A Fault Diagnosis Method Based on Acoustic Signals of Idler Bearings in Belt Conveyors

Abstract

This invention discloses a fault diagnosis method based on acoustic signals of idler roller bearings in belt conveyors. The method includes: using the Raccoon Optimization Algorithm to adaptively optimize key parameters in a Variational Mode Decomposition (VM) algorithm to obtain a globally optimal solution; feeding this globally optimal solution back into the VM algorithm; using this algorithm to adaptively decompose the acquired acoustic signals of the idler roller bearings to obtain an Integrated Mode Factor (IMF); using a comprehensive index to filter the components in the IMF to obtain the optimal IMF components and calculating their time-domain eigenvalues ​​to form a denoised signal feature vector; using the Raccoon Optimization Algorithm to optimize the hyperparameters of a convolutional neural network to obtain an optimized convolutional neural network fault diagnosis model; and inputting the denoised signal feature vector into this model for fault classification and diagnosis to obtain the diagnostic results. This invention can adaptively decompose the acoustic signals of idler roller bearings and extract fault features, effectively solving problems such as mode aliasing.

Description

Technical Field

[0001] This invention relates to the field of bearing fault diagnosis technology, and specifically to a fault diagnosis method based on acoustic signals of belt conveyor idler bearings. Background Technology

[0002] Belt conveyors are crucial continuous conveying equipment in coal mining. Idler rollers are the most frequently used rotating components on mining belt conveyors, and bearings are a vital part of these rollers, their operating status directly affecting the equipment's performance, efficiency, and lifespan. However, current fault diagnosis for idler roller bearings mostly relies on vibration signals, requiring the deployment of numerous sensors to acquire vibration data, which is time-consuming and labor-intensive. Therefore, utilizing the sound signals generated during operation for fault diagnosis can not only solve non-contact, long-distance fault diagnosis but also reduce the probability of mechanical equipment accidents, providing reliable decision support for future equipment maintenance plans, thus possessing significant practical value. The non-stationary, non-linear, and easily affected by environmental noise characteristics of rolling bearing sound signals present challenges for fault information extraction. Dragomiretskiy's VMD (variational mode decomposition) can decompose an input signal into a series of intrinsic mode functions with finite bandwidth and a certain frequency. However, the superior performance of VMD depends on appropriate parameter selection. An inappropriate preset value for the number of modes K can lead to over-decomposition or under-decomposition. Zhao et al. introduced improved integrated EMD (Empirical Mode Decomposition) and MOMOEDA (Multipoint Optimal Minimum Entropy Deconvolution Adjusted) with adaptive noise to reduce strong noise interference and extract defect features from rolling bearing signals. However, MOMEDA also faces parameter selection issues; determining the parameter period T and filter length L is crucial. Inappropriate parameter settings significantly impact signal decomposition and diagnostic classification results. The quality of these parameter values ​​directly affects the stability and accuracy of the diagnostic model. To address this issue, a fault diagnosis method based on the acoustic signals of belt conveyor roller bearings is proposed. Summary of the Invention

[0003] The purpose of this invention is to provide a fault diagnosis method based on acoustic signals of belt conveyor idler bearings, which can optimize key parameters in VMD and hyperparameters in convolutional neural networks, thereby improving the overall diagnostic accuracy and efficiency.

[0004] The present invention adopts the following technical solution:

[0005] A fault diagnosis method based on acoustic signals of belt conveyor idler roller bearings includes the following steps:

[0006] S1. Acquire acoustic signals from idler roller bearings using a sound sensor;

[0007] S2. Use COA (Coati Optimization Algorithm) to adaptively optimize the key parameters in the VMD algorithm to obtain the global optimal solution; the key parameters include the number of modes K and the penalty factor α.

[0008] S3. Feed back the key parameters of the optimization process to the VMD algorithm, and use the algorithm to perform adaptive signal decomposition on the acoustic signal in step S1 to obtain the corresponding IMF (Intrinsic Mode Function).

[0009] S4. Use comprehensive indicators to screen the components in the IMF and select the component with the largest comprehensive indicator as the optimal IMF component.

[0010] S5. Calculate the time-domain eigenvalues ​​of the optimal IMF component to form the signal feature vector after noise reduction.

[0011] S6. Use COA to optimize the hyperparameters of the convolutional neural network to obtain an optimized CNN (Convolutional Neural Networks) fault diagnosis model.

[0012] S7. Input the noise-reduced signal feature vector into the optimized CNN fault diagnosis model for fault classification and diagnosis, and obtain the diagnosis results.

[0013] Furthermore, in step S1, the sound sensor collects the acoustic signals of the idler bearings of the belt conveyor when it is unloaded under normal and fault conditions.

[0014] Furthermore, in step S2, the adaptive global optimization of key parameters includes the following:

[0015] S201. Set the parameters of COA, including population size, maximum number of iterations, and optimization upper and lower bounds.

[0016] S202. Initialize the population and calculate the fitness values ​​of key parameters in the VMD algorithm.

[0017] S203. Update the locations of all raccoons and iguanas according to the basic principles of COA.

[0018] S204. Determine whether the key parameters in the VMD algorithm exceed the set boundaries; if they do, return to step S203 and repeat the operation; if they do not exceed the boundaries, proceed to step S205.

[0019] S205. Correct the key parameters and fitness values ​​in the VMD algorithm.

[0020] S206. Repeat steps S202-S205 until the termination criterion is met, and the global optimal solution [K,α] is obtained.

[0021] Furthermore, in step S4, the comprehensive indicators include cross-correlation coefficient, permutation entropy, and mutual information entropy, with the following specific expressions:

[0022]

[0023]

[0024] MI(x i ,y i )=H(y i )-H(y i |x i )

[0025]

[0026]

[0027] Where CC, PE, MI, mi, and ci represent the cross-correlation coefficient, permutation entropy, mutual information entropy, normalized mutual information entropy, and composite index, respectively, and x i This represents the IMF component of the i-th sound signal. y represents the average of all IMF components. i This represents the i-th sound signal. p represents the average value of all sound signals, where N represents the total number of sound signals. j Let H(y) represent the probability of the j-th arrangement occurring in the k-th modal component of the i-th sound signal, where k = 1, 2, ..., n, and n represents the total number of modal components. Let x[n] represent the original sound signal. i ) represents signal y i The entropy, H(y) i |x i ) indicates that x is known i time y i The conditional entropy, H(x) i ) represents signal x i The entropy.

[0028] Furthermore, in step S5, there are nine types of time-domain feature values, namely, mean, peak-to-peak value, standard deviation, root mean square, kurtosis, peak factor, impulse factor, waveform factor, and margin factor.

[0029] Furthermore, in step S6, the hyperparameters of the convolutional neural network include the learning rate, hidden layer nodes, and regularization coefficient. Optimizing these hyperparameters includes the following:

[0030] S601. Set the parameters of COA, including population size, maximum number of iterations, and optimization upper and lower bounds.

[0031] S602. Initialize the population and calculate the fitness values ​​of the learning rate, hidden layer nodes, and regularization coefficients in the convolutional neural network.

[0032] S603. Update the locations of all raccoons and iguanas according to the basic principles of COA.

[0033] S604. Determine whether each target exceeds the boundary. If it does, return to step S603 and repeat the operation; if it does not exceed the boundary, proceed to step S605.

[0034] S605, modified learning rate, hidden layer nodes, regularization coefficients, and their fitness values.

[0035] S606. Repeat steps S602-S605 until the termination criterion is met to obtain the globally optimal learning rate, hidden layer nodes, and regularization coefficients.

[0036] S607. The optimized CNN fault diagnosis model includes a folded layer, two convolutional layers, two activation layers, and a defolded layer connected in sequence.

[0037] Furthermore, in step S7, the diagnostic results obtained include the following:

[0038] The denoised signal feature vectors are divided into training and validation sets in a 7:3 ratio. The training set is input into the optimized CNN fault diagnosis model for training until the accuracy reaches 100%. The validation set is input into the trained model to obtain the fault classification and diagnosis results, which are then output as a confusion matrix.

[0039] Furthermore, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the fault diagnosis method based on the acoustic signal of the belt conveyor roller bearing.

[0040] Furthermore, the present invention also proposes a computer-readable storage medium storing a computer program, which is executed by a processor to perform the fault diagnosis method based on acoustic signals of belt conveyor roller bearings.

[0041] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0042] 1. This invention achieves remote, non-contact fault diagnosis of belt conveyor roller bearings by noise reduction processing of sound signals, which solves the problem of needing to deploy a large number of vibration sensors and provides an effective new approach for remote, non-contact diagnosis of belt conveyor rollers.

[0043] 2. The variational mode decomposition used in this invention has good noise robustness and the ability to effectively suppress mode mixing and endpoint effects, which can better extract fault information.

[0044] 3. The combination of parameters for variational mode decomposition in this invention has a significant impact on the decomposition effect; at the same time, the selection of the optimal IMF component is not the same for different sound signals.

[0045] 4. This invention uses the Raccoon Optimization Algorithm to globally optimize the key parameters of variational mode decomposition. It can not only effectively determine the optimal parameter combination, but also adaptively decompose and extract different sound signals. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the overall implementation of the present invention.

[0047] Figure 2 This is a flowchart of the raccoon algorithm optimization process of the present invention.

[0048] Figure 3 This is a flowchart of the key parameter optimization process of this invention.

[0049] Figure 4 This is a flowchart of the hyperparameter optimization process of the present invention.

[0050] Figure 5 This is the spectrum diagram after adaptive noise reduction according to an embodiment of the present invention.

[0051] Figure 6 This is an unoptimized confusion matrix diagram according to an embodiment of the present invention.

[0052] Figure 7 This is the confusion matrix diagram of COA-VMD-CNN in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0054] To achieve the above objectives, this invention proposes a fault diagnosis method based on acoustic signals of belt conveyor idler bearings, such as... Figure 1 As shown, the specific steps are as follows:

[0055] S1. Acoustic signals of the idler bearings of the belt conveyor under no-load conditions are collected using sound sensors under normal and fault conditions respectively; the fault condition refers to the state in which iron filings enter the bearing and cause wear and damage to the inner and outer rings.

[0056] S2, such as Figure 2 As shown, COA (Coati Optimization Algorithm) is a swarm intelligence optimization algorithm that simulates the hunting behavior of raccoons to find global optimization. It combines the strategies of raccoons in the hunting attack and escape phases. In this algorithm, the raccoon's global search is divided into two main types: hunting attack (exploration phase) and escape from predation (development phase). The algorithm finds the optimal parameters through these behaviors and abstracts them into mathematical language.

[0057] (1) Initialize the population, the specific formula is:

[0058]

[0059] Where, x d,z Represents an individual; Indicates the search for the optimal lower boundary; represents the upper boundary of the search; r represents a random number between [0,1].

[0060] (2) Exploration Phase

[0061] The formula for searching the global location is as follows:

[0062]

[0063] in, G represents the new position of the d-th raccoon in the z-th dimension; z Let represent the iguana's position in the z-th dimension, i.e., the position of the best member; I represents a number randomly selected from the set {1,2}; M represents the total number of raccoons; Indicates no more than The largest integer; m represents the total number of decision variables.

[0064] Based on the iguana's random position in the search space after it falls, the raccoon moves, using the following formula:

[0065]

[0066]

[0067] in, This represents the iguana's position on the ground in the z-th dimension. F represents the objective function value after the iguana lands on the ground in the z-th dimension. d,z Let represent the objective function value of the d-th raccoon in the z-th dimension.

[0068] If the updated individual is better, then update the current individual; otherwise, leave it as is. The specific formula is as follows:

[0069]

[0070] in, F represents the objective function value of the d-th raccoon at its new location; d Let X represent the objective function value of the d-th raccoon at its previous position. d This indicates the position of the d-th raccoon. This indicates the new location of the d-th raccoon.

[0071] (3) Development stage

[0072] Generate a random location near each raccoon's location, using the following formula:

[0073]

[0074]

[0075]

[0076] in, This represents the local lower bound of the z-th decision variable; To represent the local upper bound of the z-th decision variable, q represents the iteration number, q = 1, 2, ..., Q, where Q represents the maximum iteration number. This represents the new position of the d-th raccoon in the z-th dimension.

[0077] If the updated individual is better, then update the current individual; otherwise, leave it as is. The specific formula is as follows:

[0078]

[0079] The Coding Analysis (COA) algorithm is used to adaptively optimize the key parameters of the Variational Mode Decomposition (VMD) algorithm to obtain the global optimum. The key parameters include the number of modes K and the penalty factor α. Figure 3 As shown, the specific content is as follows:

[0080] S201. Set the parameters of COA, including population size, maximum number of iterations, and optimization upper and lower bounds.

[0081] S202. Initialize the population and calculate the fitness values ​​of key parameters in the VMD algorithm.

[0082] S203. Update the locations of all raccoons and iguanas according to the basic principles of COA.

[0083] S204. Determine whether the key parameters in the VMD algorithm exceed the set boundaries; if they do, return to step S203 and repeat the operation; if they do not exceed the boundaries, proceed to step S205.

[0084] S205. Correct the key parameters and fitness values ​​in the VMD algorithm.

[0085] S206. Repeat steps S202-S205 until the termination criterion is met, and the global optimal solution [K,α] is obtained.

[0086] In this embodiment, the population size of COA is set to 50, the maximum number of iterations is set to 500, the upper and lower bounds of K are set to [3,12], and the upper and lower bounds of α are set to [100,2500].

[0087] S3. Feed back the key parameters from the optimization process to the VMD algorithm. Use this algorithm to perform adaptive signal decomposition on the acoustic signal from step S1. VMD mainly employs non-recursive techniques to construct and solve a finite variational problem, decomposing the complex initial signal into several IMF component sequences from high to low frequency, thus obtaining the corresponding IMFs (Intrinsic Mode Functions). Specifically:

[0088] S301. Construct a variational problem with the constraint that the sum of the decomposed components equals the original signal. The constraint expression is as follows:

[0089]

[0090]

[0091] Where K represents the number of decomposition modes that need to be determined in advance; {μ k}{ω k} represent the k-th modal component and center frequency after decomposition, respectively; δ(t) represents the Dirac function; * represents the convolution operator; f(t) represents the original signal.

[0092] S302. Solving the variational problem: By introducing a quadratic penalty factor and the Lagrange multiplication operator, the constrained variational problem is transformed into an unconstrained variational problem, yielding the augmented Lagrange expression, specifically:

[0093]

[0094] in, To ensure the accuracy of signal reconstruction, λ(t) represents the Lagrange multiplier operator.

[0095] S303, Update {μ k The modal components are obtained from (ω)}, and the specific formula is as follows:

[0096]

[0097] Update center frequency {ω k The specific formula is as follows:

[0098]

[0099] The Lagrange multiplication operator is updated to have the following formula:

[0100]

[0101] Where γ represents noise tolerance; They represent the (n+1)th modal components respectively. Mode function μ i Fourier transforms of f(t), λ(t), and f(t).

[0102] S304. Repeat steps S301-S303 until the following condition is met, then stop:

[0103]

[0104] Where φ represents the convergence accuracy, which is 1×10⁻⁶. -7 .

[0105] S4. Use composite indicators to filter the components in the IMF, and select the component with the largest composite indicator as the optimal IMF component; the specific content is as follows:

[0106] The comprehensive indicators include cross-correlation coefficient, permutation entropy, and mutual information entropy. The Cross-correlation Coefficient (CC) measures the strength of the linear relationship between two variables; a higher CC value indicates a higher similarity between the IMF component and the original signal. Permutation Entropy (PE) measures the degree of random variation in a signal's time series; a higher PE value indicates a more random and complex signal time series, while a lower PE value indicates a more regular and less complex signal series. Mutual Information (MI) in information theory primarily represents the correlation between two events, indicating that they are less susceptible to external interference. A higher MI value indicates a stronger correlation between the two events. For IMF components, the richer the original signal feature information they contain, the higher the MI value.

[0107] The specific formula for calculating the comprehensive index is as follows:

[0108]

[0109]

[0110] MI(x i ,y i )=H(y i )-H(y i |x i )

[0111]

[0112]

[0113] Where mi and ci represent the normalized and composite indices of mutual information entropy, respectively, and x i This represents the IMF component of the i-th sound signal. y represents the average of all IMF components. i This represents the i-th sound signal. p represents the average value of all sound signals, where N represents the total number of sound signals. j Let H(y) represent the probability of the j-th arrangement occurring in the k-th modal component of the i-th sound signal, where k = 1, 2, ..., n, n represents the total number of modal components, and j = 1, 2, ..., J, J represents the total number of arrangements. i ) represents y i The entropy, H(y) i |x i ) indicates that x is known i time y i The conditional entropy, H(x) i ) represents xi The entropy.

[0114] S5. Calculate the time-domain eigenvalues ​​of the optimal IMF component to form the signal feature vector after noise reduction. There are nine time-domain eigenvalues: mean, peak-to-peak value, standard deviation, root mean square, kurtosis, peak factor, impulse factor, waveform factor, and margin factor. Their specific expressions are as follows:

[0115]

[0116] Peak-to-peak value = x max -x min .

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124] S6. CNN (Convolutional Neural Networks) is a deep learning model similar to an artificial neural network, the multilayer perceptron. It has powerful feature extraction capabilities. Its network consists of an input layer, a hidden layer, and an output layer. The hidden layer includes convolutional layers, activation layers, pooling layers, and fully connected layers.

[0125] Convolutional layers are used to extract features. They generate output features by convolving local regions of the input with convolutional kernels. This paper increases the size of the first convolutional kernel to obtain a larger receptive field and extract short-term features. The remaining layers use small convolutional kernels, which helps prevent overfitting. Its operational expression is:

[0126]

[0127] Where x represents the input; This represents the feature output after convolution by the o-th convolutional kernel in the l-th layer; These represent the weights and biases of the o-th convolutional kernel in the l-th layer, respectively.

[0128] The purpose of the activation layer is to enable the model to learn multiple layers of nonlinear mappings by adding a nonlinear activation function. The ReLU function is chosen as the activation function. Its operational expression is:

[0129]

[0130] Pooling layers downsample the data after convolutional layers to extract key features, reducing model size and effectively avoiding overfitting.

[0131] The fully connected layer is responsible for integrating the features extracted by the model and feeding them into the classifier to output the classification result.

[0132] The learning rate is a parameter that controls the step size for updating backpropagation error. An excessively large learning rate may cause the network to fail to converge, while an excessively small learning rate will result in slow convergence. The number of hidden layers directly affects the expressive power of the neural network. Increasing the number of hidden layers can improve the network's complexity, enabling it to learn more complex patterns and features. However, the more layers there are, the more complex the training process becomes, and the greater the computational cost, leading to longer fault diagnosis times. The choice of regularization coefficient has a significant impact on the model's generalization ability. A smaller regularization coefficient will result in a weaker regularization effect, which may be insufficient to prevent overfitting. A larger regularization coefficient will result in overly strong regularization, which may affect the model's fitting ability. Therefore, choosing an appropriate regularization coefficient requires finding a balance between the model's generalization ability and training error.

[0133] The hyperparameters of a convolutional neural network include the learning rate, hidden layer nodes, and regularization coefficients. The Coding Analysis (COA) is used to optimize these hyperparameters, resulting in an optimized CNN fault diagnosis model. Figure 4 As shown, the specific content is as follows:

[0134] S601. Set the parameters of COA, including population size, maximum number of iterations, and optimization upper and lower bounds.

[0135] S602. Initialize the population and calculate the fitness values ​​of the learning rate, hidden layer nodes, and regularization coefficients in the convolutional neural network.

[0136] S603. Update the locations of all raccoons and iguanas according to the basic principles of COA.

[0137] S604. Determine whether each target exceeds the boundary. If it does, return to step S603 and repeat the operation; if it does not exceed the boundary, proceed to step S605.

[0138] S605, modified learning rate, hidden layer nodes, regularization coefficients, and their fitness values.

[0139] S606. Repeat steps S602-S605 until the termination criterion is met to obtain the globally optimal learning rate, hidden layer nodes, and regularization coefficients.

[0140] S607. The optimized CNN fault diagnosis model includes a folded layer, two convolutional layers, two activation layers, and a defolded layer connected in sequence.

[0141] In this embodiment, the lower bounds of the learning rate, hidden layer nodes, and regularization coefficient are set to 1e-3, 10, and 1e-4, respectively, and the upper bounds are set to 1e-2, 30, and 1e-1, respectively.

[0142] S7. Divide the denoised signal feature vector into a training set and a validation set in a 7:3 ratio. Input the training set into the optimized CNN fault diagnosis model for training until the accuracy reaches 100%. Then input the validation set into the trained model to obtain the fault classification and diagnosis results, and output the results as a confusion matrix.

[0143] In this embodiment, the collected idler bearing data was obtained from a laboratory-built simulation test bench for belt conveyor idler bearing failure. Three sets of idlers were set up, each spaced 1m apart, with an idler size of 250mm. The motor frequency was 1.6kW, the belt speed was 2m / s, and the sampling frequency was 48kHz. A total of 2048 sampling points were collected for the acoustic signals of the two types of idler bearings, with 1200 samples collected for each type, for a total of 2400 sets. The 2400 sets of data were numbered according to fault type: normal (1) and fault (2), and divided into training and validation sets in a 7:3 ratio. The training set for each type of acoustic signal consisted of 840 sets, and the validation set consisted of 360 sets.

[0144] Figure 5 This paper presents a schematic diagram illustrating the parameter combinations obtained from adaptive optimal mode decomposition (IMF) of two types of data: normal signals and inner / outer ring wear fault signals. Taking inner / outer ring wear fault as an example, the IMF component spectrum after optimal decomposition is shown. It can be seen that the IMF components after adaptive decomposition effectively avoid problems such as mode mixing and endpoint effects. The diagnostic results output by existing methods and the method proposed in this invention are as follows: Figure 6 and Figure 7 As shown. Figure 6 The accuracy rate was 93.5%, mainly due to inaccurate identification of faulty idlers; while Figure 7 The accuracy rate was 96.667%, and the accuracy rate for identifying faulty idlers increased, with the accuracy rates for the two types of identification being comparable.

[0145] This invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. It should be noted that when the processor executes the computer program, it corresponds to the specific steps of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0146] This invention also proposes a computer-readable storage medium storing a computer program. It should be noted that when the computer program is executed by a processor, it corresponds to the specific steps of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0147] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A fault diagnosis method based on acoustic signals of belt conveyor idler roller bearings, characterized in that, include: S1. Acquire acoustic signals from idler roller bearings using a sound sensor; S2. Use the Raccoon Optimization Algorithm to adaptively optimize the key parameters in the Variational Mode Decomposition Algorithm to obtain the global optimal solution; the key parameters include the number of modes K and the penalty factor α. S3. Feed back the key parameters of the optimization process to the variational mode decomposition algorithm, and use the algorithm to perform adaptive signal decomposition on the acoustic signal in step S1 to obtain the corresponding IMF. S4. Use composite indicators to filter the components in the IMF, and select the component with the largest composite indicator as the optimal IMF component. The composite indicators include cross-correlation coefficient, permutation entropy, and mutual information entropy, with the specific expressions as follows: MI(x i ,and i )=H(y i )-H(y i |x i ) Where CC, PE, MI, mi, and ci represent the cross-correlation coefficient, permutation entropy, mutual information entropy, normalized mutual information entropy, and composite index, respectively, and x i This represents the IMF component of the i-th sound signal. y represents the average of all IMF components. i This represents the i-th sound signal. p represents the average value of all sound signals, where N represents the total number of sound signals. j Let H(y) represent the probability of the j-th arrangement occurring in the k-th modal component of the i-th sound signal, where k = 1, 2, ..., n, and n represents the total number of modal components. Let x[n] represent the original sound signal. i ) represents signal y i The entropy, H(y) i |x i ) indicates that x is known i time y i The conditional entropy, H(x) i ) represents signal x i Entropy; S5. Calculate the time-domain eigenvalues ​​of the optimal IMF component to form the signal feature vector after noise reduction; S6. Use the Raccoon Optimization Algorithm to optimize the hyperparameters of the convolutional neural network to obtain an optimized convolutional neural network fault diagnosis model. The hyperparameters of the convolutional neural network include the learning rate, hidden layer nodes, and regularization coefficient. Optimizing these hyperparameters includes the following: S601. Set the parameters of the raccoon optimization algorithm, including population size, maximum number of iterations, and upper and lower bounds of optimization. S602. Initialize the population and calculate the fitness values ​​of the learning rate, hidden layer nodes, and regularization coefficient in the convolutional neural network. S603. Update the positions of all raccoons and iguanas according to the basic principles of the raccoon optimization algorithm; S604. Determine whether each target exceeds the boundary. If it does, return to step S603 and repeat the operation. If it does not exceed the boundary, proceed to step S605. S605, adjusting the learning rate, hidden layer nodes, regularization coefficients, and their fitness values; S606. Repeat steps S602-S605 until the termination criterion is met to obtain the globally optimal learning rate, hidden layer nodes, and regularization coefficients. S607. The optimized convolutional neural network fault diagnosis model includes a folded layer, two convolutional layers, two activation layers, and a defolded layer connected in sequence. S7. Input the noise-reduced signal feature vector into the optimized convolutional neural network fault diagnosis model to perform fault classification and diagnosis, and obtain the diagnosis results.

2. The fault diagnosis method based on acoustic signals of belt conveyor idler bearings according to claim 1, characterized in that, In step S1, the sound sensor collects the acoustic signals of the idler bearings of the belt conveyor when it is unloaded under normal and fault conditions.

3. The fault diagnosis method based on acoustic signals of belt conveyor idler roller bearings according to claim 1, characterized in that, In step S2, adaptive global optimization of key parameters includes the following: S201. Set the parameters of the raccoon optimization algorithm, including population size, maximum number of iterations, and upper and lower bounds for optimization. S202. Initialize the population and calculate the fitness values ​​of key parameters in the variational mode decomposition algorithm; S203. Update the positions of all raccoons and iguanas according to the basic principles of the raccoon optimization algorithm; S204. Determine whether the key parameters in the variational mode decomposition algorithm exceed the set boundaries; If the operation exceeds the limit, return to step S203 and repeat the operation; if the operation does not exceed the limit, proceed to step S205. S205. Key parameters and their fitness values ​​in the modified variational mode decomposition algorithm; S206. Repeat steps S202-S205 until the termination criterion is met, and the global optimal solution [K,α] is obtained.

4. The fault diagnosis method based on acoustic signals of belt conveyor idler roller bearings according to claim 1, characterized in that, In step S5, there are nine types of time-domain feature values, namely, mean, peak-to-peak value, standard deviation, root mean square, kurtosis, peak factor, impulse factor, waveform factor, and margin factor.

5. The fault diagnosis method based on acoustic signals of belt conveyor idler roller bearings according to claim 1, characterized in that, In step S7, the diagnostic results include the following: The denoised signal feature vectors are divided into training and validation sets in a 7:3 ratio. The training set is input into the optimized convolutional neural network fault diagnosis model for training until the accuracy reaches 100%. The validation set is then input into the trained model to obtain the fault classification and diagnosis results, which are output as a confusion matrix.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the fault diagnosis method based on acoustic signals of belt conveyor roller bearings as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing a computer program, characterized in that, The computer program, when executed by the processor, performs the fault diagnosis method based on acoustic signals of belt conveyor roller bearings as described in any one of claims 1 to 5.

Images