Bearing fault classification and identification method and device and computer equipment

Abstract

The invention relates to a bearing fault classification and identification method and device and computer equipment. An improved crayfish optimization algorithm is adopted to optimize parameters of variational mode decomposition to extract a more effective intrinsic mode function, so that the feature extraction precision is enhanced, spatial-temporal features, a time dependency relationship and attention to key features are comprehensively considered, and a more accurate and reliable solution is provided for bearing fault diagnosis. Meanwhile, based on a traditional crayfish optimization algorithm, the improved crayfish optimization algorithm is combined with gravity center reverse learning, non-linear control and sparrow warning mechanism strategies, more excellent optimization searching precision and robustness are shown, meanwhile, the convergence speed is higher, parameter optimization of variational mode decomposition is more accurate and reliable, and the method is suitable for large-scale popularization and application. Therefore, the accuracy of feature extraction is improved, and reliable data guarantee is provided for obtaining the accuracy of a fault classification result subsequently.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to a method, device and computer equipment for classifying and identifying bearing faults. Background Art

[0002] Industrial bearings are vital to the normal operation of mechanical equipment. With the rapid development of industrial automation and intelligence, it is particularly important to accurately identify and diagnose bearing faults. Traditional fault diagnosis methods, such as those based on double impact characteristics and dynamic modeling, perform well in accurately determining the fault size, but require diagnostic personnel to have a high degree of experience and high accuracy of the dynamic model, which may be a limiting factor in practical applications, especially when dealing with complex working conditions. These methods often have difficulty adapting to the complex characteristics and dynamic changes of signals.

[0003] In recent years, data-driven fault diagnosis methods have received widespread attention. These methods use deep learning and data analysis techniques to better capture the temporal and spatial characteristics of fault signals, thereby improving the effectiveness of fault diagnosis.

[0004] However, data-driven models still face challenges in optimization and feature extraction, and the accuracy and stability of fault diagnosis need to be further improved. Summary of the invention

[0005] Based on this, it is necessary to provide a bearing fault classification and identification method, device and computer equipment that can improve the accuracy and stability of fault diagnosis in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for classifying and identifying bearing faults. The method comprises:

[0007] Obtain bearing fault data when the industrial computer is running;

[0008] The bearing fault data is processed using the variational mode decomposition method to extract the intrinsic mode function;

[0009] Based on the intrinsic mode function, the time-space characteristics and time-dependency analysis of the bearing fault data are carried out, and the focus is introduced on the key features to obtain the fault classification results;

[0010] Among them, the parameters of the variational mode decomposition method are optimized using the improved crayfish optimization algorithm, which combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

[0011] In one embodiment, obtaining bearing fault data when the industrial computer is running includes:

[0012] Obtain the vibration signal of the target position on the industrial computer when the industrial computer is running;

[0013] Perform data preprocessing on the vibration signal to obtain bearing fault data.

[0014] In one embodiment, the improved crayfish optimization algorithm includes:

[0015] Introduce the center of gravity reverse learning to generate the spatial position matrix and determine the cutoff value;

[0016] Enter the iterative process; the iterative process includes: generating a first random number and a second random number, the value range of the second random number is 0 to 1; when the first random number is greater than the delimiting value and the second random number is less than 0.5, obtaining a first candidate solution according to the spatial position matrix and the exponential decay function that substitutes the second random number and the current number of iterations; when the first random number is greater than the delimiting value and the second random number is not less than 0.5, obtaining a second candidate solution according to the spatial position matrix; when the first random number is not greater than the delimiting value, obtaining a third candidate solution according to the spatial position matrix and the second random number; introducing a sparrow alert mechanism to compare the fitness value of the first candidate solution, the second candidate solution or the third candidate solution with the fitness value of the current optimal solution, and updating the current optimal solution according to the comparison result and the first candidate solution, the second candidate solution or the third candidate solution;

[0017] Iterate and update the current optimal solution until the preset number of iterations is reached, and obtain the target solution based on the current optimal solution corresponding to the last iteration.

[0018] In one embodiment, the parameters of the variational mode decomposition method are optimized using an improved crayfish optimization algorithm during the training phase, and the training steps include:

[0019] Obtain training data for multiple fault types;

[0020] For each fault type, the corresponding training data is extracted, and the parameters of the variational mode decomposition method are optimized using the improved crayfish optimization algorithm to obtain the corresponding target parameters.

[0021] In one embodiment, the time-space characteristic analysis and time-dependency analysis of the bearing fault data are performed based on the intrinsic mode function, and attention is paid to key characteristics, including:

[0022] Use CNN to analyze the spatiotemporal characteristics of intrinsic mode functions;

[0023] Use BiLSTM to capture the temporal dependency in the intrinsic mode function;

[0024] The self-attention mechanism is introduced to enhance the perception of key features in the intrinsic mode function.

[0025] In one embodiment, the method further comprises:

[0026] A fault recognition network including CNN, BiLSTM and self-attention layer is constructed, the fault recognition network is pre-trained, and the intrinsic mode function is input into the pre-trained fault recognition network to obtain the fault classification result.

[0027] In a second aspect, the present application also provides a bearing fault classification and identification device. The device comprises:

[0028] A data acquisition module is used to acquire bearing fault data when the industrial computer is running;

[0029] A feature extraction module is used to process bearing fault data using a variational mode decomposition method and extract intrinsic mode functions;

[0030] The fault classification module is used to perform spatiotemporal feature analysis and time dependency analysis on bearing fault data based on intrinsic mode functions, and introduce attention to key features to obtain fault classification results;

[0031] Among them, the parameters of the variational mode decomposition method are optimized using the improved crayfish optimization algorithm, which combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

[0032] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned bearing fault classification and identification method when executing the computer program.

[0033] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned bearing fault classification and identification method when executed by a processor.

[0034] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned bearing fault classification and identification method are implemented.

[0035] The above-mentioned bearing fault classification and identification method, device and computer equipment use an improved crayfish optimization algorithm to optimize the parameters of variational mode decomposition to extract more effective intrinsic mode functions, thereby enhancing the accuracy of feature extraction, and comprehensively considers spatiotemporal features, time dependencies and attention to key features, thereby providing a more accurate and reliable solution for bearing fault diagnosis. At the same time, the improved crayfish optimization algorithm, based on the traditional crayfish optimization algorithm, combines the center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategies, and shows better optimization accuracy and robustness, and also has a relatively good convergence speed, making the parameter optimization of variational mode decomposition more accurate and reliable, thereby improving the accuracy of feature extraction, and providing reliable data guarantee for the accuracy of subsequent fault classification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of a flow chart of a bearing fault classification and identification method in one embodiment;

[0037] Figure 2 A flowchart of an improved crayfish optimization algorithm in one embodiment;

[0038] Figure 3 A schematic diagram of the structure of an ICOA-VMD-CNN-BiLSTM-Attention model in one embodiment;

[0039] Figure 4 A schematic diagram of the structure of deploying the ICOA-VMD-CNN-BiLSTM-Attention model on an industrial computer in one embodiment;

[0040] Figure 5 Box plots of the CEC2022 benchmark function tests for six algorithms;

[0041] Figure 6 It is the algorithm convergence curve of F1 single peak and F2 multi-peak function;

[0042] Figure 7 It is the algorithm convergence curve of the composite and multi-modal function of F9 and F11;

[0043] Figure 8 This is the structural diagram of the bearing experimental device;

[0044] Fig. 9 This is the test result of CNN-BiLSTM-Attention model;

[0045] Fig.10 This is the test result of OOA-VMD-CNN-BiLSTM-Attention model;

[0046] Fig.11This is the test result of COA-VMD-CNN-BiLSTM-Attention model;

[0047] Fig.12 The test results of the ICOA-VMD-CNN-BiLSTM-Attention model are as follows;

[0048] Fig.13 The figure is a structural block diagram of a bearing fault classification and identification device in one embodiment. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The present application embodiment provides a method for classifying and identifying bearing faults, such as Figure 1 As shown, the following steps are included:

[0051] Step 102, obtaining bearing fault data when the industrial computer is running.

[0052] When the ICC (Industrial Control Computer) is running, the sensor is used to collect the data of the ICC during operation. If the bearing fails, it will be reflected in the data collected by the sensor, and the bearing failure data at this time is obtained. Since the bearing failure signal is transmitted in the form of vibration, the bearing failure data is usually a vibration signal.

[0053] Step 104, using the variational mode decomposition method to process the bearing fault data and extract the intrinsic mode function. The parameters of the variational mode decomposition method are optimized using an improved crayfish optimization algorithm, which combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

[0054] Variational Mode Decomposition (VMD) is a signal decomposition estimation method. In the process of obtaining decomposed components, this method determines the frequency center and bandwidth of each component by iteratively searching for the optimal solution of the variational model, so as to adaptively realize the frequency domain decomposition of the signal and the effective separation of each component.

[0055] The intrinsic mode function (IMF) is determined based on the characteristics of the data itself and can reflect the local characteristics and nonlinear dynamic behavior of the signal.

[0056] Crayfish Optimization Algorithm (COA) is an emerging swarm intelligence optimization algorithm, inspired by the feeding, summer heat and competition behaviors of crayfish in the natural environment. COA solves various optimization problems by simulating the social behavior and survival strategies of crayfish. The algorithm is mainly used to find the global optimal solution, especially in complex multi-peak optimization problems.

[0057] This embodiment proposes an improved crayfish optimization algorithm (ICOA) based on the existing crayfish optimization algorithm, combined with center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

[0058] Step 104 mainly completes the work of extracting features from bearing fault data through ICOA and VMD.

[0059] Step 106 , performing spatiotemporal feature analysis and time dependency analysis on the bearing fault data based on the intrinsic mode function, and introducing attention to key features to obtain fault classification results.

[0060] Spatiotemporal feature analysis mainly processes data that contains both temporal and spatial information in order to better understand and predict the behavior of dynamic systems. It is usually implemented using neural network architectures such as convolutional neural networks, recurrent neural networks, and graph neural networks.

[0061] Time dependency analysis refers to detecting and understanding the interdependencies between data points in the time dimension when processing time series data. This analysis is crucial for predicting future events, identifying patterns, and understanding the behavior of dynamic systems. Traditional statistical methods such as autocorrelation functions, frequency domain analysis methods, time domain analysis methods, and machine learning methods are usually used.

[0062] Key features refer to the characteristic variables that have a greater impact on model performance and prediction results. Identifying and prioritizing these features can help improve the predictive power of the model, reduce the risk of overfitting, and improve the interpretability of the model.

[0063] This embodiment uses an improved crayfish optimization algorithm to optimize the parameters of variational modal decomposition to extract more effective intrinsic mode functions, thereby enhancing the accuracy of feature extraction, and comprehensively considering spatiotemporal features, time dependencies, and attention to key features, thereby providing a more accurate and reliable solution for bearing fault diagnosis. At the same time, the improved crayfish optimization algorithm, based on the traditional crayfish optimization algorithm, combines the center of gravity reverse learning, nonlinear control, and sparrow warning mechanism strategies, showing better optimization accuracy and robustness, and also has a relatively good convergence speed, making the parameter optimization of variational modal decomposition more accurate and reliable, thereby improving the accuracy of feature extraction, and providing reliable data guarantee for the accuracy of subsequent fault classification results.

[0064] In one embodiment, step 102 includes: obtaining a vibration signal at a target position on the industrial computer when the industrial computer is running; performing data preprocessing on the vibration signal to obtain bearing fault data.

[0065] Install an acceleration sensor at the target position of the industrial computer to collect vibration signals at the target position. The target position can be selected as close to the bearing as possible, such as the bearing seat; multiple positions can also be selected at the same time. The acceleration sensors at each position collect data at the same time to form multiple independent data channels, providing basic data for subsequent fault analysis.

[0066] After obtaining the vibration signal, data preprocessing is performed on the vibration signal, including data segmentation using a sliding window, and finally the bearing fault data in matrix form is obtained.

[0067] In one embodiment, the improved crayfish optimization algorithm includes: introducing centroid reverse learning to generate a spatial position matrix and determining a demarcation value; entering an iterative process; the iterative process includes: generating a first random number and a second random number, the value range of the second random number is 0 to 1; when the first random number is greater than the demarcation value and the second random number is less than 0.5, obtaining a first candidate solution according to the spatial position matrix and an exponential decay function that substitutes the second random number and the current number of iterations; when the first random number is greater than the demarcation value and the second random number is not less than 0.5, obtaining a second candidate solution according to the spatial position matrix; when the first random number is not greater than the demarcation value, obtaining a third candidate solution according to the spatial position matrix and the second random number; introducing a sparrow alert mechanism to compare the fitness value of the first candidate solution, the second candidate solution or the third candidate solution with the fitness value of the current optimal solution, and updating the current optimal solution according to the comparison result and the first candidate solution, the second candidate solution or the third candidate solution; iteratively updating the current optimal solution until a preset number of iterations is reached, and obtaining the target solution according to the current optimal solution corresponding to the last iteration.

[0068] like Figure 2 The figure shows the flowchart of the improved crayfish optimization algorithm. The algorithm is as follows:

[0069] (1) Initialization of population

[0070] The population is generated in the initialization phase. In the initialization phase of the algorithm, the population is generated by a formula based on random values. Each individual has a position X in the jth dimension. i，j The initialization expression is

[0071] X i,j =lb j +(ub j -lb j )×rand (1)

[0072] Where: X i，j represents the position of the i-th individual in the j-th dimension, lb j andub j Represent the lower and upper limits of the jth dimension respectively, and rand is a random number uniformly distributed in the interval (0,1). The expression of the generated population matrix X is

[0073]

[0074] Where: N represents the population size, dim is the dimension.

[0075] Calculate the population mean, which is the center of gravity. The population mean vector X mean Represents the center of gravity of the entire population, which is obtained by calculating the average value of all individual positions in each dimension, and its expression is

[0076] X mean =mean(X i,j ) (3)

[0077] Introducing mutation, that is, reverse learning from the center of gravity. In order to promote the diversity and exploration of the initial population, the strategy of reverse learning from the center of gravity is adopted. The position X of each individual i，j The updated expression is

[0078] X i,j =X i,j +rand×(2×X mean -X i,j )×rand() (4)

[0079] X i，j That is, the elements in the spatial position matrix.

[0080] Boundary constraints are enforced. To ensure that the generated positions conform to the specified upper and lower bounds, boundary constraints are enforced, and the positions of the entire population are adjusted to prevent them from exceeding the bounds.

[0081] Calculate the fitness value to get XG and X L .X G represents the best position obtained so far through the number of iterations, X L Indicates the best position of the current population.

[0082] (2) Determine the temperature and intake stage of crayfish

[0083] The behavior of crayfish is affected by temperature changes, causing them to enter different physiological stages. According to the ICOA algorithm definition, the temperature is determined by formula (5). When the temperature exceeds 30°C (i.e. the cutoff value), crayfish tend to seek cool places to escape the heat; at a suitable temperature, they are more likely to forage. The optimal breeding temperature for crayfish is 25°C, because their foraging behavior is most active within the range of 20-30°C. Therefore, ICOA defines the temperature range as a random number between 20°C and 35°C. The expression for temp is

[0084] temp=rand×15+20 (5)

[0085] Wherein: temp represents the temperature of the environment where the crayfish is located, that is, the first random number, and rand is the second random number, and the value range is (0, 1).

[0086] (3) Summer vacation (exploration) stage

[0087] When the temperature is >30℃, it means the temperature is too high. At this time, the crayfish will choose to enter the cave to escape the heat. Calculate the average value of all crayfish in the current population on dimension j and store the result in the cave variable middle, The expression is

[0088]

[0089] The competition among crayfish for caves is a random event. When rand < 0.5, it means that there are no other crayfish competing for caves, and the crayfish will directly enter the cave to escape the heat. At this time, the crayfish will enter the cave to escape the heat. The expression is

[0090]

[0091] Where: is the new position of crayfish i in dimension j (i.e. the first candidate solution), is the current position of crayfish i on dimension j, median(X i,j ) is the median of the population in each dimension, and the expression of the control parameter C is

[0092]

[0093] Where: t represents the current iteration number, T represents the maximum number of iterations. The nonlinear control introduced by the ICOA algorithm is mainly reflected in the position update process, especially the exponential decay function used in calculating the control parameter C. It is linear. The nonlinear control method makes the control parameters change according to the number of iterations and random factors during the iteration process, thereby realizing dynamic adjustment of the population. This nonlinear characteristic enhances the flexibility and adaptability of the algorithm and optimizes the search process.

[0094] In the summer stage, the crayfish's goal is to approach the cave, which represents the best solution. This encourages the crayfish to move closer to the cave, making the individual closer to the optimal solution, enhancing the utilization of ICOA and accelerating the convergence of the algorithm.

[0095] (4) Competition (Development) Stage

[0096] When the temperature is >30℃ and rand≥0.5, it means that other crayfish are also interested in the cave. At this time, they will fight for the cave. Crayfish fight for the cave The expression is

[0097]

[0098] Where: is the new position of crayfish i in dimension j (i.e., the second candidate solution). is the current position of crayfish i on dimension j. is the position of a crayfish randomly selected from X on dimension j. bestpos j is the best position on dimension j.

[0099] In this stage, crayfish compete with each other. By adjusting their positions, the search range of ICOA is expanded, and the exploration ability of the algorithm is enhanced.

[0100] (5) Foraging (mining) stage

[0101] When the temperature is ≤30℃, the temperature is suitable for crayfish breeding. During the feeding stage, the expression of the new position of the crayfish (i.e., the third candidate solution) is

[0102]

[0103] Through the feeding stage, ICOA will approach the optimal solution, which enhances the utilization of the algorithm and gives it better convergence ability.

[0104] (6) Boundary constraints, sparrow warning mechanism and updated fitness value

[0105] First, boundary constraints are used to ensure the legitimacy of candidate solutions within the search space. After each candidate solution is updated, the value of the solution is limited to the predefined upper and lower bounds through the minimum and maximum functions. Whether the upper and lower bounds are unified or different for each dimension, this operation can prevent the solution from exceeding the search space, thereby ensuring the stability of the algorithm.

[0106] Secondly, the sparrow warning mechanism is introduced to determine the behavior of the candidate solution by comparing the fitness of the candidate solution with the current optimal solution. i Greater than the fitness F of the current optimal solution Li ), the alert update strategy is adopted to make it move closer to the optimal solution position to improve its fitness, simulating the crayfish at the edge of the group moving quickly to the core area when sensing danger. When the candidate solution is in an advantageous position (i.e., its fitness is less than that of the current optimal solution), the optimization update is continued, and random disturbances are used to prompt it to further improve its position, simulating the crayfish in the middle of the population randomly moving to get closer to other crayfish.

[0107] When the candidate solution is at a disadvantage, the alert update position is:

[0108]

[0109] When the candidate solution is in an advantageous position, continue to optimize as follows:

[0110]

[0111] Where: and are the positions of the i-th crayfish in the j-th dimension and the t+1-th iteration respectively; is the current optimal solution of the j-th dimension crayfish at the t-th iteration; r is a random number in the interval [0,1]. worse is the vector of inferior positions, containing the position data of the idx-th candidate solution. max is the fitness vector F i idx is the maximum value among the candidate solutions. i The index of the maximum value in , indicating the position of the maximum value.

[0112] Finally, the fitness value update process ensures that the algorithm continues to converge to the optimal solution. In each iteration, the fitness value of the new candidate solution is compared with the current optimal value. If the new fitness value is better, the individual position and the global optimal solution record are updated. Through repeated iterations, the algorithm gradually approaches the global optimal solution. These improvements enhance the search ability and convergence performance of the crayfish optimization algorithm, making it perform better when dealing with complex optimization problems.

[0113] In this embodiment, the sparrow alert mechanism is intended to enhance the algorithm's exploration capabilities and avoid falling into local optimality. In each iteration, some crayfish act as "alerts", exploring new search areas by disturbing and updating candidate solutions, and ensuring that the solution does not cross the boundary. In the fitness update step, the positions of all crayfish are evaluated by the fitness value, and the current optimal solution is selected from them, compared with the optimal solution of the previous round, and if it is better, the global optimal solution is updated. This process is repeated in each round of iteration, and each iteration generates a current optimal solution, gradually approaching the global optimal solution until the stop condition (such as reaching the maximum number of iterations or convergence accuracy) is met. Finally, the global optimal solution, as the output of the algorithm, represents the optimal solution found in multiple iterations. Through the exploration of the alert mechanism and the optimization of fitness update, a balance between exploration and utilization is achieved, and the global search capability and convergence speed are improved.

[0114] The main differences between the improved crayfish optimization algorithm and the traditional crayfish optimization algorithm are as follows:

[0115] 1. Improvement and simplification of the feeding phase

[0116] The original crayfish optimization algorithm adopts a more complex update mechanism in the feeding stage, including dynamically calculating the feeding strategy according to the individual fitness value and ambient temperature, and executing different update strategies according to the size of the food. This increases the computational complexity of the algorithm, resulting in higher computational overhead and reducing the convergence speed.

[0117] The improved crayfish algorithm is optimized in the feeding stage and adopts a more efficient linear update mechanism (see formula (10)). The individual position is updated by the difference between the individual and the average value of the population, avoiding complex mathematical operations and conditional judgments. At the same time, the improved crayfish algorithm removes the dependence on exponential decay and trigonometric functions, making the feeding stage more concise and clear. Compared with the traditional crayfish algorithm, the improved crayfish algorithm significantly reduces the algorithm complexity, improves the computational efficiency and convergence speed, and enhances its adaptability in high-dimensional optimization problems.

[0118] 2. Changes in search strategies

[0119] The improved crayfish optimization algorithm introduces the individual position mean and population median, which enhances the diversity of population search strategies.

[0120] 3. Introduction of Sparrow Sentinel Mechanism

[0121] The improved crayfish optimization algorithm integrates the sparrow warning mechanism, further enhancing the exploration ability of the population in the early stage. When the individual fitness does not exceed the global optimum, random perturbations are used to strengthen the exploration between the individual and the optimal solution, improving the ability to jump out of the local optimum.

[0122] 4. Introduction of nonlinear control

[0123] The control parameter C is obtained using an exponential decay function, which allows the algorithm to more flexibly adjust the search intensity at different stages.

[0124] 5. Improvements in initialization and boundary processing

[0125] In both the original and improved crayfish optimization algorithms, a boundary checking mechanism is set to ensure that individuals do not exceed the allowed search range. However, the improved crayfish optimization algorithm introduces a centroid-based reverse learning strategy in the initialization phase, that is, by calculating the centroid (mean) of the population, a new individual position is generated. Compared with the random initialization of the original algorithm, this initialization method can make the population more evenly distributed in the search space, avoid the local aggregation problem of the initial solution, and thus enhance the global search ability of the algorithm. This strategy not only optimizes the initialization process, but also improves the robustness and search efficiency of the algorithm in different problem scenarios.

[0126] 6. Globally optimal update method

[0127] The original crayfish optimization algorithm updates the global optimal solution and the current best individual position in rounds, while the improved crayfish optimization algorithm updates the best individual position in two stages (main search and sparrow mechanism) and strengthens the global search in each iteration.

[0128] In summary, the improved crayfish optimization algorithm enhances the global exploration ability and local development ability of the crayfish algorithm, effectively avoids falling into local optimality, and makes the algorithm more stable in different problem scenarios.

[0129] In one embodiment, the parameters of the variational modal decomposition method are optimized using an improved crayfish optimization algorithm during a training phase, and the training steps include: obtaining training data for a variety of fault types; for each fault type, extracting corresponding training data, and optimizing the parameters of the variational modal decomposition method using an improved crayfish optimization algorithm to obtain corresponding target parameters.

[0130] The ICOA-VMD feature extraction part is to optimize the parameters of VMD to extract feature vectors under different fault conditions. This step is completed during the model training phase. The training data contains data sets of various fault conditions (such as inner ring, outer ring and rolling element faults, etc.). Secondly, different fitness functions (such as minimum envelope entropy, minimum sample entropy, etc.) are selected according to the needs. This is used as the objective function and the ICOA algorithm is used to optimize the two parameters of VMD. The data set of each fault condition is optimized independently to find the best VMD parameters.

[0131] In one embodiment, the spatiotemporal feature analysis and time dependency analysis of the bearing fault data are performed based on the intrinsic mode function, and attention is introduced to the key features, including: using CNN to perform spatiotemporal feature analysis on the intrinsic mode function; using BiLSTM to capture the time dependency in the intrinsic mode function; introducing a self-attention mechanism to enhance the perception of the key features in the intrinsic mode function.

[0132] In one embodiment, the method further includes: constructing a fault identification network including a CNN, a BiLSTM and a self-attention layer, pre-training the fault identification network, inputting the intrinsic mode function into the pre-trained fault identification network, and obtaining a fault classification result.

[0133] In one embodiment, a bearing fault classification and recognition model based on ICOA-VMD-CNN-BiLSTM-Attention is proposed to fully reflect the spatiotemporal correlation and complexity of bearing fault data. First, the ICOA algorithm is used to optimize the parameters of VMD to extract effective intrinsic mode functions (IMFs) and enhance the accuracy of feature extraction. Secondly, CNN is used to analyze the spatial features of the extracted IMFs, and BiLSTM is applied to capture the temporal dependencies in the data. Thirdly, in order to further improve the model's attention to important information, Attention is introduced to enhance the perception of key information. Finally, the overall structure of the model is as follows: Figure 3 As shown in the figure, it includes data processing, ICOA algorithm, ICOA-VMD feature extraction and CNN-BiLSTM-Attention fault diagnosis. The design comprehensively considers spatiotemporal features, time dependencies and attention to key features, thus providing a more accurate and reliable solution for bearing fault diagnosis.

[0134] like Figure 3 As shown in the figure, the data processing part is first initialized, including clearing the variables in the workspace, setting the sliding window size, sample size, and the number of samples for each fault type. Secondly, the bearing vibration data files containing normal state and different fault types are loaded in turn. Thirdly, for each loaded data set, the code uses a sliding window to intercept the data and store these samples in the corresponding matrix. After completing the processing of all data sets, all the extracted sample data are merged into a large matrix. Finally, the merged data matrix is ​​saved as a file for subsequent analysis or processing.

[0135] The ICOA-VMD feature extraction part is to optimize the parameters of VMD to extract feature vectors under different fault conditions. First, the program initializes the environment and loads a data set containing 10 fault conditions (such as inner ring, outer ring and rolling element faults, etc.). Secondly, different fitness functions (such as minimum envelope entropy, minimum sample entropy, etc.) are selected according to the needs. With this as the objective function, the ICOA algorithm is used to optimize the two key parameters alpha and K of VMD. The data set for each fault state is optimized independently to find the optimal VMD parameters and the corresponding IMF components. Again, these optimized parameters are used to extract the feature vectors under each fault state, and all feature vectors are combined together and saved as files. At the same time, the optimal parameters and optimization convergence curves are also saved separately. Finally, the path is cleaned up to prevent subsequent operations from being affected.

[0136] The CNN-BiLSTM-Attention fault diagnosis section first loads and preprocesses a dataset containing features of different fault states. Labels are added to the data for each fault state, the data is divided into a training set and a test set, and normalized. Secondly, the program converts the data format of the training set and the test set into a format suitable for CNN and BiLSTM training. In terms of network structure, the program builds a deep learning model containing CNN layers, BiLSTM layers, and Attention layers, and sets hyperparameters for training, such as learning rate and training cycle. Thirdly, the program trains the network using the training set data and evaluates its classification accuracy on the test set. Finally, the confusion matrix of the test set and the scatter plot comparing the predicted results with the actual results are plotted to visualize the performance of the model.

[0137] In terms of model application, the study deployed the trained ICOA-VMD-CNN-BiLSTM-Attention model on an industrial computer, such as Figure 4 As shown in the figure. The model runs on an industrial computer, receives bearing fault data from the acceleration sensor, classifies and identifies faults in real time, and conducts reliability tests. The data is displayed on the local LCD display on the one hand, and uploaded to the IoT cloud platform for storage and analysis through wireless communication methods such as 4G on the other hand. By building a practical environment, it can provide practical technical solutions for real-time monitoring and fault diagnosis of industrial bearing status, and has high promotion and application value.

[0138] Six algorithms, including improved crayfish (ICOA), crayfish (COA), sand cat swarm (SCSO), osprey (OOA), whale (WOA), and subtraction average optimizer (SABO), were tested on the 12 benchmark functions provided by CEC2022. These benchmark functions include unimodal, multimodal, mixed and composite functions, covering different types of optimization problems. The unimodal function F1 is used to evaluate the local optimization performance of the algorithm, the multimodal functions F2-F5 are used to evaluate the global optimization performance of the algorithm, the mixed functions F6-F8 are used to evaluate the global and local optimization performance of the algorithm, and the composite and multimodal functions F9-F12 are used to evaluate the comprehensive performance of the algorithm. For each algorithm, the population size, maximum number of iterations and problem dimension are set to 30, 500 and 10 respectively. A comprehensive evaluation is performed by executing each test function 30 times. The optimal value and median value of fitness are used as evaluation indicators to examine the performance of the algorithm in terms of optimization accuracy and convergence speed. The test results are shown in Table 1 and the box plots are shown in Table 1. Figure 5 shown.

[0139] Table 1 Test results of six algorithms on CEC2022 benchmark function

[0140]

[0141]

[0142] The results show that the optimal fitness value of the ICOA algorithm on the multi-peak function F2 is 400, which is equal to the theoretical minimum value of the function 400, while the optimal values ​​of the COA, SCSO, OOA, WOA, and SABO algorithms are 400.1534, 403.2038, 736.0359, 400.6504, and 412.8405, respectively. The median value of the ICOA algorithm is 406.9689, which is closest to 400 among the medians of these six algorithms. On the composite and multimodal function F9, the optimal fitness value of the ICOA algorithm is 2529.2844, which is closest to the theoretical minimum value of the function, 2300, while the optimal values ​​of the other algorithms are 2529.4919, 2529.3289, 2658.919, 2531.0007, and 2558.4693, respectively. The median value of the ICOA algorithm is 2529.2844, which is closest to 2300 among the medians of the six algorithms. The optimization accuracy results of the six algorithms tested on the 12 benchmark functions of CEC2022 show that the optimal fitness value of the ICOA algorithm is slightly inferior to that of the WOA algorithm in the F10 function test, and the other performances are the best among the six algorithms. The median fitness value is slightly inferior to the SCSO algorithm in the F12 function test, and the other performances are the best among the six algorithms. Therefore, the ICOA algorithm shows relatively excellent optimization accuracy and robustness. Figure 6 and Figure 7The convergence curves of the six algorithms on representative functions are shown, with a maximum number of iterations of 1000. It can be further seen that the ICOA algorithm also has a relatively good convergence speed.

[0143] In order to further demonstrate the advancement of this application, the model was tested and compared. In the model test and comparison analysis of bearing fault diagnosis, the study used the bearing fault data set published by Case Western Reserve University (CWRU). Figure 8 As shown in the figure, it is a schematic diagram of the structure of the test device. The specific marks in the figure are as follows: 1, fan end bearing; 2, motor; 3, drive end bearing; 4, torque sensor and encoder; 5, power tester; 6, bearing outer ring; 7, bearing inner ring; 8, bearing rolling element; 9, bearing retainer. The core components of the bearing experimental device include a motor, a torque sensor / decoder and a power tester. The faulty bearing used in the experiment is installed at the drive end of the motor. In order to simulate faults of different positions and damage degrees, the experiment prepared multiple bearings with different fault positions and damage degrees. During the experiment, the faulty bearing was installed in the motor. When the motor rotates, the fault signal of the bearing is transmitted in the form of vibration. In order to capture these vibration signals, the experiment installed acceleration sensors at three locations, namely, on the bearing seat at the fan end and drive end of the motor, and on the base at the bottom of the experimental device. All sensors collect data at the same time and store these data in the same file, forming three independent data channels, providing basic data for subsequent fault analysis. The dataset contains vibration signals and fault labels under different working conditions, and records a variety of features, including the frequency domain components of the vibration signal and the fault category labels. Specifically, the dataset involves 10 different states, including the normal state, and the fault state with the inner ring, rolling element and outer ring having diameters of 0.007 inches, 0.014 inches and 0.021 inches respectively at a speed of 1797RPM, represented by categories 1 to 10. Each state contains 120 samples, the first 90 samples are selected for training, and the last 30 samples are selected for testing. The vibration signal data volume of each sample is 1×2048 points. During the processing and analysis process, a sliding window (w) of size 1000 is used to segment the data in order to capture the fault features more effectively. These samples provide sufficient data support for the subsequent fault diagnosis model.

[0144] In order to evaluate the accuracy and reliability of the model, the performance of four fault diagnosis models was compared, including the CNN-BiLSTM-Attention model, the OOA-VMD-CNN-BiLSTM-Attention model, the COA-VMD-CNN-BiLSTM-Attention model, and the ICOA-VMD-CNN-BiLSTM-Attention model. Under the same parameter conditions, each model was trained and tested 10 times. The classification performance of the CNN-BiLSTM-Attention model on the test set was average, with an average accuracy (Aacc) of 70.73% and a best accuracy (Bacc) of 75%. On this basis, the OOA-VMD-CNN-BiLSTM-Attention model that uses the OOA algorithm to optimize VMD parameters further improves the feature extraction capability, with Aacc of 89.6% and Bacc of 91.33%. After using the COA algorithm to optimize VMD parameters, the classification effect of the COA-VMD-CNN-BiLSTM-Attention model is further improved, with Aacc of 91.4% and Bacc of 93.33%. Among all the models, the ICOA-VMD-CNN-BiLSTM-Attention model using the ICOA algorithm performed best, with an Aacc of 98.23% and a Bacc of 99.33%. The classification confusion matrix of the four models for a single test is shown in the figure below: Figures 9 to 12As shown, the number of accurate predicted samples of the CNN-BiLSTM-Attention model on categories 0 to 10 are [30, 30, 25, 29, 2, 6, 29, 26, 16, 21] respectively; those of the OOA-VMD-CNN-BiLSTM-Attention model are [30, 30, 18, 27, 30, 28, 24, 30, 27, 30] respectively; those of the COA-VMD-CNN-BiLSTM-Attention model are [30, 30, 30, 13, 30, 29, 30, 30, 28, 30] respectively; and those of the ICOA-VMD-CNN-BiLSTM-Attention model are [30, 30, 30, 30, 30, 29, 30, 30, 29, 30] respectively. With the gradual introduction of optimization algorithms from OOA to COA and then to ICOA, the classification accuracy of the model has been continuously improved, especially in categories 4 and 9, where the ICOA-VMD-CNN-BiLSTM-Attention model performed best. These categories did not perform well in the basic model CNN-BiLSTM-Attention, but after introducing more advanced optimization algorithms, significant performance improvements were achieved. The ICOA-VMD-CNN-BiLSTM-Attention model demonstrated high classification accuracy and robustness, and has important application value in bearing fault classification and identification tasks.

[0145] In summary, the bearing fault diagnosis model based on ICOA-VMD-CNN-BiLSTM-Attention combines VMD, CNN, BiLSTM and Attention to fully capture data features from the spatiotemporal dimensions. The specific conclusions are as follows:

[0146] (1) Through the combination of VMD and CNN, the model can effectively extract the spatial features of bearing fault data. BiLSTM further captures the temporal dependency, while the introduction of Attention enhances the model's attention to key fault information and improves the accuracy and robustness of fault classification.

[0147] (2) The ICOA algorithm is introduced to optimize VMD parameters and model hyperparameters. By combining the center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategies, the ICOA algorithm performs well in terms of optimization accuracy and convergence speed. It outperforms the traditional COA and other five algorithms in most function performance tests of CEC2022, ensuring the performance of the model on complex data sets.

[0148] (3) After training, the model is deployed on an industrial computer. Experimental results show that the model exhibits high accuracy and stability in bearing fault diagnosis, providing reliable technical support for intelligent monitoring and fault prevention of industrial equipment.

[0149] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0150] Based on the same inventive concept, the embodiment of the present application also provides a bearing fault classification and identification device for implementing the bearing fault classification and identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more bearing fault classification and identification device embodiments provided below can refer to the limitations of the bearing fault classification and identification method above, and will not be repeated here.

[0151] In one embodiment, Fig.13 As shown, a bearing fault classification and identification device is provided, including: a data acquisition module 1302, a feature extraction module 1304 and a fault classification module 1306, wherein:

[0152] The data acquisition module 1302 is used to acquire bearing fault data when the industrial computer is running;

[0153] The feature extraction module 1304 is used to process the bearing fault data using the variational mode decomposition method to extract the intrinsic mode function;

[0154] A fault classification module 1306 is used to perform spatiotemporal feature analysis and time dependency analysis on bearing fault data based on intrinsic mode functions, and introduce attention to key features to obtain fault classification results;

[0155] Among them, the parameters of the variational mode decomposition method are optimized using the improved crayfish optimization algorithm, which combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

[0156] Each module in the above-mentioned bearing fault classification and identification device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0157] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in all the above method embodiments when executing the computer program.

[0158] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0159] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in all the above method embodiments when executed by a processor.

[0160] 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 laws, regulations and standards of relevant countries and regions.

[0161] Those of ordinary skill 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), magnetic 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. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0162] 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.

[0163] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for classifying and identifying bearing faults, characterized in that: The method comprises: Obtain bearing fault data when the industrial computer is running; Processing the bearing fault data using a variational mode decomposition method to extract intrinsic mode functions; Based on the intrinsic mode function, the spatiotemporal feature analysis and the time dependency analysis are performed on the bearing fault data, and attention is paid to key features to obtain a fault classification result; Among them, the parameters of the variational mode decomposition method are optimized using an improved crayfish optimization algorithm, and the improved crayfish optimization algorithm combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

2. The method according to claim 1, characterized in that The obtaining of bearing fault data when the industrial computer is running includes: Acquire a vibration signal of a target position on the industrial computer when the industrial computer is running; The vibration signal is subjected to data preprocessing to obtain the bearing fault data.

3. The method according to claim 1, characterized in that The improved crayfish optimization algorithm includes: Introduce the center of gravity reverse learning to generate the spatial position matrix and determine the cutoff value; Entering an iterative process; the iterative process includes: generating a first random number and a second random number, wherein the value range of the second random number is 0 to 1; when the first random number is greater than the demarcation value and the second random number is less than 0.5, obtaining a first candidate solution according to the spatial position matrix and an exponential decay function in which the second random number and the current number of iterations are substituted; when the first random number is greater than the demarcation value and the second random number is not less than 0.5, obtaining a second candidate solution according to the spatial position matrix; when the first random number is not greater than the demarcation value, obtaining a third candidate solution according to the spatial position matrix and the second random number; introducing a sparrow alert mechanism to compare the fitness value of the first candidate solution, the second candidate solution or the third candidate solution with the fitness value of the current optimal solution, and updating the current optimal solution according to the comparison result and the first candidate solution, the second candidate solution or the third candidate solution; The current optimal solution is iteratively updated until a preset number of iterations is reached, and a target solution is obtained according to the current optimal solution corresponding to the last iteration.

4. The method according to claim 1, characterized in that: The parameters of the variational mode decomposition method are optimized using an improved crayfish optimization algorithm during the training phase. The training steps include: Obtain training data for multiple fault types; For each of the fault types, the corresponding training data is extracted, and the parameters of the variational mode decomposition method are optimized using the improved crayfish optimization algorithm to obtain the corresponding target parameters.

5. The method according to claim 1, characterized in that The time-space characteristic analysis and time-dependency analysis of the bearing fault data based on the intrinsic mode function, and the introduction of attention to key characteristics, include: Using CNN to perform spatiotemporal feature analysis on the intrinsic mode function; Using BiLSTM to capture the time dependency in the intrinsic mode function; A self-attention mechanism is introduced to enhance the perception of key features in the intrinsic mode function.

6. The method according to claim 1, characterized in that The method further comprises: A fault identification network including CNN, BiLSTM and self-attention layer is constructed, the fault identification network is pre-trained, the intrinsic mode function is input into the pre-trained fault identification network, and the fault classification result is obtained.

7. A bearing fault classification and identification device, characterized in that: The device comprises: A data acquisition module is used to acquire bearing fault data when the industrial computer is running; A feature extraction module, used to process the bearing fault data using a variational mode decomposition method to extract an intrinsic mode function; A fault classification module, used for performing spatiotemporal feature analysis and time dependency analysis on the bearing fault data based on the intrinsic mode function, and introducing attention to key features to obtain fault classification results; Among them, the parameters of the variational mode decomposition method are optimized using an improved crayfish optimization algorithm, and the improved crayfish optimization algorithm combines the crayfish optimization algorithm, center of gravity reverse learning, nonlinear control and sparrow warning mechanism strategy.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images