is obtained.

[0034] The aging process of components is accelerated by using a high and low temperature test chamber, the soft fault of the MOSFET power drive circuit is simulated, and the drive current data under the soft fault is obtained.

[0035] The optimized variational mode decomposition algorithm is used to decompose each of the driving current signals to obtain a plurality of intrinsic mode components.

[0036] The energy of each of the eigenmode components is calculated, and several eigenmode components with the largest energy are selected to calculate and obtain the normalized energy value of the driving current data.

[0037] A fault diagnosis data set is constructed according to the normalized energy value of each driving current data and the fault type corresponding to each driving current data.

[0038] The fault diagnosis data set is used to train a convolutional neural network with the normalized energy value of the driving current data as input and the fault type corresponding to the driving current data as target output to obtain a pre-trained fault diagnosis model.

[0039] Optionally, the convolutional neural network includes an input layer, two pairs of alternating and corresponding convolutional layers, a pooling layer, a fully connected layer and an output layer.

[0040] Optionally, the energy of the eigenmode components is calculated according to the following formula:

[0041]

[0042] Among them, E j is the energy of the jth eigenmode component, u j (t) is the jth eigenmode component.

[0043] In a second aspect, the present application provides a MOSFET fault diagnosis system based on drive current data, comprising the following modules:

[0044] A drive current data acquisition module is used to acquire drive current data when the MOSFET power drive circuit is working;

[0045] A driving current data decomposition module is used to decompose the driving current signal using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after optimizing the decomposition layer number and penalty factor using the grey wolf optimization algorithm;

[0046] A normalized energy value calculation module, used to calculate the energy of each of the intrinsic mode components, and select a number of eigenmode components with the largest energy to calculate the normalized energy value of the driving current data;

[0047] The fault type diagnosis module is used to input the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

[0048] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0049] The present application provides a MOSFET fault diagnosis method and system based on driving current data. In the method, the driving current data of the MOSFET power driving circuit is first obtained when it is working, and the driving current signal is decomposed by using the variational mode decomposition algorithm after the decomposition layer number and the penalty factor are optimized by the gray wolf optimization algorithm to obtain a number of intrinsic mode components, and then the energy of each intrinsic mode component is calculated, and the several intrinsic mode components with the largest energy are selected to calculate the normalized energy value of the driving current data, and finally the fault type of the MOSFET power driving circuit is outputted using a pre-trained fault diagnosis model. The above scheme of the present application uses the variational mode decomposition algorithm after the decomposition layer number and the penalty factor are optimized by the gray wolf optimization algorithm to decompose the driving current data to obtain the intrinsic mode components, which can avoid the misjudgment and missed judgment of human experience, and calculate the normalized energy value of the intrinsic mode component as the characteristic vector of the neural network to diagnose and distinguish the fault type, avoiding directly inputting the intrinsic mode component into the neural network, greatly increasing the complexity of the calculation and the training time, resulting in the problem of dimensional disaster and overfitting, and realizing efficient and accurate judgment and identification of the fault and performance degradation mode of the MOSFET power driving circuit. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flowchart of a MOSFET fault diagnosis method based on drive current data provided in one embodiment of the present application.

[0052] Figure 2 This is a flowchart of step A1 in a MOSFET fault diagnosis method based on drive current data provided in one embodiment of the present application.

[0053] Figure 3 A flowchart of constructing a data set and training a convolutional neural network in a MOSFET fault diagnosis method based on drive current data provided in one embodiment of the present application.

[0054] Figure 4 A schematic diagram of a dual voltage time-sharing drive circuit applied to a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0055] Figure 5A schematic diagram of current waveforms of a dual voltage time-sharing drive circuit used in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0056] Figure 6 A schematic diagram of a current waveform after a high-end bootstrap suspension drive fault occurs in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0057] Figure 7 A schematic diagram of two fault forms of MOSFET1 and the current waveform after the fault in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0058] Figure 8 A schematic diagram of two fault forms of MOSFET2 and the current waveform after the fault in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0059] Fig. 9 A schematic diagram of two fault forms of MOSFET3 and current waveforms after the fault in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0060] Fig.10 A schematic diagram of an accelerated aging test platform used in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0061] Fig.11 A schematic diagram of data measurement point distribution of a dual voltage time-sharing drive circuit in a MOSFET fault diagnosis method based on drive current data provided in another embodiment of the present application.

[0062] Fig.12 A schematic diagram of functional modules of a MOSFET fault diagnosis system based on drive current data provided in one embodiment of the present application.

[0063] Fig.13 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0066] In an exemplary embodiment, Figure 1 As shown, a MOSFET fault diagnosis method based on drive current data is provided, comprising the following steps:

[0067] S1. Obtain driving current data when the MOSFET power driving circuit is working.

[0068] S2. Decompose the driving current signal using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after optimizing the decomposition layer number and penalty factor using a grey wolf optimization algorithm.

[0069] S3. Calculate the energy of each of the eigenmode components, and select several eigenmode components with the largest energy to calculate and obtain the normalized energy value of the driving current data.

[0070] S4. Input the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

[0071] Specifically in this embodiment, before the drive current signal is decomposed by using the optimized variational mode decomposition algorithm in step S2 to obtain a plurality of intrinsic mode components, the MOSFET fault diagnosis method based on drive current data further includes the following steps:

[0072] A1. The gray wolf optimization algorithm is used to optimize the number of decomposition layers and penalty factors of the variational mode decomposition algorithm to obtain the optimized variational mode decomposition algorithm; Figure 2 As shown in the flowchart, step A1 in this embodiment specifically includes the following steps:

[0073] A11. Initialize a gray wolf population and a coefficient vector; the gray wolf population includes a plurality of gray wolves, and each gray wolf corresponds to a set of decomposition layers and a penalty factor.

[0074] A12. Calculate the fitness value of each gray wolf, and use the three gray wolves with the highest fitness values ​​as target gray wolves. In this embodiment, the fitness value of each gray wolf is calculated according to the following formula:

[0075]

[0076] Where a(j) is the envelope signal obtained after Hilbert transform of the original driving current data, p jis the result of normalizing the envelope signal, N is the length of the time series of the original driving current data, E p is the envelope entropy, i.e., the fitness value.

[0077] A13. For any gray wolf in the gray wolf population except the target gray wolf, calculate the distance of the gray wolf relative to the three target gray wolves. In this embodiment, the distance of the gray wolf relative to the three target gray wolves is calculated according to the following formula:

[0078]

[0079] Among them, α, β, and δ represent the labels of the three target gray wolves, and are the distances of the gray wolf relative to the target gray wolf α, β, δ, respectively. is the coefficient vector used in this round of optimization, and They are the locations of the three target gray wolves. For the location of the Gray Wolf.

[0080] A14. According to the distance of the gray wolf relative to the three target gray wolves and the coefficient vector, the gray wolf is controlled to move in the direction of the three target gray wolves. In this embodiment, the position of the gray wolf after moving in the direction of the three target gray wolves is calculated according to the following formula:

[0081]

[0082] in, is the direction and distance that the gray wolf steps toward the target gray wolf α, is the direction and distance that the gray wolf steps toward the target gray wolf β, is the direction and distance that the gray wolf steps toward the target gray wolf δ, and are the coefficient vectors used in this round of optimization, Indicates the position of the gray wolf after it moves toward the three target gray wolves.

[0083] A15, update the coefficient vector and calculate the fitness value of each gray wolf. In this embodiment, the coefficient vector is updated according to the following formula:

[0084]

[0085] in, and are the coefficient vectors used in the Grey Wolf Optimization Algorithm. is the control vector, It decreases linearly from 2 to 0. and is a random vector in [0,1].

[0086] A16. Determine whether the maximum number of iterations has been reached. If so, execute step A17; if not, execute step A18.

[0087] A17. The decomposition level and penalty factor of the gray wolf with the highest fitness value are taken as the optimal decomposition level and optimal penalty factor.

[0088] A18, taking the three gray wolves with the highest fitness values ​​as target gray wolves, and jumping to step A13 "for any gray wolf in the gray wolf population except the target gray wolf, calculating the distance of the gray wolf relative to the three target gray wolves".

[0089] Specifically in this embodiment, before the normalized energy value is input into a pre-trained fault diagnosis model in step S4 to obtain the fault type of the MOSFET power drive circuit, the MOSFET fault diagnosis method based on drive current data also includes a process of constructing a data set and training a convolutional neural network, such as Figure 3 The flowchart shown in FIG. 1 includes the following steps:

[0090] B1. Simulate the hard fault of the MOSFET power drive circuit through experiments and simulations, and obtain the drive current data under the hard fault.

[0091] B2. Accelerate the aging process of components through high and low temperature test chambers, simulate the soft fault of MOSFET power drive circuit, and obtain the drive current data under soft fault.

[0092] B3. Use the optimized variational mode decomposition algorithm to decompose each driving current signal to obtain several intrinsic mode components.

[0093] B4. Calculate the energy of each eigenmode component, and select several eigenmode components with the largest energy to calculate the normalized energy value of the driving current data. In this embodiment, the energy of the eigenmode component is calculated according to the following formula:

[0094]

[0095] Among them, E j is the energy of the jth eigenmode component, u j (t) is the jth eigenmode component.

[0096] B5. Construct a fault diagnosis data set according to the normalized energy value of each driving current data and the fault type corresponding to each driving current data.

[0097] B6. Using the fault diagnosis data set, taking the normalized energy value of the driving current data as input and the fault type corresponding to the driving current data as the target output, a convolutional neural network is trained to obtain a pre-trained fault diagnosis model. Specifically in this embodiment, the convolutional neural network includes an input layer, two pairs of alternating and corresponding convolutional layers, a pooling layer, a fully connected layer, and an output layer.

[0098] Next, to facilitate the explanation of the improvement route of the above-mentioned MOSFET fault diagnosis method based on drive current data provided in this embodiment, a typical MOSFET drive circuit, a dual voltage time-sharing drive circuit in a fuel system drive circuit, is taken as an example for explanation. This circuit is relatively complex and can fully illustrate the technical solution and effect of the present application. However, it can be understood that the application scope of the method of the present application is not limited to this, and can also be applied to all other MOSFET power drive circuits.

[0099] The specific technical solution is explained in combination with the dual voltage time-sharing drive circuit. The schematic diagram of the dual voltage time-sharing drive circuit of the fuel system drive circuit is as follows: Figure 4 As shown, D2, D3 and D4 are all diodes. The function of diode D2 is to prevent the high voltage source from charging to the low voltage source, the function of diode D3 is to freewheel from GND, and the function of diode D4 is to prevent the high voltage source from charging to the middle position between the solenoid valve and MOSFET3; its current waveform is as follows Figure 5 As shown, I peak_max Indicates the maximum value of peak current, I peak_min Indicates the minimum peak current, I hold_max Indicates the maximum holding current, I hold_min Indicates the minimum holding current.

[0100] Before performing specific fault diagnosis, the convolutional neural network model used needs to be trained first. Before training, a fault diagnosis dataset needs to be constructed. The steps for constructing the fault diagnosis dataset include:

[0101] 1) Acquisition of original fault and performance degradation data.

[0102] The failure modes of MOSFET power drive circuits mainly include structural failure (hard failure) and parameter failure (soft failure). When a hard failure occurs, it will seriously affect the drive current waveform. Through experiments and simulations, hard failure conditions are simulated to obtain current waveform data when hard failure occurs.

[0103] 1.1) Raw data acquisition under hard fault.

[0104] Hard faults are divided into high-end bootstrap suspension drive fault, MOSFET open circuit fault and MOSFET short circuit fault. For the two fault forms of high-end bootstrap suspension drive fault and MOSFET open circuit fault, the drive current waveform is smaller than the normal value, so it can be simulated by test; while the drive current waveform of MOSFET short circuit fault is larger than the normal value, which cannot be simulated by normal test, so it is simulated by simulation. For high-end bootstrap suspension drive fault, the current waveform after the fault occurs is as follows Figure 6 As shown in the figure, for MOSFET open circuit fault and MOSFET short circuit fault, the fault forms of the three MOSFET devices and the current waveforms after the fault are as follows: Figure 7 to Figure 9 shown.

[0105] 1.2) Raw data acquisition under soft fault.

[0106] Soft failure refers to the aging of components caused by long-term use of MOSFET. In normal use, the aging process of components is relatively slow. Therefore, a MOSFET accelerated aging test method is designed in this embodiment. For semiconductor chips, in order to increase the degradation rate and shorten the aging time, it is necessary to increase the chip junction temperature T as much as possible. j . The chip junction temperature T j It is difficult to measure directly, so in this embodiment, the chip junction temperature T is estimated approximately by the following formula: j :

[0107] T j =T c +ψ JT P D .

[0108] Among them, T j is the junction temperature of the semiconductor device, T c is the ambient temperature, ψ JT P is the thermal characteristic parameter between the node and the center of the package surface, and its size is related to the properties of the semiconductor device itself. D is the package power consumption.

[0109] Therefore, in this embodiment, the ambient temperature and the operating frequency of the MOSFET are increased to increase the T c With P D , to achieve the purpose of accelerating the aging of components. In a further improved embodiment, the following can be used Fig.10 The accelerated aging test platform shown in the figure has data measurement points distributed as follows: Fig.11As shown in the figure, ①②③ are the locations of the three data measurement points. The use process includes: setting the high and low temperature test chamber to a high temperature state relative to the standard state, such as 85℃, setting the high-frequency drive operating frequency, and actively cutting off the power supply of the high and low temperature test chamber after every 500,000 power drives, and then heating it up to 85℃ again after cooling to room temperature, and accelerating all bonding and interface degradation between the semiconductor chip and the mounting surface again in an active temperature cycle.

[0110] 2) Current characteristic signal extraction.

[0111] The driving current signal is essentially a time domain signal. When a hard fault or soft fault occurs in the driving circuit, the response of the current signal flowing through the actuator during the current mutation process will also change. This change is particularly obvious in the frequency domain. In this application, the time series of the current signal flowing through the solenoid valve is regarded as a non-stationary signal. Taking the center frequency as the reference, the original signal is decomposed into K IMF components under the optimization conditions of K IMF component bandwidth and minimum. The expression is shown as follows:

[0112]

[0113] Among them, u k is the modal component sequence obtained after VMD decomposition, k = 1, 2, ..., K; ω k is the center frequency corresponding to each modal component; is the partial derivative of the function with respect to time; δ(t) is the unit pulse function; e -jωt is a complex exponential signal; f is the original driving current signal.

[0114] In order to find the optimal solution to the above variational problem, the quadratic penalty term α and the Lagrange multiplier λ are introduced. The quadratic penalty term constrains the smoothness of the signal decomposition during the optimization process to avoid over-decomposition and oscillation. The Lagrange multiplier ensures that certain conditions and constraints are met in the signal decomposition. The introduction of the two transforms the constrained variational problem into an unconstrained variational problem.

[0115] In the above process of using the variational mode decomposition algorithm VMD, the setting of the decomposition layer number K and the penalty factor α will affect the effect of VMD. Improper selection will cause information loss or mode aliasing, and the selection of parameters is irregular. However, in the discrimination of soft and hard faults, there are many types of faults and a large number of samples. The method of determining parameters based on experience takes a long time and lacks evaluation criteria. Therefore, the gray wolf algorithm is introduced in this embodiment to accelerate the optimization of VMD. The Grey Wolf Optimizer (GWO) is a swarm intelligence optimization algorithm that simulates the social behavior and hierarchical structure of the gray wolf group to solve the optimization problem. Specifically, the optimization process is as follows:

[0116] Initially, n vectors are randomly generated within the custom allowed range, that is, n gray wolves are generated, and the fitness function value of each individual in the population is calculated. The three gray wolves with the best fitness in the wolf pack are marked as α wolf, β wolf, and δ wolf in turn, and the remaining wolves are marked as ω wolf. Since the location of the optimal solution is unknown, it is temporarily believed that α wolf, β wolf, and δ wolf represent the potential location of the optimal solution. According to the above-mentioned level division, high-level gray wolves have absolute dominance over low-level gray wolves. For each ω wolf, the distance of ω wolf relative to α wolf, β wolf, and δ wolf is calculated according to the following formula:

[0117]

[0118] in, is the current position vector of the gray wolf, is the approximate distance of the current gray wolf towards α wolf, β wolf, and δ wolf. is the coefficient vector. This process uses the different positions of α wolf, β wolf, and δ wolf to estimate the position of the prey, and then changes the position of ω wolf according to the distance calculated above, as shown in the following formula:

[0119]

[0120] in, is the direction and distance that the ω wolf steps towards the α wolf, β wolf, and δ wolf, represents the final position of ω wolf. The coefficient vector The update equation is as follows:

[0121]

[0122] in, It decreases linearly from 2 to 0. is a random vector of [0, 1], that is, the direction of the gray wolf's movement has a certain randomness, which can avoid the occurrence of local optimal phenomenon to a certain extent.

[0123] In the above optimization process, the fitness function used is selected as follows. When optimizing parameters, the evaluation criterion for the decomposition effect of the VMD method is the envelope entropy E proposed by Tang Guiji et al. p , the envelope entropy is defined as follows:

[0124]

[0125] Where a(j) is the envelope signal obtained by Hilbert transforming the original driving current data time series f(j), p j is the normalized result. If the obtained component still contains high-frequency noise, the sparseness of the component signal will be weakened, the envelope entropy will be relatively large, and the extracted fault information will be submerged in the noise signal. Therefore, the fitness function for optimizing VMD parameters is set to minEp .

[0126] 3) Feature vector selection.

[0127] After VMD decomposition of the original signal, several intrinsic mode components can be obtained. A single IMF component consists of thousands of data points. If it is directly input into the neural network, it will greatly increase the complexity of the calculation and the training time, leading to the dimensionality disaster and overfitting. Therefore, in this embodiment, the energy value of each component after decomposition is first calculated according to the following formula:

[0128]

[0129] Among them, u j (t) is the signal sequence of the intrinsic mode component after VMD decomposition.

[0130] Then take the normalized energy value Q of the first four IMF components according to the following formula: * The feature vector used as input to the neural network:

[0131]

[0132] That is, the values ​​of the four components E1 to E4 are divided by the total energy value to obtain the normalized energy values ​​corresponding to the four components, and the obtained Q 1 , Q 2 , Q 3 , Q 4 The feature vector is used as input to the neural network.

[0133] In summary, a fault diagnosis data set is constructed based on the normalized energy value of each driving current data and the corresponding fault type. The normalized energy value is used as the input data, and the corresponding fault type is used as the label of the model.

[0134] 4) Training of convolutional neural networks.

[0135] In this embodiment, a convolutional neural network is used to judge faults and performance degradation. The convolutional neural network consists of an input layer, two pairs of alternating and corresponding convolutional layers, a pooling layer, a fully connected layer, and an output layer. The convolutional layer uses a convolution kernel to extract feature parameters. In order to prevent the loss of feature parameters, the maximum pooling method is used to reduce the dimension of the high-dimensional data after convolution. In the process of training the convolutional neural network using the fault diagnosis data set, the method of decreasing the learning rate is used to prevent overfitting. The fully connected layer uses the Softmax function as the activation function of the output layer, and Adam is used as an optimizer to effectively improve the training speed, convergence effect, and generalization ability of the model. The training process of the neural network is relatively conventional and will not be repeated here. After completing the training of the model, it can be applied to the MOSFET fault diagnosis method based on the drive current data provided in the above embodiment, and the diagnosis of the MOSFET fault mode based on the drive current data is completed through steps S1 to S4 of the above method.

[0136] The above-mentioned MOSFET fault diagnosis method based on driving current data proposed in this application optimizes the parameters of the variational mode decomposition algorithm based on the Gray Wolf optimization algorithm, and then obtains the intrinsic mode component according to the variational mode decomposition, and calculates the energy value of the intrinsic mode component as the characteristic vector of CNN for fault judgment, which can avoid the misjudgment and missed judgment of human experience, and realizes the judgment and identification of the short circuit, open circuit and performance degradation mode of the dual voltage time-sharing drive circuit. Specifically, it has the following advantages:

[0137] (1) The driving current signal is essentially a time domain signal. When a hard fault or a soft fault occurs in the driving circuit, the response of the current signal flowing through the actuator during the current mutation process will also change. This change is particularly obvious in the frequency domain. Starting from the time domain signal, this application extracts the driving current data under the two major fault forms of hard fault and soft fault, and extracts the characteristic parameters in the frequency domain from the current time domain signal, thus avoiding the problem of susceptibility to interference based on time domain signal judgment.

[0138] (2) This application optimizes the means of feature extraction. In the calculation process of VMD, the setting of the decomposition layer number K and the penalty factor α will affect the effect of VMD. Improper selection will cause information loss or modal aliasing, and the selection of parameters is irregular. However, in the discrimination of soft and hard faults, there are many types of faults and a large number of samples. The method of determining parameters based on experience is time-consuming and lacks evaluation criteria. Therefore, the gray wolf algorithm is introduced to accelerate the optimization of VMD.

[0139] (3) During feature extraction optimization, if the obtained component still contains high-frequency noise, the sparseness of the component signal will become weaker, the envelope entropy will be larger, and the extracted fault information will be submerged in the noise signal. Therefore, the fitness function for optimizing VMD parameters is set to minE p .

[0140] (4) After VMD decomposition, the original driving current signal can obtain several intrinsic mode components. Since a single IMF component consists of thousands of data points, if it is directly input into the neural network, it will greatly increase the complexity of the calculation and the training time, leading to the dimensionality disaster and also causing overfitting. Therefore, this application first calculates the energy value of each component after decomposition, and takes the normalized energy value Q of the first four IMF components * As the feature variable of the input neural network, the above problems are avoided.

[0141] Based on the same inventive concept, the embodiment of the present application also provides a system for implementing the above-mentioned MOSFET fault diagnosis method based on drive current data. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more system embodiments provided below can refer to the limitations of the MOSFET fault diagnosis method based on drive current data above, and will not be repeated here.

[0142] In an exemplary embodiment, Fig.12 As shown, a MOSFET fault diagnosis system based on drive current data is provided, including the following modules:

[0143] A drive current data acquisition module is used to acquire drive current data when the MOSFET power drive circuit is working;

[0144] A driving current data decomposition module is used to decompose the driving current signal using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after optimizing the decomposition layer number and penalty factor using the grey wolf optimization algorithm;

[0145] A normalized energy value calculation module, used to calculate the energy of each of the intrinsic mode components, and select a number of eigenmode components with the largest energy to calculate the normalized energy value of the driving current data;

[0146] The fault type diagnosis module is used to input the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

[0147] certainly, Fig.12 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different functions. Fig.12 One or at least two components of the system shown.

[0148] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.13 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a MOSFET fault diagnosis method based on drive current data provided in the above embodiment is implemented.

[0149] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0150] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0151] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0152] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] 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 used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0154] 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, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0155] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0156] 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.

[0157] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A MOSFET fault diagnosis method based on drive current data, characterized in that: include: Obtain the driving current data when the MOSFET power driving circuit is working; The driving current signal is decomposed by using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after the number of decomposition layers and the penalty factor are optimized by using a grey wolf optimization algorithm; Calculating the energy of each of the eigenmode components, and selecting several eigenmode components with the largest energy to calculate the normalized energy value of the driving current data; The normalized energy value is input into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

2. The MOSFET fault diagnosis method based on drive current data according to claim 1, characterized in that: Before using the optimized variational mode decomposition algorithm to decompose the driving current signal to obtain a plurality of intrinsic mode components, the method further includes: The Grey Wolf Optimization Algorithm is used to optimize the number of decomposition layers and penalty factors of the variational mode decomposition algorithm to obtain the optimized variational mode decomposition algorithm; specifically, it includes: Initialize a gray wolf population and a coefficient vector; the gray wolf population includes a plurality of gray wolves, each gray wolf corresponds to a set of decomposition layers and a penalty factor; Calculate the fitness value of each gray wolf, and select the three gray wolves with the highest fitness values ​​as target gray wolves; For any gray wolf in the gray wolf population except the target gray wolf, calculating the distance of the gray wolf relative to the three target gray wolves; Controlling the gray wolf to move toward the three target gray wolves according to the distance of the gray wolf relative to the three target gray wolves and the coefficient vector; Updating the coefficient vector and calculating the fitness value of each gray wolf; Determine whether the maximum number of iterations has been reached. If so, the decomposition level and penalty factor of the gray wolf with the highest fitness value are taken as the optimal decomposition level and optimal penalty factor; If not, the three gray wolves with the highest fitness values ​​are taken as target gray wolves, and the process jumps to step "for any gray wolf in the gray wolf population except the target gray wolf, calculating the distance between the gray wolf and the three target gray wolves".

3. The MOSFET fault diagnosis method based on drive current data according to claim 2, characterized in that: The fitness value of each gray wolf is calculated according to the following formula: Where a(j) is the envelope signal obtained after Hilbert transform of the original driving current data, p j is the result of normalizing the envelope signal, N is the length of the time series of the original driving current data, E p is the envelope entropy, i.e., the fitness value.

4. The MOSFET fault diagnosis method based on drive current data according to claim 2, characterized in that: The distance between the gray wolf and the three target gray wolves is calculated according to the following formula: Among them, α, β, and δ represent the labels of the three target gray wolves, and are the distances of the gray wolf relative to the target gray wolf α, β, δ, respectively. is the coefficient vector used in this round of optimization, and They are the locations of the three target gray wolves. For the location of the Gray Wolf.

5. The MOSFET fault diagnosis method based on drive current data according to claim 4, characterized in that: The position of the gray wolf after moving towards the three target gray wolves is calculated according to the following formula: in, is the direction and distance that the gray wolf steps toward the target gray wolf α, is the direction and distance that the gray wolf steps toward the target gray wolf β, is the direction and distance that the gray wolf steps toward the target gray wolf δ, and are the coefficient vectors used in this round of optimization, Indicates the position of the gray wolf after it moves toward the three target gray wolves.

6. The MOSFET fault diagnosis method based on drive current data according to claim 5, characterized in that: The coefficient vector is updated according to the following formula: in, and are the coefficient vectors used in the Grey Wolf Optimization Algorithm. is the control vector, It decreases linearly from 2 to 0. and is a random vector in [0,1].

7. The MOSFET fault diagnosis method based on drive current data according to claim 1, characterized in that: Before inputting the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit, the method further includes: Simulate the hard fault of MOSFET power drive circuit through experiments and simulations, and obtain the drive current data under hard fault; Accelerate the aging process of components through high and low temperature test chambers, simulate soft faults of MOSFET power drive circuits, and obtain drive current data under soft faults; Decomposing each of the driving current signals using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; Calculating the energy of each of the eigenmode components, and selecting several eigenmode components with the largest energy to calculate the normalized energy value of the driving current data; Constructing a fault diagnosis data set according to the normalized energy value of each of the driving current data and the fault type corresponding to each of the driving current data; The fault diagnosis data set is used to train a convolutional neural network with the normalized energy value of the driving current data as input and the fault type corresponding to the driving current data as target output to obtain a pre-trained fault diagnosis model.

8. The MOSFET fault diagnosis method based on drive current data according to claim 7, characterized in that: The convolutional neural network includes an input layer, two pairs of alternating and corresponding convolutional layers, a pooling layer, a fully connected layer and an output layer.

9. The MOSFET fault diagnosis method based on drive current data according to claim 1 or 7, characterized in that: The energy of the eigenmode components is calculated according to the following formula: Among them, E j is the energy of the jth eigenmode component, u j (t) is the jth eigenmode component.

10. A MOSFET fault diagnosis system based on drive current data, characterized in that: include: A drive current data acquisition module is used to acquire drive current data when the MOSFET power drive circuit is working; A driving current data decomposition module is used to decompose the driving current signal using an optimized variational mode decomposition algorithm to obtain a plurality of intrinsic mode components; the optimized variational mode decomposition algorithm is a variational mode decomposition algorithm after optimizing the decomposition layer number and penalty factor using the grey wolf optimization algorithm; A normalized energy value calculation module, used to calculate the energy of each of the eigenmode components, and select a number of eigenmode components with the largest energy to calculate the normalized energy value of the driving current data; The fault type diagnosis module is used to input the normalized energy value into a pre-trained fault diagnosis model to obtain the fault type of the MOSFET power drive circuit; the pre-trained fault diagnosis model is a pre-trained convolutional neural network.

Images