Residual network bearing fault diagnosis method and system driven by adaptive South goose algorithm

Abstract

The invention discloses a residual network bearing fault diagnosis method driven by a self-adaptive South goose algorithm. The method comprises the following steps: acquiring to-be-diagnosed bearing fault data; inputting bearing fault data to be diagnosed into the trained fault diagnosis model, and outputting a fault diagnosis result; wherein the training of the fault diagnosis model comprises the following steps: constructing a bearing fault sample data set; a fault diagnosis model is constructed, the fault diagnosis model comprises a frequency domain attention module and a flexible residual network, deep feature extraction is carried out on the enhanced features through a dual flexible residual block, the number of convolution layers of the dual flexible residual block is adjustable, and the dual flexible residual block comprises two flexible residual paths connected in series; a jump connection position dynamic selection mechanism and a cross-path cross connection weight and global average pooling attention mechanism are introduced; and introducing an adaptive Schwane algorithm to optimize network structure parameters of the fault diagnosis model, carrying out model training on the constructed sample data set based on the optimal network structure parameters, and outputting the trained fault diagnosis model.

Description

Technical Field

[0001] This invention belongs to the field of intelligent diagnostic technology for rotating equipment, and particularly relates to a bearing fault diagnosis method based on a frequency domain enhanced residual network driven by a multi-stage adaptive snow goose algorithm. Background Technology

[0002] As a core component of rotating machinery, the reliability of bearings is crucial to the safe and stable operation of the entire mechanical system. During industrial production, bearings operate under high-speed rotation and heavy loads for extended periods, making them prone to failures such as inner ring cracks, outer ring wear, and rolling element spalling. These failures not only reduce equipment efficiency but can also cause equipment damage and safety accidents. Therefore, there is an urgent need for an efficient intelligent diagnostic model to provide technical support for the safe operation of the system.

[0003] Currently, deep learning-based fault diagnosis methods are widely used in rotating machinery fields such as bearings. Convolutional Neural Networks (CNNs) have achieved remarkable results in vibration signal analysis due to their excellent local feature extraction capabilities. Applying CNNs to bearing fault diagnosis can achieve good diagnostic accuracy. However, CNNs have high computational complexity and are difficult to capture complex patterns over a wider range. Residual Networks (ResNets), through their skip connection mechanism, effectively solve the gradient vanishing problem in deep networks and exhibit excellent performance in the field of fault diagnosis. However, existing studies mostly use fixed network structures, which are difficult to adaptively adjust according to different data characteristics and diagnostic tasks, thus limiting the model's generalization ability and diagnostic accuracy.

[0004] Furthermore, traditional methods primarily focus on time-domain features, failing to fully utilize frequency-domain information and exhibiting limitations when processing complex spectral features. To overcome the limitations of fixed network structures, Neural Architecture Search (NAS) technology has gradually become a research hotspot. Researchers have begun applying automated network design concepts to the field of fault diagnosis. However, traditional NAS methods are computationally expensive, and existing methods often remain at the level of simple structural parameter searches, lacking in-depth optimization of the network's intrinsic mechanisms, and thus failing to achieve a synergistic improvement in structural flexibility and feature representation capabilities.

[0005] With the rapid development of attention mechanisms, frequency domain attention techniques have begun to receive widespread attention in the field of signal processing. Frequency domain analysis can reveal the spectral information contained in vibration signals and has unique sensitivity to different fault modes. However, how to effectively integrate frequency domain feature extraction with attention mechanisms and explore the discriminative potential of frequency domain features still requires further research.

[0006] The Snow Goose Algorithm (SGA), an emerging swarm intelligence algorithm, primarily achieves global optimization by simulating the migration and foraging behaviors of snow geese. However, traditional SGA suffers from problems such as being prone to getting trapped in local optima and having slow convergence speed, making it difficult to meet the needs of optimizing complex network structures. Therefore, how to enhance the performance of SGA in neural network optimization still requires in-depth research. Summary of the Invention

[0007] To address the challenges of intelligently selecting topology configuration parameters for deep neural networks and insufficient extraction of key frequency domain features from vibration signals in rotating machinery bearing fault diagnosis, this invention provides an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method. By deeply coupling the adaptive snow goose algorithm (ISGA) with a frequency domain attention-enhanced dual flexible residual network (DFResNet-FDAM) and synergistically utilizing time-frequency domain information, it achieves high-accuracy identification and intelligent discrimination of bearing degradation types, effectively solving the problems of insufficient network architecture flexibility and inadequate frequency domain information extraction.

[0008] According to one aspect of the present invention, an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method is provided, comprising: Obtain data on the bearing fault to be diagnosed; The bearing fault data to be diagnosed is input into the trained fault diagnosis model, and the fault diagnosis result is output; wherein, the training of the fault diagnosis model includes: Construct a bearing failure sample dataset; A fault diagnosis model is constructed, including: a frequency domain attention module, used to enhance the frequency domain perception attention features of the input bearing fault samples; and a flexible residual network, used to extract deep features from the enhanced features through dual flexible residual blocks. The dual flexible residual blocks contain two serial flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. At the same time, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to realize information exchange and feature enhancement between differentiated residual paths. An adaptive snow goose algorithm is introduced to optimize the network structure parameters of the fault diagnosis model. Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, and the trained fault diagnosis model is output.

[0009] As a further technical solution, the construction of the adaptive snow goose algorithm includes: Initialize the Snow Goose algorithm population, with each individual's encoding containing network structure parameters; Establish a multi-stage evolutionary discrimination criterion function, and dynamically identify the evolutionary stage of the algorithm based on the population dispersion index and fitness increment; A strategy for generating complementary solutions is introduced to generate complementary solutions for the current candidate solutions and combine this with guidance from elite individuals, thereby broadening the search range of the solution space. Design an elite solution memory bank management mechanism to store high-quality solutions from historical iterations and guide population evolution through a weighted selection strategy; Construct a cooperative foraging mechanism among individuals, calculate the cooperative strength index among individuals and incorporate it into the position update equation to enhance the collective search ability of the population; The flight strategy is adaptively switched between the V-formation flight strategy in the exploration phase and the linear flight strategy in the development phase based on the results of multi-stage evolution.

[0010] As a further technical solution, an adaptive snow goose algorithm is introduced to optimize the network's structural parameters, including: Set the execution parameters of the adaptive snow goose algorithm, including the number of individuals in the population, the maximum number of generations, the capacity of the elite memory bank, and the probability of adversarial learning; For each individual in the current population, construct the network topology of the corresponding fault diagnosis model; The constructed network topology is iteratively trained on the training set samples; The classification performance of the network topology is evaluated on the validation set samples, and the recognition accuracy is used as the fitness value of the individual. The population evolution operation of the adaptive snow goose algorithm is performed, and multi-stage adaptation, opposition solution generation, elite memory guidance and cooperative foraging strategies are applied. When the maximum number of generations or fitness convergence criteria are met, the parameter configuration scheme of the globally optimal individual is output.

[0011] As a further technical solution, based on the optimal network structure parameters, model training is performed on the constructed sample dataset, including: The network topology of the fault diagnosis model is constructed based on the optimal network structure parameters obtained by the adaptive snow goose algorithm. Configure the hyperparameters for network training; The optimal network structure parameters are iteratively trained on the training sample set, and the performance of the trained network model is evaluated on the test sample set. Output fault identification results, including overall accuracy, macro average precision, macro average recall, and macro average F1 score.

[0012] As a further technical solution, before iterative optimization, the following network structure parameter search space is defined: Define the search range for the number of convolutional layers for the first and second flexible residual blocks; Determine the search interval for the number of jump connections between the first flexible residual block and the second flexible residual block; A set of candidate values ​​for configuring the number of convolutional kernels.

[0013] As a further technical solution, the construction of the frequency domain attention module also includes: Perform a Fast Fourier Transform on the feature tensor of the input bearing fault samples; The amplitude and phase of the frequency domain complex result after the fast Fourier transform are decoupled and separated to obtain the spectral amplitude matrix and phase angle matrix respectively. Design a frequency channel attention weighting mechanism to achieve feature selection in the channel dimension by learning the importance coefficients of different frequency components; Construct a frequency domain spatial attention focusing mechanism to focus on salient regions at different frequency positions in the spectral amplitude matrix; Perform frequency domain feature parameterization enhancement and phase compensation operations to optimize the spectral representation through a trainable adaptive filter; The enhanced frequency domain representation is reconstructed back to the time domain using the inverse fast Fourier transform, and the original time domain features are integrated through a gating fusion mechanism to obtain the final enhanced features.

[0014] As a further technical solution, a bearing failure sample dataset is constructed, including: Collect bearing vibration data under various speed conditions; The original vibration signal is serialized and segmented using a sliding window method. The window width and sliding step are set to generate a sample sequence of fixed length. Perform Z-score normalization on the segmented sample sequences; The normalized samples are split into datasets, and the ratio of training set to test set is set.

[0015] According to one aspect of this specification, an adaptive snow goose algorithm-driven residual network bearing fault diagnosis system is provided for implementing the method, comprising: The data acquisition module is used to acquire bearing fault data to be diagnosed. The fault diagnosis module is used to input the bearing fault data to be diagnosed into the trained fault diagnosis model and output the fault diagnosis result; wherein, the training of the fault diagnosis model includes: Construct a bearing failure sample dataset; A fault diagnosis model is constructed, including: a frequency domain attention module, used to enhance the frequency domain perception attention features of the input bearing fault samples; and a flexible residual network, used to extract deep features from the enhanced features through dual flexible residual blocks. The dual flexible residual blocks contain two serial flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. At the same time, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to realize information exchange and feature enhancement between differentiated residual paths. An adaptive snow goose algorithm is introduced to optimize the network structure parameters of the fault diagnosis model. Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, and the trained fault diagnosis model is output.

[0016] According to one aspect of the present invention, an adaptive snow goose algorithm-driven residual network bearing fault diagnosis device is provided, comprising a memory and a processor, wherein the memory stores program instructions that are executed by the processor, and the processor calls the program instructions to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method.

[0017] According to one aspect of the present invention, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions that cause the computer to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By integrating ISGA and DFResNet-FDAM, a diagnostic system that combines intelligent neural architecture search with deep extraction of frequency domain features was constructed. This effectively solved the problems of poor flexibility of fixed network topology and insufficient depth of frequency domain information utilization, and significantly improved the accuracy of bearing fault identification and the efficiency of network architecture search. 2. By introducing multi-stage adaptive mechanisms, complementary solution generation, elite memory guidance, and cooperative foraging strategies, the global exploration capability and local fine-grained search capability of the traditional SGA algorithm are improved. A dual-path flexible residual structure and a frequency domain-aware attention module are designed to enable the network to adaptively adjust the topology configuration according to the fault vibration characteristics and deeply mine the spectrum discrimination information, thereby enhancing the structural elasticity and feature representation capability of the deep learning architecture. 3. A multi-level collaborative design of convolutional units, residual paths, frequency domain transformation, and attention mechanisms was realized. The intelligent topology search strategy effectively improved the model performance, providing an effective technical means for intelligent diagnosis and preventive maintenance of rotating machinery bearing faults under complex operating conditions. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method in an embodiment of the present invention. Figure 2 This is a schematic diagram of the training process of the fault diagnosis model in an embodiment of the present invention; Figure 3 This is a basic structural diagram of ResNet in an embodiment of the present invention; Figure 4 This is a basic schematic diagram of DF-ResNet in an embodiment of the present invention; Figure 5 This is a basic schematic diagram of the SGA algorithm in this embodiment of the invention; Figure 6 This is a flowchart of the ISGA-DFResNet-FDAM diagnostic process in an embodiment of the present invention. Detailed Implementation

[0021] To address the shortcomings of existing technologies, there is an urgent need for a novel adaptive deep learning network construction method. This method should effectively address the challenge of extracting salient frequency domain features from complex vibration signals, enabling intelligent adaptation of neural network architecture parameters to fault vibration modes, and improving the classification accuracy and model robustness of bearing fault diagnosis. Therefore, this invention proposes an adaptive Snow Goose algorithm-driven residual network bearing fault diagnosis method. Through deep coupling between ISGA and DFResNet-FDAM and the synergistic utilization of time-frequency domain information, it achieves high-accuracy identification and intelligent discrimination of bearing degradation types, effectively solving the problems of insufficient network architecture flexibility and inadequate frequency domain information extraction.

[0022] This invention primarily focuses on improving swarm intelligence algorithms and network topology search strategies to enhance the intelligence level of rotating machinery bearing fault diagnosis under complex spectral environments. It proposes an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method. This invention fully leverages the advantages of ISGA neural topology search and DFResNet-FDAM frequency domain feature depth extraction, achieving high-accuracy identification of multiple fault modes through their integrated implementation. A multi-level feature extraction strategy is designed for complex vibration signals, enabling intelligent collaboration between convolutional units, residual paths, frequency domain transformation, and attention mechanisms. The diagnostic results obtained through this invention are more reliable and accurate, effectively improving the accuracy of bearing fault identification and the level of equipment health monitoring under complex operating conditions.

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0024] The technical solution of this invention is to provide an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method, which aims to improve the adaptive configuration capability of deep network topology for differentiated fault modes and the deep extraction level of frequency domain discrimination features, realize the intelligent exploration of residual network architecture parameters, and ensure high-accuracy fault diagnosis and equipment health status assessment of bearings in complex operating environments.

[0025] Please see Figure 1 The present invention provides an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method. First, the bearing fault data to be diagnosed is obtained; then, the bearing fault data to be diagnosed is input into the trained fault diagnosis model, and the fault diagnosis result is output.

[0026] The training process of the fault diagnosis model is as follows: Figure 2 As shown, it includes: Step S1: Construct a bearing failure sample dataset; Step S2, constructing a fault diagnosis model, including: a frequency domain attention module, used to perform frequency domain perception attention enhancement features on the input bearing fault samples; and a flexible residual network, used to extract deep features from the enhanced features through dual flexible residual blocks. The dual flexible residual blocks contain two serial flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. At the same time, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to realize information exchange and feature enhancement between differentiated residual paths. Step S3: The adaptive snow goose algorithm is introduced to optimize the network structure parameters of the fault diagnosis model. Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, and the trained fault diagnosis model is output.

[0027] Specifically, the construction of the bearing fault sample dataset in step S1 includes the following bearing vibration signal acquisition and data preprocessing: Step 1.1: Collect bearing vibration data under various speed conditions, covering multiple operating states such as healthy condition, inner ring damage, outer ring damage, and rolling element wear; Step 1.2: The original vibration signal is serialized and segmented using the sliding window segmentation technique. The window width and sliding step are set to generate a sample sequence of fixed length. Step 1.3: Perform Z-score normalization on the segmented sample sequences to eliminate the influence of numerical units and amplitude scale;

[0028] Step 1.4: Perform dataset splitting, set the ratio of training set to test set, and ensure a balanced distribution and sufficient quantity of various types of fault samples.

[0029] The topology design of the flexible residual network described in step S2 includes the following steps: Step 2.1: Construct a flexible residual base module with variable number of convolutional layers to achieve flexible configuration of the network's vertical depth by dynamically adjusting the number of convolutional layers; Step 2.2: Design a dynamic selection mechanism for skip connection positions, and determine the access and output positions of the residual path based on the total number of skip connections. Step 2.3: Establish a cascaded topology of dual flexible residual blocks to enhance the nonlinear fitting capability of the network by cascading the two flexible residual channels. Step 2.4: Introduce cross-path cross-connection weights and global average pooling attention mechanism to achieve information exchange and feature enhancement between differentiated residual paths.

[0030] Please refer to the basic structure of ResNet. Figure 3Residual networks effectively solve the vanishing gradient problem in deep neural network training by introducing a skip connection mechanism. Let x and y be the input and output of the residual block, and W... i Let F be the weight matrix of the i-th layer, and F(-) be the residual function. The core idea of ​​the standard residual block is to learn the residual mapping rather than the direct mapping, and its mathematical expression can be represented by the following formula: .

[0031] Define ReLU(-) as the ReLU activation function, Bn(-) as the batch normalization operation, and * as the convolution operation. For a standard residual block, the residual function typically contains two convolutional layers, as shown in the following equation: .

[0032] The basic principle of DF-ResNet in the embodiments of this invention is as follows: Figure 4 As shown. The standard ResNet uses a fixed number of convolutional layers and a skip connection pattern, which limits the network's ability to adapt to different data characteristics. Therefore, this embodiment of the invention proposes a flexible residual block, which introduces variable parameters to achieve adaptive adjustment of the structure.

[0033] Define Dp(-) as the Dropout function, N c Let h be the number of convolutional layers in the flexible residual block. The output h of the i-th convolutional layer is then... i It can be represented as follows: .

[0034] Define N s Let C be the number of skip connections in the flexible residual block, t be the t-th skip connection, and c be the number of skip connections. Then the position of the skip connection is C. t It can be dynamically determined by the following formula: .

[0035] Fb1 and Fb2 are defined as the first and second flexible residual blocks, respectively. To further enhance the expressive power of the network, this embodiment of the invention designs a dual flexible residual block, which contains two cascaded flexible residual paths. This process can be represented by the following equation: .

[0036] Define W c Let G(-) be the cross-connect weight, G(-) be the global average pooling weight, A be the attention weight, and ⊙ denote element-wise multiplication. Then, the cross-connect and attention fusion mechanism can be represented as follows: .

[0037] Furthermore, the construction of the frequency domain attention module in step S2 includes the following steps: Step S21: Perform a Fast Fourier Transform operation on the input feature tensor to achieve a mathematical transformation from the time domain to the frequency domain. Step S22: Perform decoupling and phase separation processing on the frequency domain complex results to obtain the spectral amplitude matrix and phase angle matrix respectively; Step S23: Design a frequency channel attention weighting mechanism to achieve feature selection in the channel dimension by learning the importance coefficients of different frequency components; Step S24: Construct a frequency domain spatial attention focusing mechanism to focus on salient regions at different frequency positions in the spectrum matrix; Step S25: Perform frequency domain feature parameterization enhancement and phase compensation operations to optimize the spectral representation through a trainable adaptive filter; Step S26: Use inverse fast Fourier transform to reconstruct the enhanced frequency domain representation back to the time domain space, and integrate the original time domain features through a gated fusion mechanism to form a feature containing the original time domain features and the enhanced time domain features, and input it into two residual channels of the flexible residual network.

[0038] Traditional CNN networks mainly extract features in the time domain, which does not make full use of frequency domain information. Therefore, this invention proposes a frequency domain attention module (FDAM) to mine frequency domain discriminative features through Fourier transform (FFT).

[0039] Define FFT(-) as the Fourier function, and B, C, and L as the batch size, number of channels, and signal length, respectively. For the input signal x∈R B×C×L First, an FFT transformation is performed to decompose the amplitude and phase. This process can be represented by the following formula: .

[0040] Define w1 and w2 as channel attention weights. Frequency domain channel attention W ch The process of learning the importance weights of different frequency components can be represented by the following formula: .

[0041] Define W cn The spatial attention convolution weights are used. Frequency domain spatial attention W s The process of focusing on the importance of different frequency positions can be represented by the following formula: .

[0042] Define F a W is a learnable adaptive filtering parameter. p For phase correction weights. The attention-weighted frequency domain features require amplitude enhancement and phase correction, a process that can be expressed by the following equation: .

[0043] Define IFFT(-) as the inverse Fourier operation, W g These are the gating weights. Finally, enhanced features are output through inverse FFT transformation and gating mechanism, a process that can be represented by the following equation: .

[0044] Step S3, the design of the adaptive snow goose algorithm, includes the following steps: Step 3.1: Initialize the Snow Goose algorithm population. The encoding of each individual includes network topology parameters such as the number of convolutional layers, the number of convolutional kernels, and the number of hop connections. Step 3.2: Establish a multi-stage evolution discrimination criterion function, and dynamically identify the evolutionary stage of the algorithm based on the population dispersion index and fitness increment; Step 3.3: Introduce a strategy for generating complementary solutions, generate complementary solutions for the current candidate solutions, and combine this with elite individuals for guidance, thereby broadening the search range of the solution space; Step 3.4: Design an elite solution memory bank management mechanism to store high-quality solutions from historical iterations and guide population evolution through a weighted selection strategy; Step 3.5: Construct a cooperative foraging mechanism among individuals, calculate the cooperative strength index among individuals and incorporate it into the position update equation to enhance the collective search ability of the population; Step 3.6: Adaptively switch between the V-formation flight strategy in the exploration phase and the linear flight strategy in the development phase based on the multi-stage evolution discrimination results.

[0045] Optionally, before performing iterative search using the adaptive snow goose algorithm, the following step of defining the search space for network topology parameters is also included: Define the search range for the number of convolutional layers for the first and second flexible residual blocks; Determine the search interval for the number of jump connections in the first and second flexible residual blocks; A set of candidate values ​​for configuring the number of convolutional kernels.

[0046] The basic principle of the SGA algorithm in this embodiment of the invention is as follows: Figure 5 As shown. The Snow Goose Algorithm (SGA) is a novel swarm intelligence optimization algorithm proposed by Tian et al. in 2024. This algorithm achieves global optimization by simulating the migration and foraging behavior of snow geese.

[0047] Define X i,j Let l be the position of the i-th snow goose in the j-th dimension. j and u j Let `rand` be the upper and lower bounds of the problem, and `rand` be a random number between 0 and 1. In the snow goose population, each snow goose represents a candidate solution, and the population initialization process can be represented as: .

[0048] Let N be the snow goose population size, and D be the dimension. The snow goose population matrix X and fitness matrix F can be represented by the following equations: , .

[0049] Define V t Let p be the velocity vector, p be the random perturbation term, and p be the inertia weight coefficient. T is the maximum number of iterations, and t is the current number of iterations.

[0050] During long-distance migrations, snow geese form unique V-shaped formations to reduce flight drag and energy consumption. This formation effectively utilizes the updrafts generated by the leading individual, thereby improving overall flight efficiency. Based on this phenomenon, the speed update mechanism of the snow goose algorithm can be expressed by the following formula: , .

[0051] definition Let i be the position of the i-th snow goose in generation t. This is the current globally optimal position. For other randomly selected individual locations, B and D are the attraction and repulsion factors, respectively.

[0052] In actual flock migrations, snow geese develop different behavioral patterns based on their flight experience and physical condition. Dominant individuals typically occupy the front position of the formation, while other individuals adopt a follower strategy. The SGA algorithm defines the top 1 / 5 of the population as leaders, and their position update formula can be expressed as: .

[0053] definition The location of the nearest outstanding individuals. Let E be the position of adjacent individuals and E be the coordination factor. For other following individuals in the formation, their position update strategy can be expressed as: .

[0054] R is defined as a random number between 0 and 1, ⊕ is a vector addition operation, and B(D) is a random vector generated by d-dimensional Brownian motion, used to simulate small disturbances during flight.

[0055] After the snow geese complete their formation adjustments, the flock enters a relatively stable straight-line flight phase. During this phase, individuals primarily maintain consistency in flight direction through local information exchange. The SGA algorithm employs a probabilistic selection mechanism to simulate this behavioral characteristic, which can be represented by the following equation: .

[0056] The standard SGA algorithm suffers from problems such as being prone to getting trapped in local optima and having slow convergence speed. To address these issues, this invention proposes the Adaptive Snow Goose Algorithm (ISGA), which integrates multiple enhancement strategies to improve optimization performance.

[0057] (1) Multi-stage adaptive mechanism Define σ(t) to represent population diversity, θ e To explore the threshold, Δf(t) is the fitness improvement amount, ϵ s The threshold for determining stagnation is defined here. The SGA algorithm employs a fixed search strategy, which is difficult to adapt to the needs of different optimization stages. This invention designs a multi-stage adaptive mechanism that can dynamically adjust the search behavior according to the population state. The stage discrimination function can then be defined as follows: .

[0058] definition Let ε be the optimal fitness value in generation t, and ϵ be a positive number to prevent the occurrence of a zero value. Then, the improvement in population diversity and fitness at this point can be expressed by the following formula: .

[0059] (2) Oppositional learning strategies Define P b Given the base complement probability and α as the decay parameter, complement learning enhances the algorithm's search capability by simultaneously considering both the solution and its complement. This process can be represented by the following equation: .

[0060] Let β be the elite guiding coefficient. Then the formula for generating the complementary solution can be expressed as follows: .

[0061] (3) Elite Memory Management Define M as the capacity of the elite library, e i Let be the i-th elite solution. To improve the optimization performance of the SGA algorithm, an elite memory is designed to store excellent solutions and guide the search process. This process can be represented by the following formula: .

[0062] Define η as the learning rate, e s This represents an elite solution selected based on fitness and diversity weighting. The elite-guided position update can then be expressed as: .

[0063] (4) Cooperative foraging mechanism Define C i,jσ represents the strength of cooperation between individuals i and j. c f is the spatial distance parameter. a Let the average fitness be denoted as . To enhance coordination among populations, a subpopulation cooperation mechanism is designed to improve information exchange. This process can be represented by the following formula: .

[0064] Define w v Let c1 and c2 be the inertia weights, c1 and c2 be the acceleration coefficients, and r1 be a random number. The velocity update formula for cooperative guidance can then be expressed as follows: .

[0065] To verify the optimization performance of the ISGA algorithm, the Shekel benchmark test function was selected for experiments, with the number of test iterations set to 1000. ISGA was then compared and analyzed with the standard SGA, Dung Beetle Optimization (DBO), and Grey Wolf Optimization (GWO) algorithms. The mathematical expression of the Shekel test function is shown below: .

[0066] Furthermore, after designing the adaptive snow goose algorithm, the execution steps of network topology search driven by ISGA based on the adaptive snow goose algorithm are as follows: Step S31: Set the ISGA algorithm execution parameters, including the number of individuals in the population, the maximum number of generations, the elite memory bank capacity, and the opposition learning probability; Step S32: For each individual in the current population, construct the corresponding DFResNet-FDAM network topology; Step S33: Perform rapid iterative training on the constructed network topology using samples from the training set; Step S34: Evaluate the classification performance of the network topology on the validation set samples, and use the recognition accuracy as the fitness value of the individual. Step S35: Perform population evolution operations using the ISGA algorithm, applying strategies such as multi-stage adaptation, oppositional solution generation, elite memory guidance, and cooperative foraging; Step S36: Determine whether the maximum number of generations or fitness convergence criterion is met. If not, return to step 6.2 to continue the search iteration. Step S37: Output the parameter configuration scheme of the globally optimal individual.

[0067] Furthermore, after obtaining the optimal network structure parameters, the following steps for optimal network topology training and fault identification are performed: Step S38: Construct the final DFResNet-FDAM diagnostic network based on the optimal topology parameter scheme obtained from the ISGA search; Step S39: Configure the hyperparameters for network training, including the initial learning rate, mini-batch sample size, and maximum number of iterations; Step S310: Perform sufficient iterative training on the optimal network on the complete training sample set; Step S311: Evaluate the performance of the trained network model on an independent test sample set; Step S312: Output the fault identification results, including overall accuracy, macro average precision, macro average recall, and macro average F1 score.

[0068] The ISGA-DFResNet-FDAM diagnostic flowchart in this embodiment of the invention is as follows: Figure 6 As shown.

[0069] The ISGA-DFResNet-FDAM ensemble framework organically combines the adaptive snow goose algorithm, a dual-flexible residual network, and a frequency domain attention module to achieve end-to-end optimization for bearing fault diagnosis. Structural parameters such as the number of convolutional layers, kernels, and skip connections directly affect the network's feature extraction capability and diagnostic performance; therefore, the ISGA algorithm is used to optimize these key parameters.

[0070] The specific implementation steps are as follows: Step 1: Collect bearing fault data, standardize the raw signals and perform sliding window sampling, and divide the training set and test set according to a 7:3 ratio to ensure that the samples of each fault category are evenly distributed.

[0071] Step 2: Set the population size, the maximum number of iterations to 50, the initial elite memory bank capacity to 20, the adversarial learning probability to 0.3, and the cooperative foraging weight to 0.1.

[0072] Step 3: Optimize the loop execution. Construct a DF-ResNet+FDAM network based on the current population individual encoding, train it quickly using training data, and evaluate its fitness on the validation set. Next, update the population position using a multi-stage adaptive mechanism and improve the SGA algorithm by applying adversarial learning, elite memory, and cooperative foraging strategies. Finally, optimize the network hyperparameters using the improved SGA algorithm to obtain the current optimal solution.

[0073] Step 4: Determine the termination condition. Check if the maximum number of iterations or fitness convergence has been reached. If not, return to Step 3 to continue optimization; if so, terminate the search and output the optimal network structure parameters.

[0074] Step 5: Construct the final DF-ResNet-FDAM network using the optimal structural parameters obtained from ISGA search, and train it fully on the complete training set.

[0075] Step 6: Evaluate the fault diagnosis performance of the optimal model on the test set and output the final results.

[0076] Optionally, based on the trained fault diagnosis model, the following steps of multi-data source validation and robustness assessment are also performed: Step S313: Conduct the first set of verification experiments on the bearing fault data source of Jiangnan University, covering three speed conditions: 600rpm, 800rpm, and 1000rpm. Step S314: Perform cross-dataset generalization assessment on the Western Reserve University CWRU bearing fault data source, including various working conditions with different damage sizes; Step S315: Design a module ablation comparison experiment, remove the FDM module, flexible structure and ISGA improvement strategy respectively, and verify the performance contribution of each component. Step S316: Perform a model robustness assessment by changing the data structure to examine the model's ability to withstand disturbances; Step S317: Set up benchmark comparison experiments and select mainstream deep learning architectures such as ResNet, WDCNN, and Transformer for performance comparison.

[0077] This invention uses the bearing fault dataset from Jiangnan University for verification and analysis. This dataset creates bearing defects through manual cutting, and the bearing types include four categories: outer ring fault, inner ring fault, rolling element fault, and normal bearing.

[0078] Define z i For the original data, μ i and δ i These are the mean and standard deviation of the original data.

[0079] Next, the data is processed. A sliding window is used to construct samples, and the training set and test set are divided in a 7:3 ratio. The data is then preprocessed using the Z-score standardization method. This process can be represented by the following formula: .

[0080] Finally, cross-dataset validation experiments were conducted using the Western Reserve University (CWRU) bearing failure dataset.

[0081] This invention achieves intelligent diagnosis of bearing faults and automated optimization of network topology through the ISGA-DFResNet-FDAM integrated architecture, taking into account both global search performance and deep feature learning capabilities, and provides an effective technical solution for fault diagnosis of rotating machinery in complex spectrum environments.

[0082] In the above-mentioned adaptive snow goose algorithm-driven residual network bearing fault diagnosis method, all steps can be completed in Python 3.13.

[0083] The implementation of the various embodiments of this invention is based on programmed processing through a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of this invention are encapsulated into various modules. Based on this reality, and building upon the above embodiments, this invention provides an adaptive snow goose algorithm-driven residual network bearing fault diagnosis system. This system is used to execute an adaptive snow goose algorithm-driven residual network bearing fault diagnosis method from the above method embodiments.

[0084] The system includes: a data acquisition module for acquiring bearing fault data to be diagnosed; and a fault diagnosis module for inputting the bearing fault data to be diagnosed into a trained fault diagnosis model and outputting fault diagnosis results. The training of the fault diagnosis model includes: constructing a bearing fault sample dataset; constructing the fault diagnosis model includes: a frequency domain attention module for performing frequency domain awareness attention enhancement on the input bearing fault samples; a flexible residual network for performing deep feature extraction on the enhanced features through dual flexible residual blocks, wherein each dual flexible residual block contains two cascaded flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. Simultaneously, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to achieve information exchange and feature enhancement between differentiated residual paths; and introduces an adaptive snow goose algorithm to optimize the network structure parameters of the fault diagnosis model, and based on the optimal network structure parameters, trains the model on the constructed sample dataset to output the trained fault diagnosis model.

[0085] This invention provides an adaptive snow goose algorithm-driven residual network bearing fault diagnosis system. Addressing the challenges of intelligently selecting topology configuration parameters for deep neural networks and insufficient extraction of key frequency domain features of vibration signals in rotating machinery bearing fault diagnosis, this system employs several modules. Through deep coupling of the adaptive snow goose algorithm and a frequency-domain attention-enhanced dual-flexible residual network, and the synergistic utilization of time-frequency domain information, it achieves high-accuracy identification and intelligent discrimination of bearing degradation types, effectively solving the problems of inflexible network architecture and insufficient extraction of frequency domain information.

[0086] It should be noted that the system embodiments provided by the present invention are used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The only difference is that corresponding functional modules are set. The principle is basically the same as that of the above system embodiments provided by the present invention. As long as those skilled in the art can improve the modules in the above system embodiments by referring to the specific technical solutions in other method embodiments and combining technical features to obtain corresponding technical means and technical solutions composed of these technical means, on the basis of the above system embodiments, and on the premise of ensuring the practicality of the technical solutions, they can obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments.

[0087] Based on the same inventive concept as any of the foregoing embodiments, this embodiment of the invention also provides an adaptive snow goose algorithm-driven residual network bearing fault diagnosis device, including a memory and a processor. The memory stores program instructions that are executed by the processor, and the processor calls the program instructions to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method.

[0088] Based on the same inventive concept as any of the foregoing embodiments, this embodiment of the invention also provides a non-transitory computer-readable storage medium storing computer instructions that cause the computer to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method.

[0089] In summary, this invention is applicable to solving problems such as rigid topological structures in deep neural networks, superficial mining of vibration signal spectral features, and difficulty in intelligently matching network configuration parameters. First, this invention constructs a flexible residual network modeling system, designs a scalable convolutional layer configuration mechanism to achieve elastic adjustment of network depth, establishes a dynamic skip connection selection strategy to adapt to multi-path transmission of gradient information, employs a dual-path residual block cascade structure to enhance the nonlinear transformation capability of features, and combines cross-path cross-connections and attention mechanisms to achieve effective integration of multi-level features. Second, it proposes a frequency domain-aware attention module, which completes bidirectional mapping of time and frequency space through fast Fourier transform operations, designs an amplitude-phase decoupling analysis strategy to deeply analyze the multi-dimensional discriminative attributes of the spectrum, and constructs a dual-layer attention architecture of frequency channel weighting and frequency domain spatial focusing. Third, it designs a multi-stage evolutionary snow goose algorithm, through… The optimization stage of the dynamic identification algorithm based on the joint criterion of population diversity and fitness improvement is explored. Opposite solution generation and elite guidance strategies are introduced to broaden the search space coverage. A memory storage and weighted utilization mechanism for historical high-quality solutions is established, and the collective intelligence of the algorithm is enhanced by integrating inter-individual cooperation strength calculation and collective foraging behavior simulation. Finally, an integrated diagnostic architecture of ISGA-DFResNet-FDAM is constructed to achieve intelligent exploration of network topology parameters and joint training of deep model weights. Through hierarchical collaboration between upper-layer ISGA topology search and lower-layer residual network deep learning, accurate classification of multiple types of damage, such as bearing inner ring defects, outer ring cracks, and rolling element spalling, is achieved.

[0090] This invention can effectively improve the classification accuracy of bearing fault diagnosis under strong noise conditions, enhance the adaptive configuration capability and multi-condition generalization recognition level of the deep learning architecture, and ensure predictive maintenance and safe and reliable operation of rotating equipment. Its key points are:

[0091] 1. Multi-strategy collaborative improvements to the ISGA algorithm include a multi-stage evolutionary discrimination mechanism that dynamically switches search strategies based on population status, an opposing solution generation strategy to broaden the search range of the solution space, an elite memory management mechanism to preserve and utilize historical high-quality solutions, and a cooperative foraging mechanism to enhance the information exchange capabilities between individuals.

[0092] 2. Adaptive topology design of dual-flexible residual networks, including scalable convolutional layers to achieve dynamic configuration of network depth, dynamic skip connection position selection to adapt to feature transfer paths, dual-path flexible residual block concatenation to enhance nonlinear fitting ability, and cross-path cross-connection and attention fusion to strengthen multi-scale features, enabling the network to adaptively adjust topology parameters according to fault vibration characteristics.

[0093] 3. A frequency domain awareness attention enhancement module based on fast Fourier transform provides an accurate spectral discrimination basis for deep feature extraction.

[0094] The terms “comprising” and “having”, and any variations thereof, in the specification, claims, and accompanying drawings of this invention are intended to cover a non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A residual network bearing fault diagnosis method driven by an adaptive snow goose algorithm, characterized in that, include: Obtain the bearing fault data to be diagnosed; The bearing fault data to be diagnosed is input into the trained fault diagnosis model, and the fault diagnosis result is output; wherein, the training of the fault diagnosis model includes: Construct a bearing failure sample dataset; A fault diagnosis model is constructed, including: a frequency domain attention module, used to enhance the frequency domain perception attention features of the input bearing fault samples; and a flexible residual network, used to extract deep features from the enhanced features through dual flexible residual blocks. The dual flexible residual blocks contain two serial flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. At the same time, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to realize information exchange and feature enhancement between differentiated residual paths. An adaptive snow goose algorithm is introduced to optimize the network structure parameters of the fault diagnosis model. Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, and the trained fault diagnosis model is output.

2. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 1, characterized in that, The construction of the adaptive snow goose algorithm includes: Initialize the Snow Goose algorithm population, with each individual's encoding containing network structure parameters; Establish a multi-stage evolutionary discrimination criterion function, and dynamically identify the evolutionary stage of the algorithm based on the population dispersion index and fitness increment; A strategy for generating complementary solutions is introduced to generate complementary solutions for the current candidate solutions and combine this with guidance from elite individuals, thereby broadening the search range of the solution space. Design an elite solution memory bank management mechanism to store high-quality solutions from historical iterations and guide population evolution through a weighted selection strategy; Construct a cooperative foraging mechanism among individuals, calculate the cooperative strength index among individuals and incorporate it into the position update equation to enhance the collective search ability of the population; The flight strategy is adaptively switched between the V-formation flight strategy in the exploration phase and the linear flight strategy in the development phase based on the results of multi-stage evolution.

3. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 2, characterized in that, An adaptive snow goose algorithm is introduced to optimize the network's structural parameters, including: Set the execution parameters of the adaptive snow goose algorithm, including the number of individuals in the population, the maximum number of generations, the capacity of the elite memory bank, and the probability of adversarial learning; For each individual in the current population, construct the network topology of the corresponding fault diagnosis model; The constructed network topology is iteratively trained on the training set samples; The classification performance of the network topology is evaluated on the validation set samples, and the recognition accuracy is used as the fitness value of the individual. The population evolution operation of the adaptive snow goose algorithm is performed, and multi-stage adaptation, opposition solution generation, elite memory guidance and cooperative foraging strategies are applied. When the maximum number of generations or fitness convergence criteria are met, the parameter configuration scheme of the globally optimal individual is output.

4. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 3, characterized in that, Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, including: The network topology of the fault diagnosis model is constructed based on the optimal network structure parameters obtained by the adaptive snow goose algorithm. Configure the hyperparameters for network training; The optimal network structure parameters are iteratively trained on the training sample set, and the performance of the trained network model is evaluated on the test sample set. Output fault identification results, including overall accuracy, macro average precision, macro average recall, and macro average F1 score.

5. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 3, characterized in that, Before performing iterative optimization, the network structure parameter search space is bounded as follows: Define the search range for the number of convolutional layers for the first and second flexible residual blocks; Determine the search interval for the number of jump connections between the first flexible residual block and the second flexible residual block; A set of candidate values ​​for configuring the number of convolutional kernels.

6. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 1, characterized in that, The construction of the frequency domain attention module also includes: Perform a Fast Fourier Transform on the feature tensor of the input bearing fault sample; The amplitude and phase of the frequency domain complex result after the fast Fourier transform are decoupled and separated to obtain the spectral amplitude matrix and phase angle matrix respectively. Design a frequency channel attention weighting mechanism to achieve feature selection in the channel dimension by learning the importance coefficients of different frequency components; Construct a frequency domain spatial attention focusing mechanism to focus on salient regions at different frequency positions in the spectral amplitude matrix; Perform frequency domain feature parameterization enhancement and phase compensation operations to optimize the spectral representation through a trainable adaptive filter; The enhanced frequency domain representation is reconstructed back to the time domain using the inverse fast Fourier transform, and the original time domain features are integrated through a gating fusion mechanism to obtain the final enhanced features.

7. The adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to claim 1, characterized in that, Construct a bearing failure sample dataset, including: Collect bearing vibration data under various speed conditions; The original vibration signal is serialized and segmented using a sliding window method. The window width and sliding step are set to generate a sample sequence of fixed length. Perform Z-score normalization on the segmented sample sequences; The normalized samples are split into datasets, and the ratio of training set to test set is set.

8. An adaptive snow goose algorithm-driven residual network bearing fault diagnosis system, used to implement the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire bearing fault data to be diagnosed. The fault diagnosis module is used to input the bearing fault data to be diagnosed into the trained fault diagnosis model and output the fault diagnosis result; wherein, the training of the fault diagnosis model includes: Construct a bearing failure sample dataset; A fault diagnosis model is constructed, including: a frequency domain attention module, used to enhance the frequency domain perception attention features of the input bearing fault samples; and a flexible residual network, used to extract deep features from the enhanced features through dual flexible residual blocks. The dual flexible residual blocks contain two serial flexible residual paths and the number of convolutional layers in each flexible residual block is adjustable. At the same time, the flexible residual network also introduces a dynamic selection mechanism for skip connection positions to determine the access and output positions of the flexible residual paths, and introduces cross-path cross-connection weights and a global average pooling attention mechanism to realize information exchange and feature enhancement between differentiated residual paths. An adaptive snow goose algorithm is introduced to optimize the network structure parameters of the fault diagnosis model. Based on the optimal network structure parameters, the model is trained on the constructed sample dataset, and the trained fault diagnosis model is output.

9. A residual network bearing fault diagnosis device driven by an adaptive snow goose algorithm, characterized in that, The method includes a memory and a processor, wherein the memory stores program instructions that are executed by the processor, and the processor invokes the program instructions to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the adaptive snow goose algorithm-driven residual network bearing fault diagnosis method as described in any one of claims 1 to 7.

Images