Turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network

Abstract

The invention discloses a turbine blade fault diagnosis method based on a clustering coarsened graph convolutional neural network, and relates to the field of turbine fault diagnosis. According to the method, a graph structure is established between original fault samples through the KNN algorithm and converted into graph samples, then multi-level clustering coarsening processing is carried out on graph sample data, convolution operation is carried out on coarsening graphs of each level, node-level features are output, and finally fault diagnosis is carried out according to the node features. The problems that the turbine blade fault diagnosis precision is low and the model operation time is long are mainly solved, the turbine blade fault classification speed and accuracy are effectively improved, and the operation safety of a thermal power generating unit is improved.

Description

TECHNICAL FIELD

[0001] The embodiment of the disclosure belongs to the field of steam turbine blade fault diagnosis, and particularly relates to a steam turbine blade fault diagnosis method based on a clustering coarse graph convolutional neural network. BACKGROUND

[0002] As the core equipment of the power system, the operation state of the steam turbine unit directly affects the safety and economy of the power plant. As the most prone to failure components in the steam turbine, the blades are subjected to the impact of high-temperature and high-pressure steam for a long time, as well as complex loads such as centrifugal force and flutter, and are prone to cracks, water erosion, fatigue and other damages. Therefore, how to accurately diagnose the faults of the steam turbine blades is an important issue to improve the safety of the steam turbine unit. SUMMARY

[0003] The embodiment of the disclosure aims to at least solve one of the technical problems existing in the prior art, and provides a steam turbine blade fault diagnosis method, system, electronic device and computer readable storage medium based on a clustering coarse graph convolutional neural network.

[0004] In one aspect of the embodiment of the disclosure, a steam turbine blade fault diagnosis method based on a clustering coarse graph convolutional neural network is provided, comprising: Collecting vibration signal data of the steam turbine blades, performing denoising and standardization preprocessing, and generating an original data set; According to the original data set, a graph structure of the original fault sample is constructed by a KNN algorithm and is converted into graph sample data; The graph sample data is subjected to multi-level clustering coarse processing to obtain a multi-level coarse graph of the graph sample data; The multi-level coarse graph is input into a pre-trained graph convolutional network model, neighborhood node information is aggregated by multi-layer convolution, and node-level feature representation is output; According to the node features, fault classification is performed to obtain a fault diagnosis result.

[0005] Further, the multi-level clustering coarse processing of the graph sample data to obtain the multi-level coarse graph of the graph sample data comprises: A K-means or hierarchical clustering algorithm is used to cluster and compress the data points according to the vibration signal spectral features to obtain the multi-level coarse graph of the graph sample data.

[0006] Further, the graph convolutional network model comprises 2-4 layers of graph convolutional layers.

[0007] Further, the aggregation of neighborhood node information by multi-layer convolution to output node-level feature representation comprises: Batch normalization and ReLU activation functions are added after each convolution operation; By gradually aggregating the information of neighboring nodes through multiple convolutional layers, the feature vector representation of each node is output.

[0008] Furthermore, the step of classifying faults based on the node characteristics to obtain fault diagnosis results includes: The concatenation result of the node feature vector and the original signal statistical features is input into a fully connected neural network or support vector machine, and the fault diagnosis result is output.

[0009] Furthermore, the acquisition of vibration signal data from the turbine blades includes: Periodically collect vibration signals; The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network also includes: When the fault diagnosis results are the same fault type 5 times consecutively, an early warning signal is triggered.

[0010] Optionally, the fault diagnosis results may include one of cracks, wear, and loosening.

[0011] Another aspect of the embodiments of this disclosure provides a turbine blade fault diagnosis system based on a clustering coarsened graph convolutional neural network, comprising: The acquisition module is used to acquire vibration signal data of turbine blades, perform noise reduction and standardization preprocessing, and generate raw datasets. The graph sample module is used to construct a graph structure of the original fault samples based on the original dataset using the KNN algorithm and convert it into graph sample data. The coarsening module is used to perform multi-level clustering coarsening on the graph sample data to obtain a multi-level coarsened graph of the graph sample data. The feature module is used to input the multi-level coarsened graph into a pre-trained graph convolutional network model, aggregate neighborhood node information through multi-layer convolution, and output node-level feature representations. The diagnostic module is used to classify faults based on the node characteristics and obtain fault diagnosis results.

[0012] In another aspect, embodiments of this disclosure provide an electronic device comprising: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network described above.

[0013] In another aspect of the embodiments of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, can implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network described above.

[0014] The present disclosure discloses a method, system, electronic device, and computer-readable storage medium for turbine blade fault diagnosis based on a clustering coarsened graph convolutional neural network, which has the following advantages compared with the prior art: (1) Cluster coarsening is used to cluster and compress data points, which greatly reduces the dimensionality of the original data, reduces the computational load of the GCN model, and significantly reduces the running time of the model; (2) The node feature vector output by GCN is concatenated with the original signal statistical features to make up for the shortcomings of a single data source, improve the judgment ability of the classifier, and improve the accuracy of fault diagnosis. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a turbine blade fault diagnosis method based on a clustering coarsening graph convolutional neural network, according to an embodiment of this disclosure. Figure 2 This is a schematic diagram of a turbine blade fault diagnosis system based on a clustering coarsening graph convolutional neural network, according to another embodiment of the present disclosure. Figure 3 This is a schematic diagram of the structure of an electronic device according to another embodiment of the present disclosure. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this disclosure, the disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] like Figure 1 As shown, one embodiment of this disclosure provides a method for diagnosing turbine blade faults based on a clustering coarsened graph convolutional neural network, including: Step S1: Collect vibration signal data of turbine blades, perform noise reduction and standardization preprocessing to generate the original dataset.

[0018] Specifically, a non-contact measurement method is used, with sensors installed above the top of the turbine blades to measure the vibration signals of the blades passing through the sensors. The vibration signals are then denoised and standardized to generate the original dataset.

[0019] Step S2: Construct a graph structure of the original fault samples using the KNN algorithm based on the original dataset and convert it into graph sample data.

[0020] Specifically, the KNN algorithm is used to build a graph structure based on the original dataset and then transform it into graph sample data. G ={ G 1, G 2,…, G N}

[0021] Step S3: Perform multi-level clustering coarsening on the graph sample data to obtain a multi-level coarsened graph of the graph sample data.

[0022] Specifically, K-means or hierarchical clustering algorithms are used to cluster and compress data points based on the spectral characteristics of vibration signals to obtain a multi-level coarsened map of the sample data.

[0023] Image Sample Data G i ={ X i , A i , V i , y i},in X i ={ x 1 i , x 2 i ,…, xni} is the characteristic matrix, A i It is an adjacency matrix. V i ={ v 1 i , v 2 i ,…, vni} represents a set of nodes. y i The fault labels are the corresponding to the sample data in the image.

[0024] For each graph sample data G i ={ X i , A i , V i , y i Clustering algorithms can be used to analyze node sets. V i Clustering is performed to group the node sets V Divided into K A mutually exclusive cluster, namely V 1 i , V 2 i ,…, V Ki Each cluster corresponds to an inducible subgraph. GK i = G [ V Ki ].

[0025] Each graph sample data G i It can be done L Level coarsening can be expressed as: G (1) i , G (2) i , … , G ( L ) i ,in G (1) i Image sample G i itself, G ( L ) i It is a graph composed of individual nodes. For 2 ≤ l ≤ L, G(l) i Yes, the coarsened plot. G(l) i The number of nodes is the number of clusters, ultimately yielding the graph sample data. G of l Level coarsening map G (l) ={G( l ) 1,G( l ) 2,…, G( l ) N}.

[0026] Step S4: Input the multi-level coarsened graph into the pre-trained graph convolutional network model, aggregate neighborhood node information through multi-layer convolution, and output node-level feature representation.

[0027] Specifically, the obtained multi-level coarsened graph is input into a pre-trained Graph Convolutional Network (GCN) model, which aggregates neighborhood node information through multiple convolutions and outputs node-level feature representations.

[0028] The GCN model consists of 2 to 4 graph convolutional layers. After each convolutional operation, batch normalization (BatchNorm) and ReLU activation functions are added. The neighborhood node information is gradually aggregated through multiple convolutions, and finally the feature vector representation of each node is output.

[0029] Step S5: Classify the faults based on the node characteristics to obtain the fault diagnosis results.

[0030] Specifically, the concatenation result of the node feature vector output by GCN and the statistical features of the original signal is input into a fully connected neural network or support vector machine (SVM) to output fault diagnosis results, including types such as normal, crack, wear or looseness, among which crack, wear and looseness are fault types.

[0031] The turbine blade fault diagnosis method disclosed herein based on clustering coarsened graph convolutional neural network also includes an online monitoring stage: real-time acquisition of vibration signals and repetition of the above steps S1 to S5, triggering an early warning signal when the diagnosis results are the same fault type for 5 consecutive times.

[0032] Compared with existing technologies, the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network disclosed in this paper has the following advantages: (1) Cluster coarsening is used to cluster and compress data points, which greatly reduces the dimensionality of the original data, reduces the computational load of the GCN model, and significantly reduces the running time of the model; (2) The node feature vector output by GCN is concatenated with the original signal statistical features to make up for the shortcomings of a single data source, improve the judgment ability of the classifier, and improve the accuracy of fault diagnosis.

[0033] like Figure 2 As shown, another embodiment of this disclosure provides a turbine blade fault diagnosis system based on a clustering coarsened graph convolutional neural network, comprising: The acquisition module 210 is used to acquire vibration signal data of turbine blades, perform noise reduction and standardization preprocessing, and generate raw datasets. The sample module 220 is used to construct a graph structure of the original fault samples based on the original dataset using the KNN algorithm and convert it into graph sample data; The coarsening module 230 is used to perform multi-level clustering coarsening on the graph sample data to obtain a multi-level coarsened graph of the graph sample data. Feature module 240 is used to input the multi-level coarsened graph into a pre-trained graph convolutional network model, aggregate neighborhood node information through multi-layer convolution, and output node-level feature representations. The diagnostic module 250 is used to classify faults based on the node characteristics and obtain fault diagnosis results.

[0034] Specifically, the turbine blade fault diagnosis system based on clustering coarsened graph convolutional neural network disclosed herein is used to implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network described above. The specific implementation process will not be repeated here.

[0035] Compared with existing technologies, the turbine blade fault diagnosis system based on clustering coarsened graph convolutional neural network disclosed in this invention has the following advantages: (1) Cluster coarsening is used to cluster and compress data points, which greatly reduces the dimensionality of the original data, reduces the computational load of the GCN model, and significantly reduces the running time of the model; (2) The node feature vector output by GCN is concatenated with the original signal statistical features to make up for the shortcomings of a single data source, improve the judgment ability of the classifier, and improve the accuracy of fault diagnosis.

[0036] like Figure 3 As shown, another embodiment of this disclosure provides an electronic device, including: At least one processor 301; and a memory 302 communicatively connected to the at least one processor 301 for storing one or more programs that, when executed by the at least one processor 301, enable the at least one processor 301 to implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network described above.

[0037] The memory 302 and processor 301 are connected via a bus, which can include any number of interconnecting buses and bridges. The bus connects various circuits of one or more processors 301 and memory 302 together. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 301 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 301.

[0038] Processor 301 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 302 can be used to store data used by processor 301 during operation.

[0039] Another embodiment of this disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network described above.

[0040] The computer-readable medium may be included in the apparatus, device, or system of the present invention, or it may exist independently.

[0041] Computer-readable storage media can be any tangible medium that contains or stores a program, and can be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, optical fibers, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0042] Computer-readable storage media may also include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code, specific examples of which include, but are not limited to, electromagnetic signals, optical signals, or any suitable combination thereof.

[0043] It is understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of this disclosure, and this disclosure is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of this disclosure, and these modifications and improvements are also considered to be within the scope of protection of this disclosure.

Claims

1. A method for fault diagnosis of steam turbine blades based on clustering coarsened graph convolutional neural networks, characterized in that, The method includes: Vibration signal data of turbine blades are collected, and noise reduction and standardization preprocessing are performed to generate the original dataset; Based on the original dataset, a graph structure of the original fault samples is constructed using the KNN algorithm and converted into graph sample data; Multi-level clustering coarsening is performed on the graph sample data to obtain a multi-level coarsened graph of the graph sample data. The multi-level coarsened graph is input into a pre-trained graph convolutional network model, and neighborhood node information is aggregated through multiple convolutions to output node-level feature representations. Based on the node characteristics, fault classification is performed to obtain fault diagnosis results.

2. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to claim 1, characterized in that, The step of performing multi-level clustering coarsening on the graph sample data to obtain a multi-level coarsened graph of the graph sample data includes: Using K-means or hierarchical clustering algorithms, data points are clustered and compressed based on the spectral characteristics of vibration signals to obtain a multi-level coarsened map of the sample data.

3. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to claim 1, characterized in that, The graph convolutional network model includes 2 to 4 graph convolutional layers.

4. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to claim 3, characterized in that, The process of aggregating neighborhood node information through multi-layer convolution and outputting node-level feature representations includes: Add batch normalization and ReLU activation functions after each convolutional operation; By gradually aggregating the information of neighboring nodes through multiple convolutions, the feature vector representation of each node is output.

5. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to claim 1, characterized in that, The step of classifying faults based on the node characteristics to obtain fault diagnosis results includes: The concatenation result of the node feature vector and the original signal statistical features is input into a fully connected neural network or support vector machine, and the fault diagnosis result is output.

6. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to claim 1, characterized in that, The acquisition of vibration signal data from turbine blades includes: Real-time acquisition of vibration signals; The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network also includes: When the fault diagnosis results are the same fault type 5 times consecutively, an early warning signal is triggered.

7. The turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network according to any one of claims 1 to 6, characterized in that, The fault diagnosis results include one of the following: cracks, wear, and loosening.

8. A turbine blade fault diagnosis system based on clustering coarsened graph convolutional neural network, characterized in that, The system includes: The acquisition module is used to acquire vibration signal data of turbine blades, perform noise reduction and standardization preprocessing, and generate raw datasets. The sample module is used to construct a graph structure of the original fault samples based on the original dataset using the KNN algorithm and convert it into graph sample data; The coarsening module is used to perform multi-level clustering coarsening on the graph sample data to obtain a multi-level coarsened graph of the graph sample data. The feature module is used to input the multi-level coarsened graph into a pre-trained graph convolutional network model, aggregate neighborhood node information through multi-layer convolution, and output node-level feature representations. The diagnostic module is used to classify faults based on the node characteristics and obtain fault diagnosis results.

9. An electronic device, characterized in that, include: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it can implement the turbine blade fault diagnosis method based on clustering coarsened graph convolutional neural network as described in any one of claims 1 to 7.

Images