Method and system for mechanical and electrical equipment fault diagnosis based on neural network

Abstract

This application discloses a method and system for fault diagnosis of electromechanical equipment based on neural networks. The method includes: acquiring normal data and fault data of different fault types of electromechanical equipment, and performing normalization operations; performing Fourier transform on the normal data and the fault data to convert them into frequency domain data; converting the frequency domain data into a graph structure, and dividing the data into training set and test set according to a preset ratio; training the training set and test set through a graph Chebyshev network model to obtain features of different fine granularities and concatenating them; importing the features into the trained model to view the prediction results, and visualizing the prediction results. This application solves the problem of feature extraction and noise reduction for long and short time series, and can learn local features from many surrounding nodes. It combines local features with long-term features through pyramid attention to achieve the extraction of long and short-term features.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis of electromechanical equipment, and in particular to a method and system for fault diagnosis of electromechanical equipment based on neural networks. Background Technology

[0002] With the rapid development of the manufacturing industry, electromechanical equipment has become a major component. As the service life of this equipment increases, its failure rate also rises, leading to increased maintenance costs, repair cycles, and downtime costs. Therefore, research into fault diagnosis for electromechanical equipment is imperative to address the reduced production efficiency or even shutdowns caused by sudden equipment failures.

[0003] Traditional maintenance methods, which rely on reactive approaches to repair after a machine malfunction, lead to a passive maintenance environment. Furthermore, the increasing complexity and sophistication of modern machinery significantly increases the time and labor costs of troubleshooting and repair. In automated production lines with a vast number of machines, the cost of relying on traditional, reactive maintenance methods is immeasurable. Existing time-series fault diagnosis methods suffer from several problems: 1) Extracting short-term features from local data results in incomplete feature extraction and unsatisfactory results. 2) While using multi-layer convolutional networks, fully connected attention mechanisms, or recurrent neural networks with long short-term memory to extract features from relatively comprehensive data, these methods are costly and inefficient. Additionally, while existing methods offer greater flexibility in selecting different attention mechanisms and exhibit good noise resistance, they also have drawbacks: a lack of processing and denoising methods for time-series data; the simultaneous use of dot products (DP), single-layer neural networks, and hybrid approaches when using attention models; and the need to combine all neighboring nodes for feature extraction at each node, leading to excessive space and time complexity, especially for time-series data where combining too much data is unnecessary. They also lack practical application in long-term and short-term fault prediction.

[0004] Therefore, the aforementioned technical problems in the relevant technologies urgently need to be solved. Summary of the Invention

[0005] This application aims to solve one of the technical problems in related technologies. To this end, embodiments of this application provide a method and system for diagnosing electromechanical equipment faults based on neural networks, which can perform fault diagnosis on electromechanical equipment based on neural networks.

[0006] According to one aspect of the embodiments of this application, a method for diagnosing electromechanical equipment faults based on neural networks is provided, the method comprising:

[0007] Acquire normal data and fault data of different fault types from electromechanical equipment, and perform normalization operations;

[0008] The normal data and the fault data are subjected to Fourier transform and converted into frequency domain data;

[0009] The frequency domain data is transformed into a graph structure, and the data is divided into a training set and a test set according to a preset ratio;

[0010] The training set and the test set are trained using a Tuchebyshev network model to obtain features of different fine granularities, which are then concatenated.

[0011] Import the features into the trained model to view the prediction results, and then visualize the prediction results.

[0012] In one embodiment, after obtaining and stitching together features of different fine granularities, the method further includes:

[0013] The pyramid attention mechanism is used to process the features of different fine granularities to obtain the prediction results of the training data;

[0014] Update the weights and backpropagate until training is complete.

[0015] In one embodiment, the normalization operation includes:

[0016] The normal data and the fault data are normalized, and the result after normalization is controlled within the range of [0,1].

[0017] In one embodiment, the normal data and the fault data are subjected to Fourier transform to convert them into frequency domain data. The Fourier transform formula includes:

[0018]

[0019] The original data is partially output as x′(t), the original data is x(t), and the length of the original data is L.

[0020] In one embodiment, the Tuchebyshev network model employs a Graph Laplace matrix.

[0021] In one embodiment, the training set and the test set are trained using a Tuchebyshev network model to obtain features of different fine granularities, which are then concatenated, including:

[0022] Use faulty data as input for multi-scale convolution;

[0023] Features between nodes are extracted using the Laplacian matrix.

[0024] According to one aspect of an embodiment of this application, a fault diagnosis system for electromechanical equipment based on neural networks is provided, the system comprising:

[0025] The first module is used to acquire normal data and fault data of different fault types of electromechanical equipment, and perform normalization operations.

[0026] The second module is used to perform Fourier transform on the normal data and the fault data to convert them into frequency domain data;

[0027] The third module is used to convert the frequency domain data into a graph structure and divide the data into a training set and a test set according to a preset ratio.

[0028] The fourth module is used to train the training set and test set using a Tuchebyshev network model to obtain features of different fine granularities and then concatenate them.

[0029] The fifth module is used to import features into the trained model to view the prediction results and visualize the prediction results.

[0030] In one embodiment, the system includes:

[0031] The sixth module is used to process the features of different fine granularities through the pyramid attention mechanism to obtain the prediction results of the training data;

[0032] The seventh module is used to update the weights and backpropagate until training is complete.

[0033] In one embodiment, the system includes an eighth module, configured to normalize the normal data and the fault data, and control the result after normalization within the range of [0,1].

[0034] According to one aspect of an embodiment of this application, a fault diagnosis system for electromechanical equipment based on neural networks is provided, the system comprising:

[0035] At least one processor;

[0036] At least one memory for storing at least one program;

[0037] When at least one of the programs is executed by at least one of the processors, the neural network-based electromechanical equipment fault diagnosis method as described in the preceding embodiments is implemented.

[0038] The beneficial effects of the neural network-based electromechanical equipment fault diagnosis method and system provided in this application are as follows: The method of this application includes: acquiring normal data and fault data of different fault types of electromechanical equipment, and performing normalization operations; performing Fourier transform on the normal data and the fault data to convert them into frequency domain data; converting the frequency domain data into a graph structure, and dividing the data into training set and test set according to a preset ratio; training the training set and test set through a graph Chebyshev network model to obtain features of different fine granularities and concatenating them; importing the features into the trained model to view the prediction results, and visualizing the prediction results. This application solves the feature extraction and noise reduction of long and short time series: by converting local time series features into frequency domain data to construct a directed weighted graph, nodes can learn local features through many surrounding nodes, which is different from the noise problem caused by relying on fewer preceding and following time series features. By combining local features with long-term features through pyramid attention, the extraction of long and short-term features is achieved.

[0039] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating a neural network-based electromechanical equipment fault diagnosis method provided in this application embodiment;

[0042] Figure 2 A flowchart illustrating the specific implementation of the neural network-based electromechanical equipment fault diagnosis method provided in this application embodiment;

[0043] Figure 3 A schematic diagram of a neural network-based electromechanical equipment fault diagnosis system provided in an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of another electromechanical equipment fault diagnosis system based on neural networks, provided as an embodiment of this application. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0046] The terms "first," "second," "third," and "fourth," etc., used in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0047] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0048] With the rapid development of the manufacturing industry, electromechanical equipment has become a major component. As the service life of this equipment increases, its failure rate also rises, leading to increased maintenance costs, repair cycles, and downtime costs. Therefore, research into fault diagnosis for electromechanical equipment is imperative to address the reduced production efficiency or even shutdowns caused by sudden equipment failures.

[0049] Traditional maintenance methods, which rely on reactive approaches to repair after a machine malfunction, lead to a passive maintenance environment. Furthermore, the increasing complexity and sophistication of modern machinery significantly increases the time and labor costs of troubleshooting and repair. In automated production lines with a vast number of machines, the cost of relying on traditional, reactive maintenance methods is immeasurable. Existing time-series fault diagnosis methods suffer from several problems: 1) Extracting short-term features from local data results in incomplete feature extraction and unsatisfactory results. 2) While using multi-layer convolutional networks, fully connected attention mechanisms, or recurrent neural networks with long short-term memory to extract features from relatively comprehensive data, these methods are costly and inefficient. Additionally, while existing methods offer greater flexibility in selecting different attention mechanisms and exhibit good noise resistance, they also have drawbacks: a lack of processing and denoising methods for time-series data; the simultaneous use of dot products (DP), single-layer neural networks, and hybrid approaches when using attention models; and the need to combine all neighboring nodes for feature extraction at each node, leading to excessive space and time complexity, especially for time-series data where combining too much data is unnecessary. They also lack practical application in long-term and short-term fault prediction.

[0050] To address the aforementioned problems, this application proposes a method and system for fault diagnosis of electromechanical equipment based on neural networks. The steps of this application include: 1) collecting several points of equipment fault data and normal monitoring data (such as speed, voltage, and location); 2) preprocessing the data into a graph data node format, and converting the time-domain data into frequency-domain data using Fast Fourier Transform; 3) constructing a Tuchebyshev neural network and using pyramidal attention to extract long-term and short-term data features; 4) dividing the data into training and testing sets, training the model's attention weight parameters using the training set, and then using the testing set to check the fault classification accuracy. This invention utilizes the high robustness and adaptability of the Tuchebyshev neural network, combined with a tree-like attention mechanism, to more efficiently extract long-term and short-term features of time-series data with lower time and space complexity. This improves the efficiency of fault diagnosis for electromechanical equipment.

[0051] Specifically, such as Figure 1 As shown, this application proposes a method for fault diagnosis of electromechanical equipment based on neural networks, comprising:

[0052] S101. Obtain normal data of electromechanical equipment and fault data of different fault types, and perform normalization operation.

[0053] S102. Perform Fourier transform on the normal data and the fault data to convert them into frequency domain data.

[0054] S103. The frequency domain data is converted into a graph structure, and the data is divided into a training set and a test set according to a preset ratio.

[0055] S104. The training set and test set are trained using a Tuchebyshev network model to obtain features of different fine granularities and then spliced ​​together.

[0056] S105. Import the features into the trained model, view the prediction results, and visualize the prediction results.

[0057] Optionally, after obtaining and concatenating features of different fine granularities, the method further includes: processing the features of different fine granularities through a pyramid attention mechanism to obtain the prediction results of the training data; updating the weights and backpropagating until training is completed.

[0058] In addition, this embodiment also includes a normalization operation, including: normalizing the normal data and the fault data, and controlling the result after the normalization operation within the range of [0,1].

[0059] In this embodiment, the normal data and the fault data are subjected to Fourier transform to convert them into frequency domain data. The Fourier transform formula includes:

[0060]

[0061] The original data is partially output as x′(t), the original data is x(t), and the length of the original data is L.

[0062] In this embodiment, the Graph Chebyshev network model uses a graph Laplace matrix.

[0063] It should be noted that the training set and test set are trained using a Tuchebyshev network model to obtain features of different fine granularities and then concatenated, including: using fault data as input for multi-scale convolution; and extracting features between nodes using the Laplacian matrix.

[0064] like Figure 2 As shown, the method for fault diagnosis of electromechanical equipment based on neural networks proposed in this application includes the following steps in its specific application:

[0065] (1) Collect normal data of electromechanical equipment and monitoring data of different fault types, and normalize them to prevent some data from affecting the overall effect.

[0066] (2) The time-domain data is converted into frequency-domain data after Fourier transform.

[0067] (3) Transform the data into a graph structure and divide the data into training and testing sets in an 8:2 ratio.

[0068] (4) Load the training set into the Chebyshev network, obtain features of different fine granularities through model training, and then splice them together.

[0069] (5) Add a pyramid attention mechanism to process the features extracted by the model, obtain the prediction results of the training data, update the weights, and backpropagate until the training epoch is completed.

[0070] (6) Import the test data into the trained model to view the prediction effect and visualize it.

[0071] The relevant algorithm formulas used in this invention are described below:

[0072] Data processing related functions:

[0073] The data is normalized to keep the processed result within the range [0,1]. Let the original data be x, the maximum value be max, and the minimum value be min. The normalized output is y, and the normalization is shown in formula (1).

[0074]

[0075] The normalized data is subjected to a fast Fourier transform to convert it into frequency domain data, where the original data is x(t), ω is the harmonic order, and the Fourier series of x(t) is shown in formula (2):

[0076]

[0077] The Fourier transform after converting it to complex form is shown in formula (3):

[0078]

[0079] Since the transformed data is symmetrical, it is then halved and output as x′(t), where L is the length of the original data. The transformation is as shown in formula (4):

[0080]

[0081] Pyramid-attention-enhanced GraphChebyshev Neural Network:

[0082] Network model structure as follows Figure 2 As shown, the fault sample x′(t) after the above processing is used as the input of multi-scale convolution. The purpose of multi-scale convolution is to extract multi-fine-grained features of the fault sample graph structure, which facilitates subsequent fault diagnosis. The network mainly uses the graph Laplacian matrix L as shown in formula (5), where I n Let A be the identity matrix, D be the degree matrix, and A be the edge matrix. Features between nodes are extracted using the Laplacian matrix.

[0083] L = I n -D -1 / 2 AD -1 / 2 (5)

[0084] The Laplace matrix is ​​normalized to U, and g is extracted using Chebyshev polynomials. θ Filter approximation, Let T be the characteristic matrix, k be the order of the Chebyshev polynomial, and T be the characteristic matrix. k Chebyshev polynomials are expressed as in formulas (6)-(7):

[0085]

[0086]

[0087] To improve the feature extraction and noise reduction of fault diagnosis data, a Tuchebyshev network is used instead of a traditional sequence-based neural network. This method aggregates multiple neighboring nodes of a node into a graph, allowing nodes to learn each other's features. Compared to traditional feature extraction methods that rely on a small number of sequential node features and learn incorrect features on noisy data, leading to poor extraction of correct features and ineffective noise reduction, this method overcomes these problems. Furthermore, it employs multi-fine-grained convolution. The input data x′(t) is processed by Tuchebyshev convolution at scale C to extract features and output Che. n :

[0088] Che n =Checonv(x'1, ..., x') n (8)

[0089] The input to the pyramid attention is obtained by concatenating convolutional features of different fine granularities.

[0090] G in Where B is the batch size and D is the feature dimension:

[0091]

[0092] G is used to represent different fine-grained features. in If the input is the pyramid attention mechanism, and a directed graph is formed using the features of C nodes, then... A node only needs to focus on its neighbors at the same scale. (Including itself), and its C child nodes (with itself as the parent node). and its parent node Finally, they form a set of features. s = 1, ..., S represents the scale (the number of layers from bottom to top), and l represents the l-th node at that scale.

[0093]

[0094] In this way, a node only needs to focus on the features of its three adjacent scale neighbors, thus the maximum path length is controlled at O(1). In contrast, traditional methods, being sequence-based, require a receptive field length of at least O(L) to obtain the features of the first L nodes. Fine-grained features are aggregated into coarse-grained features and propagated in the tree-like path. The time and space complexity of the model are linearly related to the number of nodes, controlled at O(N), where N is the number of nodes. After a representative feature is propagated at most N times in the model, all nodes can learn its features. Traditional fault diagnosis methods have a time complexity ranging from O(LN) to O(L). 2 Achieve low complexity between ) and ).

[0095] The attention formula can be expressed as y i This can serve as a basis for fault diagnosis and classification, thus realizing a fault diagnosis method with efficient feature extraction, good noise reduction, and low network model complexity.

[0096]

[0097] This application addresses feature extraction and noise reduction for both short-term and long-term time series: it constructs a directed weighted graph by converting local time-series features into frequency domain data. This allows nodes to learn local features from their surrounding nodes, avoiding the noise problem associated with relying on fewer preceding and following time-series features. Furthermore, it combines local and long-term features through pyramidal attention, achieving the extraction of both short-term and long-term features. This application implements a low-complexity attention model: it constructs pyramid-shaped features from each local graph, adds a small number of parameters to obtain features of varying coarseness, and each node only learns its child nodes, C neighboring nodes, and its parent node. This significantly reduces the number of node features to focus on in long-term predictions while still transmitting long-term features through pyramidal attention. This low-complexity approach improves accuracy, enabling an efficient fault diagnosis model.

[0098] Furthermore, this application can use other network models instead of the Chebyshev network to extract features. Different network models will result in different feature extraction methods and thus alter the effectiveness of feature extraction. Therefore, a network model more suitable for this dataset can be selected. Noise reduction processing is implemented during data collection to reduce noise from the data source, thereby reducing the number of erroneous features learned by the algorithm.

[0099] Furthermore, this application also proposes a fault diagnosis system for electromechanical equipment based on neural networks, such as... Figure 3 As shown, the system includes:

[0100] The first module 301 is used to acquire normal data of electromechanical equipment and fault data of different fault types, and to perform normalization operations.

[0101] The second module 302 is used to perform Fourier transform on the normal data and the fault data to convert them into frequency domain data;

[0102] The third module 303 is used to convert the frequency domain data into a graph structure and divide the data into a training set and a test set according to a preset ratio.

[0103] The fourth module 304 is used to train the training set and the test set through a Tuchebyshev network model to obtain features of different fine granularities and then splice them together.

[0104] The fifth module 305 is used to import features into the trained model to view the prediction results and visualize the prediction results.

[0105] The system further includes: a sixth module, used to process the features of different fine granularities through a pyramid attention mechanism to obtain the prediction results of the training data; and a seventh module, used to update the weights and backpropagate until training is completed.

[0106] Furthermore, this application also proposes a fault diagnosis system for electromechanical equipment based on neural networks, such as... Figure 4 As shown, the system includes:

[0107] At least one processor 401;

[0108] At least one memory 402, the memory 402 being used to store at least one program;

[0109] When at least one of the programs is executed by at least one of the processors 401, the neural network-based electromechanical equipment fault diagnosis method as described in the preceding embodiments is implemented.

[0110] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0111] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical system and / or software module, or one or more functions and / or features may be implemented in a separate physical system or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the system disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0112] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0113] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0114] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0115] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for fault diagnosis of electromechanical equipment based on neural networks, characterized in that, The method includes: Acquire normal data and fault data of different fault types from electromechanical equipment, and perform normalization operations; The normal data and the fault data are subjected to Fourier transform and converted into frequency domain data; The frequency domain data is transformed into a graph structure, a directed weighted graph is constructed, and the data is divided into a training set and a test set according to a preset ratio; The training set and the test set are trained using a Tuchebyshev network model to obtain features of different fine granularities, which are then concatenated. Import the features into the trained model to view the prediction results, and then visualize the prediction results; After obtaining and stitching together features of different fine granularities, the method further includes: Different fine-grained features are concatenated and used as input to the pyramid attention mechanism. The pyramid attention mechanism processes the different fine-grained features to obtain the prediction results of the training data. Update the weights and backpropagate until training is complete; The process of concatenating different fine-grained features and using them as input to the pyramid attention mechanism, and then processing these features through the pyramid attention mechanism, includes: The input to the pyramid attention is obtained by concatenating convolutional features of different fine granularities. By using different fine-grained features as input to the pyramid attention mechanism, and employing C node features to form a directed graph, then... A node only needs to focus on its neighbors at the same scale. C child nodes of it as the parent node and its parent node Finally, a set of features is formed. Indicates scale. Indicates the first at this scale One node; in, It is an edge matrix. It is a Laplace matrix.

2. The method for fault diagnosis of electromechanical equipment based on neural networks according to claim 1, characterized in that, Normalization operations include: The normal data and the fault data are normalized, and the result after normalization is controlled within the range of [0,1].

3. The method for fault diagnosis of electromechanical equipment based on neural networks according to claim 1, characterized in that, The normal data and the fault data are subjected to Fourier transform to convert them into frequency domain data. The Fourier transform formula includes: The original data half-output is The original data is x(t), and the length of the original data is L.

4. The method for fault diagnosis of electromechanical equipment based on neural networks according to claim 1, characterized in that, The Tuchebyshev network model uses a Graph Laplace matrix.

5. The method for fault diagnosis of electromechanical equipment based on neural networks according to claim 1, characterized in that, The training and test sets are trained using a Tuchebyshev network model to obtain features of different fine granularities, which are then concatenated, including: Use faulty data as input for multi-scale convolution; Features between nodes are extracted using the Laplacian matrix.

6. A fault diagnosis system for electromechanical equipment based on neural networks, characterized in that, The system includes: The first module is used to acquire normal data and fault data of different fault types of electromechanical equipment, and perform normalization operations. The second module is used to perform Fourier transform on the normal data and the fault data to convert them into frequency domain data; The third module is used to convert the frequency domain data into a graph structure and divide the data into a training set and a test set according to a preset ratio. The fourth module is used to train the training set and test set using a Tuchebyshev network model to obtain features of different fine granularities and then concatenate them. The fifth module is used to import features into the trained model to view the prediction results and visualize the prediction results; The sixth module is used to input different fine-grained features as input to the pyramid attention mechanism, and to process the different fine-grained features through the pyramid attention mechanism to obtain the prediction results of the training data. The seventh module is used to update the weights and backpropagate until training is complete. The process of concatenating different fine-grained features and using them as input to the pyramid attention mechanism, and then processing these features through the pyramid attention mechanism, includes: The input to the pyramid attention is obtained by concatenating convolutional features of different fine granularities. By using different fine-grained features as input to the pyramid attention mechanism, and employing C node features to form a directed graph, then... A node only needs to focus on its neighbors at the same scale. C child nodes of it as the parent node and its parent node Finally, a set of features is formed. Indicates scale. Indicates the first at this scale One node; in, It is an edge matrix. It is a Laplace matrix.

7. The electromechanical equipment fault diagnosis system based on neural networks according to claim 6, characterized in that, The system includes: The eighth module is used to normalize the normal data and the fault data, and to control the result after normalization within the range of [0,1].

8. A fault diagnosis system for electromechanical equipment based on neural networks, characterized in that, The system includes: At least one processor; At least one memory for storing at least one program; The method for diagnosing electromechanical equipment based on neural networks as described in any one of claims 1-5 is implemented when at least one of the programs is executed by at least one of the processors.

Images