Sliding bearing fault detection method and system based on IGWO-VMD-CNN

Abstract

The invention discloses a sliding bearing fault detection method and system based on IGWO-VMD-CNN. The method comprises the following steps: acquiring an acoustic emission signal of a sliding bearing; introducing improved tent mapping, a nonlinear convergence operator and an improved wolf position updating mechanism, and constructing an improved grey wolf optimization algorithm (IGWO); based on an improved grey wolf optimization algorithm, reconstructing the acoustic emission signal of the sliding bearing in combination with a variational mode decomposition (VMD) algorithm to obtain a reconstructed acoustic emission signal of the sliding bearing; and constructing a sliding bearing fault detection neural network (CNN), and carrying out fault detection identification processing on the reconstructed sliding bearing acoustic emission signal to obtain a sliding bearing fault detection result. According to the embodiment of the invention, the influence of the noise signal on fault signal detection can be reduced, and the fault detection and identification precision of the sliding bearing is improved. The sliding bearing fault detection method and system based on the IGWO-VMD-CNN can be widely applied to the technical field of sliding bearing fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of sliding bearing fault diagnosis, and in particular to a sliding bearing fault detection method and system based on IGWO-VMD-CNN. Background Art

[0002] Sliding bearings are important supporting components of high-end equipment. They work at high speeds, heavy loads, and in complex and harsh environments for a long time. If a failure occurs, it will cause huge economic losses. Therefore, it is extremely important to study the fault diagnosis of sliding bearings. Traditional vibration analysis has been widely used in bearing fault diagnosis, but the early fault signals are extremely weak and can easily be submerged in the noise signals of other components and cannot be detected. Acoustic emission signals are signals generated by factors such as elastic deformation, crack propagation, and wear and corrosion of the internal materials of the bearing. Therefore, the detected signal is much higher than the background noise band, which can reflect the real defect information inside the material and help to effectively extract the fault signal.

[0003] For the extraction of fault information from acoustic emission signals, the variational mode decomposition algorithm (VMD) has strong robustness, higher decomposition accuracy, better sparsity, can effectively avoid the occurrence of endpoint effects and modal aliasing, has strong analysis and processing capabilities for complex nonlinear and non-stationary signals, and can successfully solve the problem of early weak fault signals being difficult to detect and diagnose. However, in actual application, the decomposition effect of VMD is easily affected by the number of decomposition layers and penalty factors, which leads to inaccurate fault information of the acquired acoustic emission signals, making it difficult to accurately judge the fault problem of the sliding bearing. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a sliding bearing fault detection method and system based on IGWO-VMD-CNN, which can reduce the influence of noise signals on the detection of fault signals, thereby improving the fault detection and identification accuracy of sliding bearings.

[0005] The first technical solution adopted by the present invention is: a sliding bearing fault detection method based on IGWO-VMD-CNN, comprising the following steps:

[0006] Acquiring an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes a sliding bearing normal acoustic emission signal and a sliding bearing fault acoustic emission signal;

[0007] Improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism are introduced to construct an improved grey wolf optimization algorithm (IGWO);

[0008] Based on the improved Grey Wolf optimization algorithm and combined with the variational mode decomposition (VMD) algorithm, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing.

[0009] A sliding bearing fault detection neural network (CNN) was constructed, and the reconstructed sliding bearing acoustic emission signals were processed for fault detection and identification to obtain the sliding bearing fault detection results.

[0010] Furthermore, the improved tent mapping is used to initialize the population, and the expression of the improved tent mapping is specifically as follows:

[0011]

[0012] In the above formula, t represents the t-th iteration, β represents the chaos parameter, x(t) represents the chaos mapping function of the t-th iteration, x(t+1) represents the chaos mapping function of the t+1 iteration, N represents the number of particles in the sequence, and rand(1) / N represents the random variable.

[0013] Furthermore, the expression of the nonlinear convergence operator is specifically as follows:

[0014] a=1+sin(π / 2+π×iter / Max_iter)

[0015] In the above formula, a represents the nonlinear convergence operator, iter represents the current number of iterations, and Max_iter represents the maximum number of iterations.

[0016] Furthermore, the improved wolf position update mechanism includes improving the position update of the Alpha wolf and the population position update based on the adaptive hunting weight coefficient, wherein:

[0017] The expression for improving the position update of the Alpha wolf is specifically as follows:

[0018]

[0019] X 1 (t+1)=X α (t)-a 1 ×D α (t)

[0020] In the above formula, C 1 is a random number between [0, 1], X α (t) represents the position vector of Alpha wolf in the tth iteration, X(t) represents the position of the current population, and D α (t) represents the distance between the other individuals in the group and the Alpha wolf at this time, X 1(t+1) represents the updated position vector of Alpha, and rand(1) represents a random variable;

[0021] The expression of population position update based on the adaptive hunting weight coefficient is specifically as follows:

[0022] X(t+1)=θ α ×X 1 +θ β ×X 2 +θ δ ×X 3

[0023]

[0024] In the above formula, α, β, δ represent Alpha, Beta, and Delta respectively, ω i represents the inertia coefficient corresponding to each head, f i represents the fitness function value corresponding to each wolf, θ i Indicates the weight corresponding to each wolf.

[0025] Furthermore, the step of reconstructing the acoustic emission signal of the sliding bearing based on the improved Grey Wolf optimization algorithm to obtain the reconstructed acoustic emission signal of the sliding bearing specifically includes:

[0026] Based on the improved grey wolf optimization algorithm, the optimal number of decomposition layers and the optimal penalty factor of the variational mode decomposition algorithm are determined, and the variational mode decomposition algorithm with optimized parameters is obtained;

[0027] The acoustic emission signal of the sliding bearing is decomposed by the variational mode decomposition algorithm after parameter optimization to obtain several IMF components.

[0028] The sample entropy function is used as the fitness function, and the IMF component corresponding to the minimum sample entropy among several IMF components is selected based on the principle of minimum fitness function;

[0029] Based on the IMF component corresponding to the minimum sample entropy, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing.

[0030] Furthermore, the step of determining the optimal number of decomposition layers and the optimal penalty factor of the variational mode decomposition algorithm based on the improved grey wolf optimization algorithm to obtain the variational mode decomposition algorithm after parameter optimization specifically includes:

[0031] Initialize the values ​​of various parameters α, β, δ in the improved gray wolf optimization algorithm and generate the positions of individual wolves in the wolf pack based on the improved tent mapping;

[0032] Define the sample entropy function as the fitness function;

[0033] The individual position in the wolf pack corresponds to the parameter combination [K, α] of the variational mode decomposition algorithm. The variational mode decomposition algorithm is decomposed according to [K, α], and the individual fitness value of each wolf in the wolf pack is calculated according to the fitness function;

[0034] Based on the improved position update method of Alpha wolf and the population position update method of adaptive hunting weight coefficient, α, β, and δ are updated to recalculate the fitness value of each wolf;

[0035] Determine whether the maximum iteration is reached. If not, return to the individual fitness value calculation step of the wolf to continue iterating. If reached, output the parameter combination [K, α] corresponding to the optimal fitness value to obtain the variational mode decomposition algorithm after parameter optimization.

[0036] Furthermore, the step of decomposing the acoustic emission signal of the sliding bearing by the variational mode decomposition algorithm after parameter optimization to obtain a plurality of IMF components specifically includes:

[0037] Based on the variational mode decomposition algorithm after parameter optimization, the optimal parameter combination [K, α] is used as the selected mode number K and penalty parameter α to decompose the acoustic emission signal of the sliding bearing;

[0038] The decomposition is stopped until the decomposition iteration conditions of the variational mode decomposition algorithm are met, and the intrinsic mode functions and residual vectors of K modes, that is, several IMF components, are obtained.

[0039] Furthermore, the step of constructing a sliding bearing fault detection neural network, performing fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtaining a sliding bearing fault detection result specifically includes:

[0040] Label the reconstructed acoustic emission signal according to the fault type;

[0041] Constructing a sliding bearing fault detection neural network, wherein the sliding bearing fault detection neural network includes an input layer, a feature extraction layer and a classification layer;

[0042] The reconstructed acoustic emission signal is input into the feature extraction layer through the input layer to extract features and obtain the features of the reconstructed acoustic emission signal;

[0043] The reconstructed acoustic emission signal features are input into the classification layer, and the classification output is realized through the Softmax regression classifier to obtain the sliding bearing fault detection result.

[0044] The second technical solution adopted by the present invention is: a sliding bearing fault detection system based on IGWO-VMD-CNN, comprising:

[0045] An acquisition module, used for acquiring an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes an acoustic emission signal of a sliding bearing without fault and an acoustic emission signal of a sliding bearing with fault;

[0046] Building a module to introduce improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism to build an improved grey wolf optimization algorithm (IGWO);

[0047] A reconstruction module is used to reconstruct the acoustic emission signal of the sliding bearing based on the improved Grey Wolf optimization algorithm combined with the variational mode decomposition (VMD) algorithm to obtain the reconstructed acoustic emission signal of the sliding bearing;

[0048] The detection module is used to construct a sliding bearing fault detection neural network (CNN), perform fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtain the sliding bearing fault detection result.

[0049] The beneficial effects of the method and system of the present invention are as follows: the present invention acquires the acoustic emission signal of the sliding bearing, and then introduces the improved tent mapping, the nonlinear convergence operator and the improved wolf position update mechanism, constructs the improved gray wolf optimization algorithm, applies the improved gray wolf optimization algorithm to the determination of the variational mode decomposition parameters, solves the problem that the VMD parameters need to be set based on experience, introduces the improved tent mapping, the nonlinear convergence operator and the improved wolf position update mechanism, the population diversity of the adaptive hunting weight coefficient enhancement algorithm, and the global and local search capabilities of the balance algorithm to improve the initial gray wolf algorithm, avoids the problem that the algorithm is prone to fall into the local optimum in the later stage of iteration, and reconstructs the acoustic emission signal of the sliding bearing based on the improved gray wolf optimization algorithm, reduces the influence of the noise signal on the detection of the fault signal, and thus improves the fault detection and identification accuracy of the sliding bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of the steps of a sliding bearing fault detection method based on IGWO-VMD-CNN of the present invention;

[0051] Figure 2 It is a structural block diagram of a sliding bearing fault detection system based on IGWO-VMD-CNN of the present invention;

[0052] Figure 3 It is an improved tent chaos map provided by a specific embodiment of the present invention;

[0053] Figure 4 It is an improved tent chaos value histogram provided by a specific embodiment of the present invention;

[0054] Figure 5It is a schematic diagram comparing optimization curves of the improved grey wolf algorithm and the grey wolf algorithm provided in a specific embodiment of the present invention;

[0055] Figure 6 It is a schematic diagram of the IGWO-VMD decomposition result of the acoustic emission signal provided by a specific embodiment of the present invention;

[0056] Figure 7 is a schematic diagram of sample entropy of each component of an acoustic emission signal provided by a specific embodiment of the present invention;

[0057] Figure 8 is a schematic diagram comparing the reconstructed acoustic emission signal and the original acoustic emission signal provided by a specific embodiment of the present invention;

[0058] Fig. 9 is a schematic diagram of a CNN fault diagnosis result based on raw data provided by a specific embodiment of the present invention;

[0059] Fig.10 It is a schematic diagram of a CNN fault diagnosis confusion matrix based on raw data provided by a specific embodiment of the present invention;

[0060] Fig.11 It is a schematic diagram of IGWO-VMD-CNN fault diagnosis results provided by a specific embodiment of the present invention;

[0061] Fig.12 It is a schematic diagram of the IGWO-VMD-CNN fault diagnosis confusion matrix provided by a specific embodiment of the present invention. DETAILED DESCRIPTION

[0062] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only provided for the convenience of explanation and description, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0063] Reference Figure 1 The present invention provides a sliding bearing fault detection method based on IGWO-VMD-CNN, which comprises the following steps:

[0064] S1. Acquire an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes a sliding bearing normal acoustic emission signal and a sliding bearing fault acoustic emission signal;

[0065] Specifically, the acoustic emission signals of the sliding bearing under no-fault conditions and various fault conditions are collected.

[0066] S2, introduce improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism to construct improved grey wolf optimization algorithm (IGWO);

[0067] Specifically, the initial gray wolf algorithm is improved by introducing improved tent mapping, nonlinear convergence operator, improved alpha position update mechanism, and adaptive hunting weight coefficient. Based on the improved gray wolf algorithm, the optimal decomposition layer number K and penalty factor α of the variational mode decomposition algorithm are determined, as follows: Figure 5 The figure shows a comparison diagram of the optimization curves between the improved grey wolf algorithm and the grey wolf algorithm.

[0068] S21, using improved tent chaotic mapping to initialize the population;

[0069] Specifically, the improved tent chaotic mapping introduces a random variable rand(1) / N into the standard tent chaotic mapping function. While retaining the randomness, ergodicity and regularity of the tent chaotic mapping, it avoids the tent chaotic sequence from falling into small periodic points and unstable periodic points during iteration, such as Figure 3 and Figure 4 As shown, the improved tent chaos mapping expression is:

[0070]

[0071] In the above formula, t represents the t-th iteration, β represents the chaos parameter, x(t) represents the chaos mapping function of the t-th iteration, x(t+1) represents the chaos mapping function of the t+1 iteration, N represents the number of particles in the sequence, and rand(1) / N represents the random variable.

[0072] S22, introduce nonlinear convergence operators to improve the global and local search capabilities of the Grey Wolf Algorithm;

[0073] Specifically, the use of a sinusoidal convergence operator a effectively balances the global and local search capabilities of the algorithm, and its expression is:

[0074] a=1+sin(π / 2+π×iter / Max_iter)

[0075] In the above formula, a represents the nonlinear convergence operator, iter represents the current number of iterations, and Max_iter represents the maximum number of iterations.

[0076] S23. Improve the position update method of Alpha Wolf to further improve the globality of the algorithm and avoid local optimality;

[0077] Specifically, the position update method of the Alpha wolf is improved by combining the sine and cosine algorithms to balance the exploration and development processes and accelerate the convergence of the algorithm. The update method for the remaining wolves is the same as the traditional GWO algorithm. The position update formula guided by the Alpha wolf is as follows:

[0078]

[0079] X 1 (t+1)=X α (t)-a 1 ×D α (t)

[0080] In the above formula, C 1 is a random number between [0, 1], X α (t) represents the position vector of Alpha wolf in the tth iteration, X(t) represents the position of the current population, and D α (t) represents the distance between the other individuals in the group and the Alpha wolf at this time, X 1 (t+1) represents the updated position vector of Alpha, and rand(1) represents a random variable.

[0081] S24. Update the population position based on the adaptive hunting weight coefficient.

[0082] Specifically, based on the adaptive hunting weight coefficient to update the population position, the mass update formula of the gravitational search algorithm is introduced. The Alpha wolf provides the main movement direction for the gray wolf community, while the Beta wolf and the Delta wolf provide auxiliary directions to speed up the encirclement and attack on the prey. The formula for obtaining the updated population position is:

[0083] X(t+1)=θ α ×X 1 +θ β ×X 2 +θ δ ×X 3

[0084]

[0085] In the above formula, α, β, δ represent Alpha wolf, Beta wolf and Delta wolf respectively, ω i represents the inertia coefficient of each wolf, f i represents the fitness function value corresponding to each wolf, θ i Indicates the weight corresponding to each wolf.

[0086] S3. Based on the improved Grey Wolf optimization algorithm and combined with the variational mode decomposition (VMD) algorithm, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing;

[0087] Specifically, based on the improved Grey Wolf optimization algorithm, the optimal decomposition layer number and the optimal penalty factor of the variational mode decomposition algorithm are determined to obtain the variational mode decomposition algorithm after parameter optimization; the acoustic emission signal of the sliding bearing is decomposed by the variational mode decomposition algorithm after parameter optimization to obtain several IMF components; the sample entropy function is used as the fitness function, and the IMF component corresponding to the minimum sample entropy among several IMF components is selected based on the principle of minimum fitness function; based on the IMF component corresponding to the minimum sample entropy, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing.

[0088] In the embodiment of the present invention, the acquired acoustic emission signal is decomposed into a series of modal components using a variational modal decomposition algorithm with optimized parameters, such as Figure 6 As shown. Based on the principle of minimum fitness function, effective IMF components are selected, and the sample entropy function is used as the fitness function. The sample entropy diagram of each component is shown in Figure 7 As shown, the three IMF components with the smallest sample entropy are selected to reconstruct the acoustic emission signal. The reconstructed acoustic emission signal is shown in Figure 8 As shown in the figure, the effective IMF components obtained by decomposing the acoustic emission signals under fault-free and faulty conditions are constructed into the initial sample set of the convolutional neural network, and the fault state label is set according to the fault state corresponding to the data.

[0089] First, the parameters are initialized. The values ​​of various parameters α, β, and δ in GWO are initialized, and the positions of individual wolves in the wolf pack are generated based on the improved tent mapping. The sample entropy function is defined as the fitness function according to the problem requirements. The individual positions in the wolf pack correspond to the parameter combination [K, α] of VMD. VMD is decomposed according to [K, α]. The individual fitness value of each wolf in the wolf pack is calculated according to the fitness function. α, β, and δ are updated based on the improved Alpha wolf position update method and the population position update method of the adaptive hunting weight coefficient. The fitness value of each wolf is recalculated to determine whether the maximum iteration has been reached. If not, return to the individual fitness value calculation step of each wolf in the wolf pack to continue the iteration; otherwise, stop the iteration and obtain the parameter combination [K, α] corresponding to the optimal fitness value.

[0090] It should be noted that the optimal parameter combination [K, α] is used as the selected mode number K and penalty parameter α to decompose the original sliding bearing acoustic emission signal until the VMD decomposition iteration condition is met and the decomposition is stopped. The intrinsic mode functions IMF1, IMF2, ..., IMFK and residual vector RES of K modes are obtained, and the sample entropy function is calculated for the obtained IMF components of each order, and the three components with the smallest sample entropy are selected to reconstruct the acoustic emission signal.

[0091] S4. Construct a sliding bearing fault detection neural network (CNN), perform fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtain the sliding bearing fault detection result.

[0092] Specifically, a sliding bearing convolutional neural network model is constructed, and part of the sample sets are used as the training sets of the sliding bearing convolutional neural network model. The corresponding fault state labels are used as the expected outputs of the sliding bearing convolutional neural network model. The sliding bearing convolutional neural network is trained, and the remaining sample sets are input as test sets into the trained sliding bearing convolutional neural network model to verify the reliability of the model.

[0093] The reconstructed acoustic emission signals are labeled according to the fault type and divided into training set and test set at a ratio of 5:1. A network model is constructed, including an input layer, a feature extraction layer, and a classification layer. The data is input into the feature extraction layer through the input layer, including 2 convolutional layers and 2 pooling layers, to extract features from the reconstructed acoustic emission signals. The pooling layer selects the maximum pooling operator to reduce the dimension of the feature vector and improve the robustness of the nonlinear features. The extracted features are input into the classification layer, including a fully connected layer and a Softmax classifier. The number of neurons in the fully connected layer is consistent with the number of label faults, and the Softmax regression classifier is used to achieve classification output.

[0094] In summary, the embodiment of the present invention applies the improved GWO gray wolf optimization algorithm to the determination of variational mode decomposition parameters, solving the problem that VMD parameters need to be set based on experience. At the same time, since the GWO algorithm is prone to fall into the local optimum in the later iteration, the initial gray wolf algorithm is improved by introducing improved tent mapping, nonlinear convergence operator, improved leader position update mechanism, adaptive hunting weight coefficient enhancement algorithm population diversity, and global and local search capabilities of the balance algorithm. Then, the optimized VMD is used to reconstruct the acoustic emission signal of the sliding bearing; finally, the reconstructed signal is imported into the CNN model for training and testing to obtain the fault diagnosis result. Through comparative analysis, the proposed method obtains a higher diagnostic accuracy.

[0095] like Fig. 9 , Fig.10 , Fig.11 and Fig.12 As shown in the figure, the diagnosis model optimized by the improved grey wolf algorithm for VMD-CNN has better effect. The fault diagnosis accuracy of sliding bearings based on the original data CNN is only 65.93%, while the fault diagnosis accuracy of sliding bearings based on IGWO-VMD-CNN reaches 98.94%.

[0096] Reference Figure 2 , a sliding bearing fault detection system based on IGWO-VMD-CNN, comprising:

[0097] An acquisition module, used for acquiring an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes an acoustic emission signal of a sliding bearing without fault and an acoustic emission signal of a sliding bearing with fault;

[0098] Building a module to introduce improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism to build an improved grey wolf optimization algorithm (IGWO);

[0099] A reconstruction module is used to reconstruct the acoustic emission signal of the sliding bearing based on the improved Grey Wolf optimization algorithm combined with the variational mode decomposition (VMD) algorithm to obtain the reconstructed acoustic emission signal of the sliding bearing;

[0100] The detection module is used to construct a sliding bearing fault detection neural network (CNN), perform fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtain the sliding bearing fault detection result.

[0101] The contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0102] The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A sliding bearing fault detection method based on IGWO-VMD-CNN, characterized in that: The following steps are involved: Acquiring an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes a sliding bearing normal acoustic emission signal and a sliding bearing fault acoustic emission signal; Improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism are introduced to construct an improved grey wolf optimization algorithm (IGWO); Based on the improved Grey Wolf optimization algorithm and combined with the variational mode decomposition (VMD) algorithm, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing. A sliding bearing fault detection neural network (CNN) was constructed, and the reconstructed sliding bearing acoustic emission signals were processed for fault detection and identification to obtain the sliding bearing fault detection results.

2. According to claim 1, a sliding bearing fault detection method based on IGWO-VMD-CNN is characterized in that: The improved tent mapping is used to initialize the population, and the expression of the improved tent mapping is specifically as follows: In the above formula, t represents the t-th iteration, β represents the chaos parameter, x(t) represents the chaos mapping function of the t-th iteration, x(t+1) represents the chaos mapping function of the t+1 iteration, N represents the number of particles in the sequence, and rand(1) / N represents the random variable.

3. According to claim 2, a sliding bearing fault detection method based on IGWO-VMD-CNN is characterized in that: The expression of the nonlinear convergence operator is specifically as follows: a=1+sin(π / 2+π×iter / Max_iter) In the above formula, a represents the nonlinear convergence operator, iter represents the current number of iterations, and Max_iter represents the maximum number of iterations.

4. The sliding bearing fault detection method based on IGWO-VMD-CNN according to claim 3 is characterized in that: The improved wolf position update mechanism includes improving the position update of the Alpha wolf and the population position update based on the adaptive hunting weight coefficient, wherein: The expression for improving the position update of the Alpha wolf is specifically as follows: X1(t+1)=X α (t)-a1×D α (t) In the above formula, C1 is a random number between [0,1], X α (t) represents the position vector of Alpha wolf in the tth iteration, X(t) represents the position of the current population, and D α (t) represents the distance between the other individuals in the group and the Alpha wolf at this time, X1(t+1) represents the updated position vector of Alpha, and rand(1) represents a random variable; The expression of population position update based on the adaptive hunting weight coefficient is specifically as follows: X(t+1)=θ α ×X1+θ β ×X2+θ δ ×X3 In the above formula, α, β, δ represent Alpha wolf, Beta wolf and Delta wolf respectively, ω i represents the inertia coefficient of each wolf, f i represents the fitness function value corresponding to each wolf, θ i Indicates the weight corresponding to each wolf.

5. The sliding bearing fault detection method based on IGWO-VMD-CNN according to claim 4 is characterized in that: The step of reconstructing the acoustic emission signal of the sliding bearing based on the improved Grey Wolf optimization algorithm to obtain the reconstructed acoustic emission signal of the sliding bearing specifically includes: Based on the improved grey wolf optimization algorithm, the optimal number of decomposition layers and the optimal penalty factor of the variational mode decomposition algorithm are determined, and the variational mode decomposition algorithm with optimized parameters is obtained; The acoustic emission signal of the sliding bearing is decomposed by the variational mode decomposition algorithm after parameter optimization to obtain several IMF components. The sample entropy function is used as the fitness function, and the IMF component corresponding to the minimum sample entropy among several IMF components is selected based on the principle of minimum fitness function; Based on the IMF component corresponding to the minimum sample entropy, the acoustic emission signal of the sliding bearing is reconstructed to obtain the reconstructed acoustic emission signal of the sliding bearing.

6. The sliding bearing fault detection method based on IGWO-VMD-CNN according to claim 5 is characterized in that: The step of determining the optimal number of decomposition layers and the optimal penalty factor of the variational mode decomposition algorithm based on the improved grey wolf optimization algorithm to obtain the variational mode decomposition algorithm after parameter optimization specifically includes: Initialize the values ​​of various parameters α, β, δ in the improved gray wolf optimization algorithm and generate the positions of individual wolves in the wolf pack based on the improved tent mapping; Define the sample entropy function as the fitness function; The individual position in the wolf pack corresponds to the parameter combination [K, α] of the variational mode decomposition algorithm. The variational mode decomposition algorithm is decomposed according to [K, α], and the individual fitness value of each wolf in the wolf pack is calculated according to the fitness function; Based on the improved position update method of Alpha wolf and the population position update method of adaptive hunting weight coefficient, α, β, and δ are updated to recalculate the fitness value of each wolf; Determine whether the maximum iteration is reached. If not, return to the individual fitness value calculation step of the wolf to continue iterating. If reached, output the parameter combination [K, α] corresponding to the optimal fitness value to obtain the variational mode decomposition algorithm after parameter optimization.

7. The sliding bearing fault detection method based on IGWO-VMD-CNN according to claim 6 is characterized in that: The step of decomposing the acoustic emission signal of the sliding bearing by using the variational mode decomposition algorithm after parameter optimization to obtain a plurality of IMF components specifically includes: Based on the variational mode decomposition algorithm after parameter optimization, the optimal parameter combination [K, α] is used as the selected mode number K and penalty parameter α to decompose the acoustic emission signal of the sliding bearing; The decomposition is stopped until the decomposition iteration conditions of the variational mode decomposition algorithm are met, and the intrinsic mode functions and residual vectors of K modes, that is, several IMF components, are obtained.

8. The sliding bearing fault detection method based on IGWO-VMD-CNN according to claim 7 is characterized in that: The step of constructing a sliding bearing fault detection neural network, performing fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtaining the sliding bearing fault detection result specifically includes: Label the reconstructed acoustic emission signal according to the fault type; Constructing a sliding bearing fault detection neural network, wherein the sliding bearing fault detection neural network includes an input layer, a feature extraction layer and a classification layer; The reconstructed acoustic emission signal is input into the feature extraction layer through the input layer to extract features and obtain the features of the reconstructed acoustic emission signal; The reconstructed acoustic emission signal features are input into the classification layer, and the classification output is realized through the Softmax regression classifier to obtain the sliding bearing fault detection result.

9. A sliding bearing fault detection system based on IGWO-VMD-CNN, characterized in that: Includes the following modules: An acquisition module, used for acquiring an acoustic emission signal of a sliding bearing, wherein the acoustic emission signal of the sliding bearing includes an acoustic emission signal of a sliding bearing without fault and an acoustic emission signal of a sliding bearing with fault; Building a module to introduce improved tent mapping, nonlinear convergence operator and improved wolf position update mechanism to build an improved grey wolf optimization algorithm (IGWO); A reconstruction module is used to reconstruct the acoustic emission signal of the sliding bearing based on the improved Grey Wolf optimization algorithm combined with the variational mode decomposition (VMD) algorithm to obtain the reconstructed acoustic emission signal of the sliding bearing; The detection module is used to construct a sliding bearing fault detection neural network (CNN), perform fault detection and identification processing on the reconstructed sliding bearing acoustic emission signal, and obtain the sliding bearing fault detection result.

Images