Rotating machine fault diagnosis method and device, fault diagnosis equipment and storage medium

Abstract

The invention relates to a rotating machinery fault diagnosis method and device, fault diagnosis equipment and a storage medium, and belongs to the technical field of fault diagnos.The rotating machinery fault diagnosis method comprises the steps that vibration signal data of target rotating machinery equipment are acquired and preprocessed; converting the preprocessed vibration signal data into multi-scale data based on a CEEMDAN algorithm; and taking the multi-scale data and the preprocessed vibration signal data as input of a fault diagnosis model to obtain a fault diagnosis result output by the fault diagnosis model. Accurate diagnosis of unknown compound faults of the rotating machinery can be realized.

Description

Technical Field

[0001] This invention relates to the field of equipment fault diagnosis technology, and in particular to a method, apparatus, fault diagnosis equipment and storage medium for diagnosing faults in rotating machinery. Background Technology

[0002] With the rapid development of modern industry, the use of large-scale machinery and equipment is becoming increasingly widespread. Among these, rotating machinery, as a key component, directly affects the overall operating efficiency and safety of the equipment. However, due to long-term operation, rotating machinery is prone to degradation, wear, and even failure. If faulty components are not maintained in a timely manner, a single fault may escalate into a complex fault, potentially leading to equipment downtime or even personal injury. Therefore, complex fault diagnosis technology for rotating machinery has become an important means to ensure the safe operation of mechanical equipment and reduce maintenance costs and economic losses.

[0003] Existing fault diagnosis technologies for rotating machinery typically only diagnose known single fault types, making it difficult to effectively handle unknown fault types that arise during the degradation process of mechanical equipment, especially compound faults. Compound faults, due to their inclusion of multiple individual fault characteristics and the complex coupling relationships between these fault types, pose a significant challenge to intelligent fault diagnosis technologies.

[0004] Therefore, how to accurately diagnose unknown complex faults using existing data has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, apparatus, equipment and storage medium for diagnosing faults in rotating machinery, so as to solve the problem that existing fault diagnosis technologies for rotating machinery are unable to accurately diagnose complex faults.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a method for diagnosing faults in rotating machinery, comprising: Acquire vibration signal data of the target rotating mechanical equipment and perform preprocessing; The preprocessed vibration signal data is converted into multi-scale data based on the CEEMDAN algorithm. The multi-scale data and the preprocessed vibration signal data are used as inputs to the fault diagnosis model to obtain the fault diagnosis results output by the fault diagnosis model. The fault diagnosis model consists of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels. The dynamic feedforward Transformer module consists of a global attention module and a dynamic deep convolution module.

[0007] In one possible implementation, the fault diagnosis model is composed of a first multi-scale feedforward depth convolution module, a dynamic feedforward Transformer module, a second multi-scale feedforward depth convolution module, a first fully connected module, and a second fully connected module connected in sequence.

[0008] In one possible implementation, the first multi-scale feedforward depth convolution module and the second multi-scale feedforward depth convolution module are composed of a first convolution module, a multi-scale depth convolution module, a feature fusion module and a second convolution module connected in sequence, for fusing multi-scale features; The first convolution module includes a convolutional layer, an activation layer, and a batch normalization layer; the multi-scale deep convolution module includes multiple deep convolutional layers with different kernel sizes; the feature fusion module includes an activation layer and a batch normalization layer; and the second convolution module includes a convolutional layer and a batch normalization layer.

[0009] In one possible implementation, the global attention module consists of a first deep convolutional module, a second deep convolutional module, a third deep convolutional module, a first linear layer, a second linear layer, a third linear layer, and a multi-head attention module, used to extract global features; The first deep convolutional module includes a deep convolutional layer, an activation layer, and a batch normalization layer. The output of the first deep convolutional module is linearly transformed by the first linear layer and used as the query vector of the multi-head attention module. The second deep convolutional module includes a deep convolutional layer and a batch normalization layer. The third deep convolutional module includes a deep convolutional layer. The second deep convolutional module and the third deep convolutional module are connected by a residual. The output obtained by the residual connection of the second deep convolutional module and the third deep convolutional module is linearly transformed by the second linear layer and used as the value vector of the multi-head attention module. After being linearly transformed by the third linear layer, it is used as the key vector of the multi-head attention module.

[0010] In one possible implementation, the dynamic deep convolution module is composed of an adaptive average pooling module, a third convolution module, a fourth convolution module, and a deep convolution layer connected in sequence, used to extract local features; The third convolutional module includes a convolutional layer, an activation layer, and a batch normalization layer. The fourth convolutional module includes a convolutional layer. The kernel parameters of the depthwise convolutional layer are determined based on the output of the fourth convolutional module.

[0011] In one possible implementation, the fault diagnosis model uses cross-entropy loss as the loss function during training and updates the model parameters based on an adaptive moment estimation algorithm.

[0012] In one possible implementation, the conversion of the preprocessed vibration signal data into multi-scale data based on the CEEMDAN algorithm includes: The preprocessed vibration signal data is converted into multiple IMFs components and a residual signal based on the CEEMDAN algorithm.

[0013] On the other hand, the present invention also provides a rotating machinery fault diagnosis device, comprising: The acquisition module is used to acquire vibration signal data of the target rotating mechanical equipment and perform preprocessing. The conversion module is used to convert the preprocessed vibration signal data into multi-scale data based on the CEEMDAN algorithm. The diagnostic module is used to take the multi-scale data and the preprocessed vibration signal data as input to the fault diagnosis model and obtain the fault diagnosis result output by the fault diagnosis model. The fault diagnosis model consists of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels. The dynamic feedforward Transformer module consists of a global attention module and a dynamic deep convolution module.

[0014] Secondly, the present invention also provides a fault diagnosis device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the rotating machinery fault diagnosis method described in any of the above implementations.

[0015] Thirdly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps in the rotating machinery fault diagnosis method described in any of the above implementations.

[0016] The beneficial effects of the present invention are as follows: The rotating machinery fault diagnosis method, device, equipment and storage medium provided in the embodiments of the present invention convert the composite fault information in the vibration signal data into fault information at multiple scales through the CEEMDAN algorithm, thereby reducing the complexity of composite fault diagnosis. Then, based on the data characteristics of multi-scale data, fault diagnosis is performed through a fault diagnosis model that can realize multi-scale feature fusion, local feature and global feature extraction, thereby achieving accurate diagnosis of unknown composite faults in rotating machinery. Attached Figure Description

[0017] Figure 1This is a schematic flowchart of an embodiment of the rotating machinery fault diagnosis method provided by the present invention; Figure 2 A schematic flowchart of an embodiment of the rotating machinery fault diagnosis process provided by the present invention; Figure 3 A schematic diagram of an embodiment of the fault diagnosis model provided by the present invention; Figure 4 A schematic diagram of an embodiment of the multi-scale feedforward depth convolution module provided by the present invention; Figure 5 A schematic diagram of an embodiment of the global attention module provided by the present invention; Figure 6 A schematic diagram of an embodiment of the dynamic depthwise convolution module provided by the present invention; Figure 7 A schematic diagram of an embodiment of the confusion matrix diagram of the model test results provided by the present invention; Figure 8 A schematic diagram of an embodiment of the rotating machinery fault diagnosis device provided by the present invention; Figure 9 This is a schematic diagram of an embodiment of the fault diagnosis device provided by the present invention. Detailed Implementation

[0018] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0021] 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 the invention. 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.

[0022] This invention provides a method, apparatus, equipment, and storage medium for diagnosing faults in rotating machinery, which will be described below.

[0023] Figure 1 This is a schematic flowchart of an embodiment of the rotating machinery fault diagnosis method provided by the present invention, as shown below. Figure 1 As shown, the methods for diagnosing rotating machinery faults include: S101. Acquire vibration signal data of the target rotating mechanical equipment and perform preprocessing.

[0024] It should be noted that, in order to diagnose faults in target rotating machinery, vibration signal data of the target rotating machinery can first be acquired. For example, a vibration acceleration sensor can be installed on the outer ring of the bearing of the target rotating machinery to collect vibration signal data. Preprocessing of the vibration signal data can involve time-series segmentation of the vibration signal. This segmentation can be performed using a sliding window, the size of which corresponds to the sampling frequency of the acceleration sensor. For example, if the sampling frequency of the acceleration sensor is 1024Hz, then the size of the sliding window can be set to 1024.

[0025] S102. Based on the CEEMDAN algorithm, the preprocessed vibration signal data is converted into multi-scale data.

[0026] It should be noted that by using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) algorithm to convert the preprocessed vibration signal data into multi-scale data, the composite fault information in the vibration signal data can be converted into fault information at multiple scales, thereby reducing the difficulty of diagnosing composite faults.

[0027] S103. The multi-scale data and the preprocessed vibration signal data are used as inputs to the fault diagnosis model to obtain the fault diagnosis results output by the fault diagnosis model.

[0028] The fault diagnosis model consists of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels. The dynamic feedforward Transformer module consists of a global attention module and a dynamic deep convolution module.

[0029] It should be noted that, in order to fully extract the characteristic information from multi-scale data, this invention proposes a fault diagnosis model consisting of a multi-scale feedforward depthwise convolutional module, a dynamic feedforward Transformer module, and a fully connected module. The multi-scale feedforward depthwise convolutional module is used to fuse multi-scale features, the dynamic feedforward Transformer module is used to extract global and local features, and finally, the characteristic information is fused through a fully connected model to output the fault diagnosis result.

[0030] In summary, the rotating machinery fault diagnosis method provided in this embodiment of the invention converts the composite fault information in vibration signal data into fault information at multiple scales using the CEEMDAN algorithm, reducing the complexity of composite fault diagnosis. Then, based on the data characteristics of multi-scale data, a fault diagnosis model capable of multi-scale feature fusion, local feature extraction, and global feature extraction is used for fault diagnosis, thereby achieving accurate diagnosis of unknown composite faults in rotating machinery.

[0031] Combination Figure 2 The rotating machinery fault diagnosis process provided by this invention specifically includes the following steps: 1. Collect vibration signal data of rotating machinery in both healthy and faulty states.

[0032] 2. Organize and store the data in a unified format, and use the health status and single fault status data as the training set, and the composite fault status data as the test set.

[0033] 3. Use fully adaptive noise set empirical mode decomposition to obtain IMFs components at different scales, and then combine them with the original vibration signal to construct a multi-scale signal to obtain training and test datasets.

[0034] 4. Construct a multi-scale feedforward deep convolutional model to fuse vibration signals and information from different IMFs components. Then, based on self-attention and adaptive deep convolution, construct a dynamic feedforward Transformer to extract features for global information and local details. Output the features of the two network models, concatenate the output global and local features, construct another multi-scale feedforward deep convolutional model to capture multi-scale information, and finally output the final result of the fused features through a fully connected layer.

[0035] 5. Train the model parameters using the training set data to learn the fault characteristics of different rotating machinery. Then, test the model using the test set data to identify fault type samples.

[0036] Combination Figure 3 In some embodiments of the present invention, the fault diagnosis model is composed of a first multi-scale feedforward depth convolution module, a dynamic feedforward Transformer module, a second multi-scale feedforward depth convolution module, a first fully connected module, and a second fully connected module connected in sequence.

[0037] Combination Figure 4 In some embodiments of the present invention, the first multi-scale feedforward depth convolution module and the second multi-scale feedforward depth convolution module are formed by sequentially connecting a first convolution module, a multi-scale depth convolution module, a feature fusion module and a second convolution module, for fusing multi-scale features. The first convolution module includes a convolutional layer, an activation layer, and a batch normalization layer; the multi-scale deep convolution module includes multiple deep convolutional layers with different kernel sizes; the feature fusion module includes an activation layer and a batch normalization layer; and the second convolution module includes a convolutional layer and a batch normalization layer.

[0038] It's important to note that: In a regular convolutional layer, the kernel convolves with all channels, while in a depthwise convolution layer, it convolves with only a single channel. Using depthwise convolution significantly reduces the number of parameters. Activation layers can improve the model's feature extraction performance. Batch normalization can effectively accelerate convergence and prevent gradient vanishing, improving the model's training efficiency and stability.

[0039] Combination Figure 5 In some embodiments of the present invention, the global attention module is composed of a first deep convolution module, a second deep convolution module, a third deep convolution module, a first linear layer, a second linear layer, a third linear layer, and a multi-head attention module, and is used to extract global features. The first deep convolutional module includes a deep convolutional layer, an activation layer, and a batch normalization layer. The output of the first deep convolutional module is linearly transformed by the first linear layer and used as the query vector of the multi-head attention module. The second deep convolutional module includes a deep convolutional layer and a batch normalization layer. The third deep convolutional module includes a deep convolutional layer. The second deep convolutional module and the third deep convolutional module are connected by a residual. The output obtained by the residual connection of the second deep convolutional module and the third deep convolutional module is linearly transformed by the second linear layer and used as the value vector of the multi-head attention module. After being linearly transformed by the third linear layer, it is used as the key vector of the multi-head attention module.

[0040] It should be noted that the query, key, and value of the multi-head attention mechanism are obtained from the input through three depthwise convolutions and linear transformations, while the query of the multi-head attention mechanism is obtained only through linear transformations. This structure greatly reduces the parameters of the global attention branch and improves the training speed of the model.

[0041] Combination Figure 6 In some embodiments of the present invention, the dynamic deep convolution module is composed of an adaptive average pooling module, a third convolution module, a fourth convolution module, and a deep convolution layer connected in sequence, and is used to extract local features. The third convolutional module includes a convolutional layer, an activation layer, and a batch normalization layer. The fourth convolutional module includes a convolutional layer. The kernel parameters of the depthwise convolutional layer are determined based on the output of the fourth convolutional module.

[0042] It should be noted that: obtaining low-dimensional feature representations from high-dimensional features through pooling operations can effectively reduce information redundancy, improve the scale invariance and rotation invariance of the model, and prevent overfitting.

[0043] In some embodiments of the present invention, the fault diagnosis model uses cross-entropy loss as the loss function during training and updates the model parameters based on the adaptive moment estimation algorithm.

[0044] It should be noted that during the training of the fault diagnosis model, cross-entropy loss can be used as the loss function to reduce the loss function value. The model parameters are updated based on the adaptive moment estimation (ADAM) algorithm. The learning ability of the model is improved through repeated iterative training, and finally a well-trained model is obtained.

[0045] In some embodiments of the present invention, the conversion of the preprocessed vibration signal data into multi-scale data based on the CEEMDAN algorithm includes: The preprocessed vibration signal data is converted into multiple IMFs components and a residual signal based on the CEEMDAN algorithm.

[0046] It should be noted that the iteration termination condition of the CEEMDAN algorithm can be that the residual signal reaches a preset energy threshold, or that the residual signal has no more obvious fluctuation trend.

[0047] To verify the effectiveness of the rotating machinery fault diagnosis method proposed in this invention, vibration acceleration data of ten different types of bearings under fault and health conditions were collected during operation. All tested bearings were tested under a load of 0 Nm and a rotational speed of 400 rpm. The collected data was preprocessed to obtain a time-domain sample dataset in the format [1024, 1, 1]. The first and second digits in the format represent the dimension of the sample, and the third digit represents the number of channels, i.e., the number of feature categories. CEEMDAN decomposition was performed on the time-domain data to obtain IMFs components at different scales. The decomposed IMFs components were merged with the time-domain data to obtain multi-scale data as input to the model. The multi-scale data was input into the model, and sensitive fault features were extracted from the input data through a layer-by-layer forward propagation method. Figure 7 As shown in the figure, the diagonal line represents the prediction accuracy, and the off-diagonal line represents the prediction error rate. The experiment used 10 types of faulty bearings and conducted ten trials, taking the average value as the experimental result. It can be seen that the overall accuracy of faulty bearing identification is above 99%, and the identification rate for each category of faults also reached 95%. The experimental results demonstrate that this model can accurately diagnose rotating machinery faults and meets the requirements for rotating machinery fault diagnosis.

[0048] This invention proposes a rotating machinery fault diagnosis method based on a dynamic feedforward Transformer using CEEMDAN. First, the input data is decomposed into Information Modeling Functions (IMFs) at different scales using CEEMDAN technology, and these IMFs are then input into a multi-branch model along with the original data to fuse the information from each component. Subsequently, a dynamic feedforward Transformer is constructed based on a self-attention mechanism and adaptive deep convolution to enhance the feature extraction capability for both global information and local details. Finally, a multi-scale feedforward deep convolution model is established to fully capture multi-scale information, enhancing the model's inductive bias tolerance and thus achieving accurate fault diagnosis for rotating machinery. Experimental analysis shows that this invention can efficiently and accurately achieve the goal of rotating machinery fault diagnosis.

[0049] To better implement the rotating machinery fault diagnosis method in the embodiments of the present invention, based on the rotating machinery fault diagnosis method, correspondingly, as follows: Figure 8 As shown, this embodiment of the invention also provides a rotating machinery fault diagnosis device, the rotating machinery fault diagnosis device 800 comprising: The acquisition module 801 is used to acquire vibration signal data of the target rotating mechanical equipment and perform preprocessing. The conversion module 802 is used to convert the preprocessed vibration signal data into multi-scale data based on the CEEMDAN algorithm; The diagnostic module 803 is used to take the multi-scale data and the preprocessed vibration signal data as input to the fault diagnosis model to obtain the fault diagnosis result output by the fault diagnosis model. The fault diagnosis model consists of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels. The dynamic feedforward Transformer module consists of a global attention module and a dynamic deep convolution module.

[0050] The rotating machinery fault diagnosis device 800 provided in the above embodiments can realize the technical solutions described in the above rotating machinery fault diagnosis method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above rotating machinery fault diagnosis method embodiments, and will not be repeated here.

[0051] like Figure 9 As shown, the present invention also provides a fault diagnosis device 900. The fault diagnosis device 900 includes a processor 901, a memory 902, and a display 903. Figure 9 Only some components of the fault diagnosis device 900 are shown; however, it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented instead.

[0052] In some embodiments, processor 901 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 902 or process data, such as the magnetic resonance image optimization method of the present invention.

[0053] In some embodiments, processor 901 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 901 may be local or remote. In some embodiments, processor 901 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, etc., or any combination thereof.

[0054] In some embodiments, memory 902 may be an internal storage unit of fault diagnosis device 900, such as a hard disk or memory of fault diagnosis device 900. In other embodiments, memory 902 may also be an external storage device of fault diagnosis device 900, such as a pluggable hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on fault diagnosis device 900.

[0055] Furthermore, the memory 902 may include both internal storage units of the fault diagnosis device 900 and external storage devices. The memory 902 is used to store the application software and various types of data installed on the fault diagnosis device 900.

[0056] In some embodiments, display 903 may be an LED display, a liquid crystal display, a touch-screen liquid crystal display, or an organic light-emitting diode (OLED) touchscreen, etc. Display 903 is used to display information from the fault diagnosis device 900 and to display a visual user interface. Components 901-903 of the fault diagnosis device 900 communicate with each other via a system bus.

[0057] In one embodiment, when the processor 901 executes the rotating machinery fault diagnosis program in the memory 902, the following steps can be implemented: Acquire vibration signal data of the target rotating mechanical equipment and perform preprocessing; The preprocessed vibration signal data is converted into multi-scale data based on the CEEMDAN algorithm. The multi-scale data and the preprocessed vibration signal data are used as inputs to the fault diagnosis model to obtain the fault diagnosis results output by the fault diagnosis model. The fault diagnosis model consists of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels. The dynamic feedforward Transformer module consists of a global attention module and a dynamic deep convolution module.

[0058] It should be understood that when the processor 901 executes the rotating machinery fault diagnosis program in the memory 902, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0059] Furthermore, this embodiment of the invention does not specifically limit the type of the fault diagnosis device 900 mentioned. The fault diagnosis device 900 can be a portable electronic device such as a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, or laptop computer. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic devices can also be other portable electronic devices, such as laptop computers with touch-sensitive surfaces (e.g., touch panels). It should also be understood that in some other embodiments of the invention, the fault diagnosis device 900 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0060] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions in the rotating machinery fault diagnosis methods provided in the above-described method embodiments.

[0061] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0062] The rotating machinery fault diagnosis method, apparatus, fault diagnosis equipment, and storage medium provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A rotating machine fault diagnosis method characterized by comprising: The method comprises: acquiring vibration signal data of a target rotating machinery device and preprocessing the vibration signal data; converting the preprocessed vibration signal data into multi-scale data based on a CEEMDAN algorithm; inputting the multi-scale data and the preprocessed vibration signal data into a fault diagnosis model to obtain a fault diagnosis result output by the fault diagnosis model; wherein the fault diagnosis model is composed of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module and a full connection module, and is trained based on sample vibration signal data with fault labels, and the dynamic feedforward Transformer module is composed of a global attention module and a dynamic deep convolution module.

2. The rotating machine fault diagnostic method according to claim 1, characterized by, The fault diagnosis model is composed of a first multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, a second multi-scale feedforward deep convolution module, a first full connection module and a second full connection module connected in sequence.

3. The rotating machine fault diagnostic method according to claim 2, characterized by, The first multi-scale feedforward deep convolution module and the second multi-scale feedforward deep convolution module are composed of a first convolution module, a multi-scale deep convolution module, a feature fusion module and a second convolution module connected in sequence, and are used for fusing multi-scale features. The first convolution module includes a convolution layer, an activation layer and a batch normalization layer, the multi-scale deep convolution module includes deep convolution layers with different convolution kernel sizes, the feature fusion module includes an activation layer and a batch normalization layer, and the second convolution module includes a convolution layer and a batch normalization layer.

4. The rotating machine fault diagnostic method according to claim 2, characterized by, The global attention module is composed of a first deep convolution module, a second deep convolution module, a third deep convolution module, a first linear layer, a second linear layer, a third linear layer and a multi-head attention module, and is used for extracting global features. The first deep convolution module includes a deep convolution layer, an activation layer and a batch normalization layer, the output of the first deep convolution module is linearly transformed by the first linear layer and used as a query vector of the multi-head attention module, the second deep convolution module includes a deep convolution layer and a batch normalization layer, the third deep convolution module includes a deep convolution layer, the second deep convolution module and the third deep convolution module are connected through a residual connection, the output obtained by connecting the second deep convolution module and the third deep convolution module through the residual connection is linearly transformed by the second linear layer and used as a value vector of the multi-head attention module, and is linearly transformed by the third linear layer and used as a key vector of the multi-head attention module.

5. The rotating machine fault diagnostic method according to claim 2, characterized by, The dynamic deep convolution module is composed of an adaptive average pooling module, a third convolution module, a fourth convolution module and a deep convolution layer connected in sequence, and is used for extracting local features. The third convolution module includes a convolution layer, an activation layer and a batch normalization layer, the fourth convolution module includes a convolution layer, and the convolution kernel parameters of the deep convolution layer are determined based on the output of the fourth convolution module.

6. The rotating machine fault diagnostic method according to any one of claims 1 to 5, characterized by, The fault diagnosis model takes cross-entropy loss as a loss function in a training process, and updates model parameters based on an adaptive M-estimation algorithm.

7. The rotating machine fault diagnostic method according to any one of claims 1 to 5, characterized by, The CEEMDAN algorithm is used to convert the preprocessed vibration signal data into multi-scale data, including: The CEEMDAN algorithm is used to convert the preprocessed vibration signal data into a plurality of IMF components and a residual signal.

8. A rotating machine failure diagnosing device characterized by comprising: Including: An acquisition module is configured to acquire vibration signal data of a target rotating machinery device and perform preprocessing; A conversion module is configured to convert the preprocessed vibration signal data into multi-scale data based on the CEEMDAN algorithm; A diagnosis module is configured to take the multi-scale data and the preprocessed vibration signal data as inputs of a fault diagnosis model, and obtain a fault diagnosis result output by the fault diagnosis model. The fault diagnosis model is composed of a multi-scale feedforward deep convolution module, a dynamic feedforward Transformer module, and a fully connected module, and is trained based on sample vibration signal data with fault labels.

9. A failure diagnosing apparatus characterized by comprising: It includes a memory and a processor, wherein The memory is configured to store a program; The processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the rotating machinery fault diagnosis method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer-readable program or instruction is stored, and the program or instruction is executed by a processor to implement the steps of the rotating machinery fault diagnosis method of any one of claims 1 to 7.

Images