Industrial data diagnosis method based on SEResNet and attention mechanism

Abstract

The invention discloses an industrial data diagnosis method based on SEResNet and an attention mechanism, and belongs to the field of mechanical equipment signal processing. The method comprises the following steps: designing a central network and a SEResNet branch network, constructing an SRMMF fault diagnosis model to diagnose faults in an industrial process, extracting features through the central network and the SEResNet branch network, and combining an AMF attention mechanism to realize multi-feature fusion; and finally, outputting a final diagnosis result through the fault identification sub-network. And the accurate and stable diagnosis of the fault is realized.

Description

Technical Field

[0001] The present invention relates to an industrial data diagnosis method based on SEResNet and an attention mechanism, and belongs to the field of mechanical equipment signal processing. Background Art

[0002] With the advent of industrial intelligence, the scale of mechanical equipment has continued to expand and the system has continued to become more complex, resulting in an increase in sensor monitoring points. After long-term data collection, massive amounts of data have been formed. These massive amounts of data have greatly increased the difficulty of industrial data fault diagnosis.

[0003] At present, common multi-sensor fusion fault diagnosis methods include multi-channel input CNN model method, fusion method based on attention mechanism and fusion fault diagnosis method based on image splicing. Among them, the multi-channel input CNN model collects different types of signals through multiple channels, and then uses the CNN model to extract and fuse features, and finally realizes fault classification; however, the CNN model has high requirements on data quality and complex preprocessing process; and with the increase of network layers, there will be problems of gradient disappearance or gradient explosion; the fusion method based on attention mechanism constructs dynamic graphs and uses attention mechanism to enhance the model's learning ability for key features, but this method is more sensitive to noise and the generalization ability of the model is limited; the fusion method based on image splicing realizes fault diagnosis by splicing and fusing image data from different sensors, but the feature extraction and fusion of this method are more difficult, and it is highly dependent on the environment; in addition, the above methods cannot guarantee the stability and generalization of the network model while realizing the extraction of multi-level and deep features.

[0004] In addition, the existing methods only use residual networks as feature extractors, which fails to effectively eliminate the interference of useless information and the diagnostic results are not accurate enough. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides an industrial data diagnosis method based on SEResNet and attention mechanism, the method comprising: Step 1: Collect historical data from industrial processes and preprocess them; Step 2: Divide the preprocessed data in step 1 into training set, validation set and test set; Step 3: By designing the feature extraction and fusion sub-network and the fault identification sub-network, a fault diagnosis model based on the SEResNet multi-sensor attention mechanism fusion (hereinafter referred to as the SRMMF fault diagnosis model) is constructed; Step 4: Use the training set obtained in step 2 to train the model constructed in step 3; Step 5: Input the data to be diagnosed into the model trained in step 4 for fault diagnosis.

[0006] The industrial process data in step 1 include: process data of waste heat recovery fan in cold rolling mill, process data of phosphorus removal pump in hot rolling mill, process data of coal mine air compressor, etc.; Preprocessing includes: fast Fourier transform (FFT) and variational mode decomposition (VMD); In step 2, the ratio of training set, validation set and test set is 7:2:1; The SRMMF fault diagnosis model in step 3 includes a feature extraction and fusion subnetwork and a fault identification subnetwork; the feature extraction and fusion subnetwork can adaptively combine multiple fusion layers through a multi-layer fusion framework and an attention-based fusion strategy, and extract relevant information from multiple sensor features. The fused multi-sensor features are then further fed into the fully connected layer in the fault identification subnetwork to achieve fault identification.

[0007] The feature extraction and fusion sub-network includes 1 central network and 2 SEResNet branch networks; the central network includes AMF module, dimension conversion module and global pooling module; the SEResNet branch network includes convolution layer, pooling layer, SEResNet feature extraction network and global pooling; The feature extraction and fusion subnetwork is built based on a multi-layer fusion framework. First, the features of different signals are extracted through two SEResNet branch networks. The fusion stage is activated after the convolution layer, and the fusion point is set after the pooling layer. The pooling layer is mainly used for the aggregation of effective information. The fusion point is set after the pooling layer to reduce the number of model parameters. The attention fusion algorithm is used to fuse multi-sensor features at each fusion point. Due to the existence of convolution and pooling layers, the feature dimensions between different fusion points are not consistent. In order to fuse features of different levels, the fusion features of the current level need to be dimensionally converted. The dimension conversion module is used to match the feature dimension of the next fusion point. All pooling operations are maximum pooling, and batch normalization is used after each convolution layer.

[0008] The SEResNet branch network automatically extracts deep features of single sensor data, fuses the extracted multi-level multi-sensor data features with the central network, significantly enhances the information interaction between multi-sensor data, and realizes adaptive hierarchical fusion of information.

[0009] Specifically, the first convolutional layer is used to extract shallow features of the signal. The extracted shallow feature map is provided to the SEResNet layer, and 16 SEResNet blocks are used to extract deep features to obtain the most important features and suppress redundant features. Then, in order to improve accuracy and reduce computational costs, shallow features are fused with deep features through global residuals. The features extracted from each SEResNet block are combined and connected through the convolutional layer, and then the effective information is aggregated through the pooling layer for fusion of the central network.

[0010] The fault recognition sub-network uses the global average pooling layer to receive high-level fusion features. It should be noted that although the recognition results are also required in the SEResNet branch network, it is only necessary to ensure the performance of the branch network, and the final recognition result is still determined by the central network.

[0011] The SEResNet feature extraction network in the SEResNet branch network combines the signal reconstruction of SE-Net and the residual feature extraction network of ResNet. SE-Net can automatically learn a set of weights through a small sub-network and calculate the weights for each channel of the feature map. Therefore, in this way, the useful feature channels are enhanced and the redundant feature channels are weakened. In addition, the residual network is easy to optimize and alleviates the problem of gradient disappearance due to increased depth, so the residual network is selected as the feature extractor.

[0012] The specific process is: SE-Net includes: squeezing, excitation and scaling. Given the number of feature channels is Input data , through a series of convolution operations, the number of feature channels is obtained Features The implementation process is as follows:

[0013] Represents input data, Represents a convolution operation.

[0014] Through global average pooling Compress each feature map into a real number with a global sensitivity field on the feature map, and obtain a size of Global attention information The calculation process is as follows:

[0015] in, Indicates the number of feature channels Features, It is the global attention information extracted from the corresponding feature map through the squeezing operation. Represents global average pooling.

[0016] Next, we perform excitation on two fully connected layers in SE-Net to obtain the relationship between feature channels. The training results are used to increase the weight of more important feature information of the task and reduce the weight of unimportant feature information. The calculation process is as follows:

[0017] in and Respectively represent the weight matrices of different fully connected layers in SE-Net, represents the ReLU activation function, represents the sigmoid activation function, Represents the feature weights of each resulting feature map.

[0018] The feature weight of each feature channel is estimated according to the value of the loss function, and its expression is:

[0019] in, represents the total number of categories, represents the true value, Represents the predicted value.

[0020] The weight parameters are updated through the back propagation of the loss function, and the optimal weight is output according to the error between the predicted value and the true value. Then, based on the feature weight, useful information is enhanced and useless information is suppressed, so that the model can achieve better performance.

[0021] Finally, the Scale operation performs a recalibration of the original features in the channel dimension by multiplying the feature weights output by the Excitation operation by the previous feature channel by channel. Therefore, the model can distinguish the characteristics of each channel. The formula is as follows:

[0022] in, is the feature map that rescales the original features.

[0023] SE-Net is used to learn the importance of multi-channel features, enhance the feature learning ability of each channel, and show higher accuracy.

[0024] The residual network contains many residual blocks, and the basic residual learning block is defined as:

[0025]

[0026] in is the feature map that rescales the original features, is the residual function, represents the weight of the first layer of convolution in the residual block, represents the weight of the second layer of convolution in the residual block, is the nonlinear function ReLU, Represents the output after learning the network.

[0027] The central network fuses the features extracted from the two branch networks through the AMF module, and realizes the fusion between the fusion points through dimension conversion, which greatly improves the comprehensiveness of feature extraction and fusion; The specific process of the AMF module is as follows: first, the features of a single signal are extracted through two single-branch feature extraction networks, and then the extracted features are input into the AMF module of the central network to fuse the signal features from two different sensors, and finally the fault classification is performed through the SoftMax classification function.

[0028] Take two monitoring signals as an example. Assume They are the signal features extracted from two different sensor monitoring signals. Represents a three-dimensional real number tensor; the specific fusion process is as follows: (1) Use global average pooling operation to compress the signal features of monitoring signals from different sensors in the spatial dimension:

[0029]

[0030]

[0031] in and Respectively represent the two signal features in The compression characteristics of the channels, The spatial dimension of the feature representation is , The number of channels representing features, i Indicates c Channel No. i Features, express Middle The characteristics of the channels, express Middle The characteristics of a channel.

[0032] (2) Combine the compressed features of the two sensor signals to generate global representation information ; Its expression is:

[0033] in, Respectively represent signal characteristics Features after compression; In addition, in order to enable the excitation signal to completely correct the characteristics of each sensor, the feature learning process should be nonlinear. Therefore, a full connection operation is added after the global information feature to improve the nonlinearity; the expression is:

[0034] in, is a compact feature after dimensionality reduction, Represents dimensions in the range of real numbers; and Represent weight and bias respectively; Represents the nonlinear function ReLU.

[0035] (3) Based on compact features after dimensionality reduction , and then generate an incentive signal with soft attention and ,in The features of each channel can be adaptively selected. In addition, a SoftMax function is added to obtain the excitation probability of each channel feature:

[0036]

[0037] here , respectively corresponding to the two signal features The learnable weight matrix, The dimensions of the matrix are OK, Column, the elements are real numbers. and Represents two different learnable weight matrices and , for the same feature To transform, and Respectively and The excitation signal.

[0038] (4) Each feature from different sensor data is recalibrated and fused by the excitation signal through a gating mechanism:

[0039] in, represents fusion features; ⊗ represents channel direction product.

[0040] After the above four steps, the features of the signals of the two sensors can be fused at each fusion point. Due to the existence of convolutional layers and pooling layers, the feature dimensions between different fusion points are not consistent.

[0041] In order to fuse features at different levels, the fused features at the current level will first pass through the Dimension Transformation Module (DTM). The Dimension Transformation Module includes 1×1 convolutional layers and pooling layers. The 1×1 convolutional layer can scale the channel dimension and increase the nonlinearity of the network, and the pooling layer can scale the spatial dimension of the feature. When the feature dimensions of the two fusion points are consistent, the features of the current layer and the next layer can be obtained by the following formula:

[0042] in Indicates The output features of the fusion points are represents the weight coefficient, Indicated in Multi-sensor features at three fusion points passing through the AMF module.

[0043] The dimension conversion module converts the dimension of the previous fusion point into the same dimension as the next fusion point, thereby realizing the feature fusion of the two fusion points.

[0044] Step 4: During the model training process, the number of model iterations and learning rate are continuously adjusted. When the accuracy tends to be stable and the loss value curve is stable and no longer decreases, the model training is completed; then the model is verified through the validation set. When the validation set also achieves the above results, the trained model can be used for fault diagnosis of the data to be diagnosed.

[0045] Step 5: Input the data to be diagnosed into the model trained in step 4 for fault diagnosis.

[0046] The beneficial effects of the present invention are: The present invention proposes an industrial data diagnosis method based on SEResNet and attention mechanism. By combining SE-Net signal reconstruction and ResNet residual feature extraction network, a SEResNet feature extraction network is constructed, which improves the extraction efficiency of useful information of fault signals and solves the problem of useless information interference in feature extraction. By combining the attention mechanism to fuse signals from two different sensors, the comprehensiveness and stability of fault diagnosis are greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 It is a structural diagram of SE-Net in the industrial data diagnosis method based on SEResNet and attention mechanism proposed in the present invention; Figure 2 It is a structural diagram of the SRMMF fault diagnosis model in the industrial data diagnosis method based on SEResNet and attention mechanism proposed in the present invention; Figure 3 It is a flow chart of the industrial data diagnosis method based on SEResNet and attention mechanism proposed in the present invention; Figure 4 It is a curve chart showing the accuracy and loss value of the industrial data diagnosis method based on SEResNet and attention mechanism proposed in the present invention; Figure 5 It is a curve chart of the accuracy and loss value of the existing MCFCNN method; Figure 6 It is a curve chart of the accuracy and loss value of the RMF method in the prior art; Figure 7 It is a confusion matrix diagram of the industrial data diagnosis method based on SEResNet and attention mechanism proposed in the present invention; Figure 8 It is a schematic diagram of the confusion matrix of the prior art MCFCNN method; Fig. 9 It is a schematic diagram of the confusion matrix of the prior art RMF method. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0050] Embodiment 1 The present invention proposes an industrial data diagnosis method based on SEResNet and attention mechanism, which specifically includes: Step 1: Collect historical data from industrial processes and preprocess them; Step 2: Divide the preprocessed data in step 1 into training set, validation set and test set; Step 3: Construct the SRMMF fault diagnosis model by designing the feature extraction and fusion subnetwork and the fault identification subnetwork; Step 4: Use the training set obtained in step 2 to train the model constructed in step 3; Step 5: Input the data to be diagnosed into the model trained in step 4 for fault diagnosis.

[0051] The industrial process data in step 1 include: process data of waste heat recovery fan in cold rolling mill, process data of phosphorus removal pump in hot rolling mill, process data of coal mine air compressor, etc.; Preprocessing includes: fast Fourier transform (FFT) and variational mode decomposition (VMD); In step 2, the ratio of training set, validation set and test set is 7:2:1; The SRMMF fault diagnosis model in step 3 includes a feature extraction and fusion subnetwork and a fault identification subnetwork; the feature extraction and fusion subnetwork can adaptively combine multiple fusion layers through a multi-layer fusion framework and an attention-based fusion strategy, and extract relevant information from multiple sensor features. The fused multi-sensor features are then further fed into the fully connected layer in the fault identification subnetwork to achieve fault identification.

[0052] The feature extraction and fusion sub-network includes 1 central network and 2 SEResNet branch networks; the central network includes AMF module, dimension conversion module and global pooling; the SEResNet branch network includes convolution layer, pooling layer, SEResNet feature extraction network and global pooling; The feature extraction and fusion subnetwork is built based on a multi-layer fusion framework. First, the features of different signals are extracted through two SEResNet branch networks. The fusion stage is activated after the convolution layer, and the fusion point is set after the pooling layer. The pooling layer is mainly used for the aggregation of effective information. The fusion point is set after the pooling layer to reduce the number of model parameters. The attention fusion algorithm is used to fuse multi-sensor features at each fusion point. Due to the existence of convolution and pooling layers, the feature dimensions between different fusion points are not consistent. In order to fuse features of different levels, the fusion features of the current level need to be dimensionally converted. The dimension conversion module is used to match the feature dimension of the next fusion point. All pooling operations are maximum pooling, and batch normalization is used after each convolution layer.

[0053] The SEResNet branch network automatically extracts deep features of single sensor data, fuses the extracted multi-level multi-sensor data features with the central network, significantly enhances the information interaction between multi-sensor data, and realizes adaptive hierarchical fusion of information.

[0054] Specifically, the first convolutional layer is used to extract shallow features of the signal. The extracted shallow feature map is provided to the SEResNet layer, and 16 SEResNet blocks are used to extract deep features to obtain the most important features and suppress redundant features. Then, in order to improve accuracy and reduce computational costs, shallow features are fused with deep features through global residuals. The features extracted from each SEResNet block are combined and connected through the convolutional layer, and then the effective information is aggregated through the pooling layer for fusion of the central network.

[0055] The fault recognition sub-network uses the global average pooling layer to receive high-level fusion features. It should be noted that although the recognition results are also required in the SEResNet branch network, it is only necessary to ensure the performance of the branch network, and the final recognition result is still determined by the central network.

[0056] The SEResNet feature extraction network in the SEResNet branch network combines the signal reconstruction of SE-Net and the residual feature extraction network of ResNet. SE-Net can automatically learn a set of weights through a small sub-network and calculate the weights for each channel of the feature map. Therefore, in this way, the useful feature channels are enhanced and the redundant feature channels are weakened. In addition, the residual network is easy to optimize and alleviates the problem of gradient disappearance due to increased depth, so the residual network is selected as the feature extractor.

[0057] The specific process is: SE-Net includes: squeezing, excitation and scaling. Given the number of feature channels is Input data , through a series of convolution operations, the number of feature channels is obtained Features The implementation process is as follows:

[0058] Represents input data, Represents a convolution operation.

[0059] Through global average pooling Compress each feature map into a real number with a global sensitivity field on the feature map, and obtain a size of Global attention information The calculation process is as follows:

[0060] in, Indicates the number of feature channels Features, It is the global attention information extracted from the corresponding feature map through the squeezing operation. Represents global average pooling.

[0061] Next, we perform excitation on two fully connected layers in SE-Net to obtain the relationship between feature channels. The training results are used to increase the weight of more important feature information of the task and reduce the weight of unimportant feature information. The calculation process is as follows:

[0062] in and Respectively represent the weight matrices of different fully connected layers in SE-Net, represents the ReLU activation function, represents the sigmoid activation function, Represents the feature weights of each resulting feature map.

[0063] The feature weight of each feature channel is estimated according to the value of the loss function, and its expression is:

[0064] in, represents the total number of categories, represents the true value, Represents the predicted value.

[0065] The weight parameters are updated through the back propagation of the loss function, and the optimal weight is output according to the error between the predicted value and the true value. Then, based on the feature weight, useful information is enhanced and useless information is suppressed, so that the model can achieve better performance.

[0066] Finally, the Scale operation performs a recalibration of the original features in the channel dimension by multiplying the feature weights output by the Excitation operation by the previous feature channel by channel. Therefore, the model can distinguish the characteristics of each channel. The formula is as follows:

[0067] in, is the feature map that rescales the original features.

[0068] SE-Net is used to learn the importance of multi-channel features, enhance the feature learning ability of each channel, and show higher accuracy.

[0069] The residual network contains many residual blocks, and the basic residual learning block is defined as:

[0070]

[0071] in is the feature map that rescales the original features, is the residual function, represents the weight of the first layer of convolution in the residual block, represents the weight of the second layer of convolution in the residual block, is the nonlinear function ReLU, Represents the output after learning the network.

[0072] The central network fuses the features extracted from the two branch networks through the AMF module, and realizes the fusion between the fusion points through dimension conversion, which greatly improves the comprehensiveness of feature extraction and fusion; The specific process of the AMF module is as follows: first, the features of a single signal are extracted through two single-branch feature extraction networks, and then the extracted features are input into the AMF module of the central network to fuse the signal features from two different sensors, and finally the fault classification is performed through the SoftMax classification function.

[0073] Take two monitoring signals as an example. Assume They are the signal features extracted from two different sensor monitoring signals. Represents a three-dimensional real tensor; the specific fusion process is as follows: (1) Use global average pooling operation to compress the signal features of monitoring signals from different sensors in the spatial dimension:

[0074]

[0075]

[0076] in and Respectively represent the two signal features in The compression characteristics of the channels, The spatial dimension of the feature representation is , The number of channels representing features, i Indicates c Channel No. i Features, express Middle The characteristics of the channels, express Middle The characteristics of a channel.

[0077] (2) Combine the compressed features of the two sensor signals to generate global representation information ; Its expression is:

[0078] in, Respectively represent signal characteristics Features after compression; In addition, in order to enable the excitation signal to completely correct the characteristics of each sensor, the feature learning process should be nonlinear. Therefore, a full connection operation is added after the global information feature to improve the nonlinearity; the expression is:

[0079] in, is a compact feature after dimensionality reduction, Represents dimensions in the range of real numbers; and Represent weight and bias respectively; Represents the nonlinear function ReLU.

[0080] (3) Based on compact features after dimensionality reduction , and then generate an incentive signal with soft attention and ,in The features of each channel can be adaptively selected. In addition, a SoftMax function is added to obtain the excitation probability of each channel feature:

[0081]

[0082] here , which correspond to two signals The learnable weight matrix, The dimensions of the matrix are OK, Column, the elements are real numbers. and Represents two different learnable weight matrices and , for the same feature To transform, and Respectively and The excitation signal.

[0083] (4) Each feature from different sensor data is recalibrated and fused by the excitation signal through a gating mechanism:

[0084] in, represents fusion features; ⊗ represents channel direction product.

[0085] After the above four steps, the features of the signals of the two sensors can be fused at each fusion point. Due to the existence of convolutional layers and pooling layers, the feature dimensions between different fusion points are not consistent.

[0086] In order to fuse features at different levels, the fused features at the current level will first pass through the Dimension Transformation Module (DTM). The Dimension Transformation Module includes 1×1 convolutional layers and pooling layers. The 1×1 convolutional layer can scale the channel dimension and increase the nonlinearity of the network, and the pooling layer can scale the spatial dimension of the feature. When the feature dimensions of the two fusion points are consistent, the features of the current layer and the next layer can be obtained by the following formula:

[0087] in Indicates The output features of the fusion points are represents the weight coefficient, Indicated in Multi-sensor features at three fusion points passing through the AMF module.

[0088] The dimension conversion module converts the dimension of the previous fusion point into the same dimension as the next fusion point, thereby realizing the feature fusion of the two fusion points.

[0089] Step 4: During the model training process, the number of model iterations and learning rate are continuously adjusted. When the accuracy tends to be stable and the loss value curve is stable and no longer decreases, the model training is completed; then the model is verified through the validation set. When the validation set also achieves the above results, the trained model can be used for fault diagnosis of the data to be diagnosed.

[0090] Step 5: Input the data to be diagnosed into the model trained in step 4 for fault diagnosis.

[0091] Embodiment 2 This embodiment provides an industrial data diagnosis method based on SEResNet and attention mechanism, which is implemented based on the SRMMF fault diagnosis model described in Example 1.

[0092] Step 1: Collect the vibration signal and current signal of the bearing during operation and pre-process them; Among them, preprocessing includes fast Fourier transform (FFT) and variational mode decomposition (VMD). FFT is used to convert the signal from the time domain to the frequency domain to obtain the signal's spectrum information. VMD is used to decompose the signal into a series of modal functions to obtain the signal's time domain feature information at different scales. Through the above preprocessing methods, the signal's time domain and frequency domain feature information are obtained at the same time, thereby improving the globality of feature extraction.

[0093] Step 2: Input the preprocessed data in step 1 into the SRMMF fault diagnosis model for diagnosis and output the final diagnosis results.

[0094] In order to evaluate the performance of the SRMMF fault diagnosis model proposed in the present invention, the model is compared with the existing multi-sensor fusion methods, including the multi-channel input CNN model (MCFCNN) and the multi-sensor fusion method (RMF) using ResNet alone. The multi-channel input CNN model (MCFCNN) can be specifically referred to the introduction in "Li Hongmei. Research on intelligent fault diagnosis method based on convolutional neural network [D]. North University of China, 2021.", and the multi-sensor fusion method (RMF) using ResNet alone can be referred to the introduction in "Hong Liang, Yu Qiyuan, Qin Chaoqun, et al. Bearing fault diagnosis based on information fusion and dual-connection attention residual network [J]. Vibration and Shock, 2023, 42(20): 114-123".

[0095] The SRMMF fault diagnosis model proposed in the present invention is constructed based on the pytorch framework. The network model can output accuracy graphs, loss value graphs, and confusion matrix graphs. The above results are used to analyze the fault diagnosis accuracy and classification effect, and then evaluate the performance of the proposed network model.

[0096] A comparative analysis is carried out from three aspects: accuracy change curve, loss value change curve and confusion matrix.

[0097] The accuracy and loss value change curves of the SRMMF model in the method proposed by the present invention are as follows: Figure 4 As shown, the accuracy and loss value change curves of the MCFCNN model and the RMF model in the prior art are respectively as follows Figure 5 and Figure 6As shown; in the accuracy curve, the SRMMF model proposed in the present invention tends to be stable after a small number of iterations, the fluctuation is reduced, and the accuracy reaches more than 99%; the MCFCNN model accuracy curve fluctuation is greater than that of SRMMF, but the accuracy can still reach more than 95% after stabilizing; and the RMF model accuracy is similar to that of SRMMF after stabilization, but the fluctuation is large, and the model does not have better stability. In the loss value curve, the SRMMF model curve proposed in the present invention converges faster and is smoother, and successfully achieves more than 99% accurate extraction and identification of different types of bearing fault features. The other two model curves converge more slowly and fluctuate more, and the effect is worse than that of SRMMF.

[0098] The confusion matrix of the SRMMF model in the method proposed by the present invention is as follows: Figure 7 As shown, the confusion matrix of the MCFCNN model and the RMF model in the prior art is as follows Figure 8 and Fig. 9 As shown, the confusion matrix is ​​a table used to evaluate the performance of classification models in machine learning, with actual categories and predicted categories as rows and columns, including true positive examples, true negative examples, false positive examples and false negative examples, which can be used to calculate indicators such as accuracy, precision, and recall, and help to comprehensively evaluate model performance. For the SRMMF model confusion matrix diagram proposed by the present invention, it can be observed that the prediction accuracy of the model in all categories has reached 100%, and no misclassification has occurred. The MCFCNN model has some confusion in label 0 and label 3. In label 0, 1 sample is mistakenly predicted as label 2, and in label 3, 4 samples are mistakenly predicted as label 0. The RMF model has confusion in label 0, and 19 samples are mistakenly predicted as label 3.

[0099] Compared with the MCFCNN model and the RMF model in the prior art, the SRMMF model in the method proposed in the present invention has the advantages of high accuracy and model stability.

[0100] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An industrial data diagnosis method based on SEResNet and attention mechanism, characterized in that: The method comprises: Step 1: Collect historical data from industrial processes and preprocess them; Step 2: Divide the preprocessed data in step 1 into training set, validation set and test set; Step 3: By designing the feature extraction and fusion sub-network and the fault identification sub-network, a fusion fault diagnosis model based on the SEResNet multi-sensor attention mechanism is constructed; Step 4: Use the training set obtained in step 2 to train the model constructed in step 3; Step 5: Input the data to be diagnosed into the model trained in step 4 for fault diagnosis; The feature extraction and fusion sub-network in step 3 includes a central network and a SEResNet branch network; The SEResNet branch network includes: a convolutional layer, a pooling layer, a SEResNet feature extraction network and a global pooling; The central network includes: an AMF module, a dimension conversion module and a global pooling; The AMF module extracts signal features from two sensor monitoring signals. Perform global average pooling operation in the spatial dimension to obtain compressed features ; Then compress the features Combined into global representation information F g , and then use the full connection operation and ReLU function to perform nonlinear feature learning to obtain compact features after dimensionality reduction ; Then, using the corresponding two signal features Learnable weight matrix For the compact features after dimensionality reduction Transform and combine with SoftMax function to generate signal features with soft attention and The excitation signal ; Finally, through the gating mechanism, the excitation signal Signal characteristics Recalibrate and fuse to obtain the final fusion feature F ; The fault identification subnetwork includes a fully connected layer and an output, and implements fault identification by receiving the output of the feature extraction and fusion subnetwork; The features of different signals are extracted through the SEResNet feature extraction network in the SEResNet branch network, and the extracted features are fused through the AMF module in the central network. The fusion between different AMF modules is then achieved through the dimension conversion module. The fusion result is output through the fully connected layer to produce the fault identification result.

2. The method according to claim 1, characterized in that The SEResNet feature extraction network is composed of a signal reconstruction of SE-Net and a residual feature extraction network of ResNet; The specific process of SE-Net signal reconstruction is as follows: SE-Net includes: squeezing, excitation and scaling. The number of feature channels is given as Input data , the number of feature channels obtained through convolution operation is Features , the expression is: in, Represents input data, Represents the convolution operation; Through global average pooling, the features Compress each feature map into a real number with a global sensitivity field on the feature map, and obtain a size of Global attention information , the expression is: in, Indicates the number of feature channels Features, is the global attention information extracted from the corresponding feature map through the squeezing operation; represents global average pooling; Excitation is performed on the two fully connected layers of SE-Net to obtain the relationship between feature channels; the expression is: in and Respectively represent the weight matrices of different fully connected layers in SE-Net, represents the ReLU activation function, represents the sigmoid activation function, Represents the feature weight of each resulting feature map; The feature weight of each feature channel is estimated according to the value of the loss function. The expression of the loss function is: in, represents the total number of categories, represents the true value, Represents the predicted value; updates the weight parameters through back propagation of the loss function; Finally, the feature weights output by the Excitation operation are multiplied by the previous feature channel by channel through the Scale operation, and the original features are recalibrated in the channel dimension; its expression is: in, is the feature map that rescales the original features.

3. The method according to claim 2, characterized in that The expression of the ResNet residual feature extraction network is: in, is the feature map that rescales the original features, is the residual function, Represents the weight of the first convolution layer in the ResNet residual feature extraction network, Represents the weight of the second convolution layer in the ResNet residual feature extraction network, is the nonlinear function ReLU, Represents the output of the SEResNet feature extraction network after learning.

4. The method according to claim 3, characterized in that The processing process of the AMF module is: Take two monitoring signals as an example, assuming They are the signal features extracted from two different sensor monitoring signals. Represents a three-dimensional real number tensor; the specific fusion process is as follows: Step S1: Use the global average pooling operation to compress the signal features from different sensor monitoring signals in the spatial dimension: in, and Represents two signal characteristics respectively In the The compression characteristics of the channels, The spatial dimension of the feature representation is , The number of channels representing features, i Indicates The first i Features, express Middle The characteristics of the channels, express Middle The characteristics of each channel; Step S2: Combine the compressed features of the two sensor signals to generate global representation information ; Its expression is: in, Respectively represent signal characteristics Features after compression; A full connection operation is added after the global information feature, and its expression is: in, is a compact feature after dimensionality reduction, Represents dimensions in the range of real numbers; and Represent weight and bias respectively; Represents the nonlinear function ReLU; Step S3: Based on the compact features after dimensionality reduction , generating an incentive signal with soft attention and , by adding the SoftMax function, the excitation probability of each channel feature is obtained: in, , which correspond to two features respectively. The learnable weight matrix, The dimensions of the matrix are OK, Column, elements are real numbers; and Represents two different learnable weight matrices and , for the same feature To transform, and Respectively represent the corresponding and The incentive signal; Step S4: Each feature from different sensor data is recalibrated and fused by the excitation signal through a gating mechanism: in, represents fusion features; ⊗ represents channel direction product.

5. The method according to claim 4, characterized in that The dimension conversion module converts the dimensions output by different AMF modules into the same one and fuses the converted results; its expression is: in, Indicates The output features of the fusion points are represents the weight coefficient, Indicated in The dimension conversion module converts the dimension of the previous fusion point into the same dimension as the next fusion point, thereby realizing the feature fusion of the output results of different AMF modules.

6. The method according to claim 5, characterized in that During the model training process of step 4: the model is optimized by adjusting the number of model iterations and the learning rate. When the accuracy and loss value curves do not change with the training process, the model training is completed.

7. The method according to claim 6, characterized in that The historical data of the industrial process in step 1 includes: process data of waste heat recovery fan in cold rolling mill, process data of phosphorus removal pump in hot rolling mill or process data of coal mine air compressor; The preprocessing includes: fast Fourier transform and variational mode decomposition.

8. The method according to claim 7, characterized in that In step 2, the ratio of the training set, the validation set and the test set is 7:2:1.

Images