A hydroelectric generating unit guide bearing health management method and device and electronic equipment
By combining a deep fusion of an autoencoder network and a linear t-SNE model, the nonlinearity and high-dimensional feature interference problems in guide bearing health status identification are solved, enabling efficient and accurate identification and management of guide bearing health status.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUAIYIN INSTITUTE OF TECHNOLOGY
- Filing Date
- 2026-03-09
- Publication Date
- 2026-06-23
AI Technical Summary
Existing methods for identifying the health status of bearings fail to effectively address the nonlinear and non-stationary characteristics of vibration signals, and redundant information interference in the high-dimensional feature space limits the accuracy of pattern recognition. Furthermore, there is a lack of adaptive model building methods.
A deep fusion autoencoder network (DIAE) is used to extract the health status features of the bearing by combining random noise terms and Jacobian matrix. The features are then mapped to low dimensions using a linear t-SNE model, and finally, a Softmax classifier is used for pattern recognition.
It improves the accuracy and interpretability of guide bearing health status identification, overcomes the interference of nonlinear and non-stationary signals, realizes efficient feature reduction and visualization analysis, and provides scientific and reliable support for guide bearing health management.
Smart Images
Figure CN122262902A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of condition monitoring and fault diagnosis of hydropower units, and in particular to a method, device and electronic equipment for health management of guide bearings of hydropower units. Background Technology
[0002] Ensuring the safe, stable, and reliable operation of hydropower units is of paramount importance to the productivity and efficiency of power enterprises and the security of the power grid. As a key component supporting the rotating parts of the unit and suppressing vibration and sway, the health status of guide bearings directly affects the long-term stable operation of the entire unit. Considering the need for on-site maintenance personnel to monitor the real-time operating status of equipment, rapid and accurate guide bearing health status detection results can provide necessary data support for maintenance personnel to formulate reasonable maintenance plans and implement effective health management. Furthermore, researching effective fault identification model construction methods for guide bearings operating in abnormal or faulty states to achieve accurate and efficient fault type identification has become a pressing technical challenge to be solved in ensuring the safety of hydropower station equipment.
[0003] Existing methods for identifying the health status of bearings often directly input the collected vibration signals into a classifier for status discrimination, neglecting the removal of noise components from the signal and the mining of deep health features. This makes it difficult to effectively address the nonlinear and non-stationary characteristics of bearing vibration signals. Furthermore, the interference of redundant information in the high-dimensional feature space also limits the accuracy of pattern recognition. Therefore, researching scientifically reliable models and methods for deep extraction and efficient reduction of bearing health status features can provide a new approach to solving these problems. Currently, systematic and in-depth research on such models and methods for deep extraction and efficient reduction of health status features has not been conducted, and there is a lack of adaptive model building methods, which cannot meet the functional requirements for accurate detection of bearing health status. Summary of the Invention
[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a method, device and electronic equipment for health management of guide bearings of hydropower units, which aims to solve the problem of insufficient accuracy in identifying the health status of guide bearings of existing hydropower units.
[0005] To achieve the above objectives, in a first aspect, this application provides a method for health management of guide bearings in hydropower units, comprising:
[0006] Step S1: Based on the online monitoring device, collect the vibration signal of the guide bearing position during unit operation, and use it as the original sample input for the health management model;
[0007] Step S2: Construct a deep fusion autoencoder network to extract the set of guide bearing health status features from the original samples;
[0008] Step S3: Using the health status feature set as input, establish a linear t-SNE model to capture the low-dimensional mapping of the feature set;
[0009] Step S4: Use a pattern classifier to complete the pattern recognition of the health status of the guide bearing.
[0010] Further, step S1 includes the following sub-steps:
[0011] (S1-1) Based on the structural composition and layout of the hydropower unit, several acceleration sensors are arranged at its guide bearing.
[0012] (S1-2) Using the deployed accelerometers, collect a set of vibration signal samples during the operation of the guide bearing. ,in , Indicates the sample dimension. Indicates the number of samples.
[0013] Further, step S2 includes the following sub-steps:
[0014] (S2-1) Using vibration signal samples Using the input as input, a deep fusion autoencoder network (DIAE) is constructed, with the following hidden layer number: ;
[0015] (S2-2) Add random noise term to the DIAE layer. In order to eliminate the influence of noise components mixed in the original input;
[0016] (S2-3) Integrate the Jacobian matrix into the 2nd... The loss function of the layer is used to suppress minor disturbances in the feature transfer process between different layers;
[0017] The output of the top hidden layer of the DIAE constructed in (S2-4) is the extracted set of health status features of the guide bearing. The DIAE loss function can be expressed as follows:
[0018]
[0019] in, and These represent the original input and the DIAE output, respectively. The loss function of the autoencoder. For random noise coefficients, It is the identity matrix. This is the underlying output of DIAE. As a regularization factor, This represents the Jacobian matrix.
[0020] Furthermore, step S3 includes the following sub-steps:
[0021] (S3-1) Establish a linear t-SNE model to realize high-dimensional spatial features. Representation in low-dimensional space The parameterized mapping process can be represented as follows:
[0022]
[0023] in, These are the parameters corresponding to the sample data in the low-dimensional space. For random, fixed sample points, This is the Gaussian kernel function.
[0024] (S3-2) The set of health status features of the guide bearing extracted in step S2 As input, a low-dimensional mapping representation of the feature set is captured by a linear t-SNE model constructed based on (S3-1).
[0025] Further, step S4 includes the following sub-steps:
[0026] (S4-1) Select the Softmax model as the equipment fault type classifier;
[0027] (S4-2) Using the low-dimensional mapping representation of the feature set obtained in step S3 as the input of the Softmax model, pattern recognition of the health status of the guide bearing is completed.
[0028] Secondly, this application provides a health management device for the guide bearing of a hydroelectric generator, comprising:
[0029] The signal acquisition module is used to collect vibration signals of the guide bearing position during unit operation using online monitoring devices, which serve as the raw sample input for the health management model.
[0030] The feature extraction module is used to construct a deep fusion autoencoder network to extract a set of bearing health status features from the original samples obtained by the signal acquisition module.
[0031] The dimension reduction mapping module is used to establish a linear t-SNE model with the set of health status features extracted by the feature extraction module as input, and capture the low-dimensional mapping of the feature set.
[0032] The pattern recognition module is used to perform pattern recognition of the health status of the guide bearing by using a pattern classifier based on the low-dimensional mapping representation output by the dimensionality reduction mapping module.
[0033] Thirdly, this application provides an electronic device, comprising: at least one memory for storing a computer program; and one or more processors for executing the program stored in the memory, wherein when the program stored in the memory is executed, the processors are configured to perform the method described in the first aspect or any possible implementation thereof.
[0034] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.
[0035] Fifthly, this application provides a computer program product that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.
[0036] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.
[0037] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art:
[0038] (1) This application is the first to integrate random noise terms and Jacobian matrices into an autoencoder network to construct a DIAE model. By introducing random noise terms at the bottom layer, the robustness of the model to noise components in the original vibration signal is effectively improved, avoiding the impact of noise interference on the accuracy of subsequent feature extraction. At the same time, by integrating the Jacobian matrix into the hidden layer loss function, the small perturbations of features during the layer-by-layer transmission process are significantly suppressed, ensuring the stability and completeness of the extracted health status features. This method overcomes the limitations of traditional autoencoders in adapting to nonlinear and non-stationary vibration signals, realizes the accurate mining and adaptive extraction of deep health features of guide bearings, and fills the theoretical gap in the construction of a robust feature extraction model for hydroelectric guide bearings.
[0039] (2) This application is the first to consider constructing a linear t-SNE model for low-dimensional mapping of the high-dimensional health status feature set extracted by DIAE. This model realizes the parameterized mapping from the high-dimensional feature space to the low-dimensional visualization space. While effectively preserving the local structure and global distribution characteristics of the original features, it significantly reduces the interference of redundant feature information on pattern recognition, providing a more discriminative low-dimensional representation input for the Softmax classifier. This method overcomes the technical difficulties of efficient reduction and visualization analysis of high-dimensional health features, significantly improves the accuracy and interpretability of guide bearing health status pattern recognition, and provides scientific and reliable technical support for the intelligent health management of hydropower unit guide bearings. Attached Figure Description
[0040] Figure 1 A flowchart illustrating the health management method for hydroelectric generator guide bearings provided in this application embodiment;
[0041] Figure 2 A schematic diagram of the original time series curve of the vibration signal of the guide bearing of the hydroelectric generator provided in the embodiments of this application;
[0042] Figure 3 This is a schematic diagram of the structure of the DIAE model provided in the embodiments of this application;
[0043] Figure 4 This application provides the low-dimensional spatial mapping distribution results of the vibration signal samples of the guide bearing of the hydropower unit after linear t-SNE dimensionality reduction;
[0044] Figure 5 This is a schematic diagram of the structure of the hydropower unit guide bearing health management device provided in the embodiments of this application;
[0045] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0046] The following is a detailed description of this application with reference to the accompanying drawings.
[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0048] This application uses the analysis of the health management method of the guide bearing of a certain hydroelectric power unit as a case study. Figure 1 The diagram shown is a flowchart of the hydropower unit guide bearing health management method provided in this application, which specifically includes the following steps:
[0049] Step S1: Based on the online monitoring device, collect the vibration signal of the guide bearing position during the unit operation, and use it as the original sample input for the health management model.
[0050] (S1-1) Based on the structural composition and layout of the hydropower unit, several acceleration sensors are arranged at its guide bearing.
[0051] (S1-2) Using the deployed accelerometers, collect a set of vibration signal samples during the operation of the guide bearing. ,in , Indicates the sample dimension. This indicates the number of samples. Taking the health management method analysis of a certain unit's guide bearing as a case study, the included guide bearing health status types and vibration signal samples are shown in Table 1. A schematic diagram of the original time series curves of the vibration signal samples is shown below. Figure 2 As shown.
[0052] Table 1. Description of Health Status Types and Vibration Signal Samples for Hydropower Unit Guide Bearings
[0053]
[0054] Step S2: Construct a deep fusion autoencoder network (DIAE) to extract a set of guide bearing health status features from the original samples.
[0055] (S2-1) Using vibration signal samples Using the input as input, a deep fusion autoencoder network (DIAE) is constructed, with the following hidden layer number: .like Figure 3 This is a structural implementation diagram of the DIAE model constructed in this embodiment, showing the number of hidden layers. The value is 5.
[0056] (S2-2) Add random noise term to the DIAE layer. This is to eliminate the influence of noise components mixed in the original input.
[0057] (S2-3) Integrate the Jacobian matrix into the 2nd... The loss function of the layer is used to suppress minor disturbances in the feature transfer process between different layers.
[0058] The output of the top hidden layer of the DIAE constructed in (S2-4) is the extracted set of health status features of the guide bearing. The DIAE loss function can be expressed as follows:
[0059]
[0060] in, and These represent the original input and the DIAE output, respectively. The loss function of the autoencoder. For random noise coefficients, It is the identity matrix. This is the underlying output of DIAE. As a regularization factor, This represents the Jacobian matrix. In this embodiment, the random noise coefficients... The value is 0.5, which is the regularization factor. The value is 0.7.
[0061] Step S3: Using the health status feature set as input, establish a linear t-SNE model to capture the low-dimensional mapping of the feature set.
[0062] (S3-1) Establish a linear t-SNE model to realize high-dimensional spatial features. Representation in low-dimensional space The parameterized mapping process can be represented as follows:
[0063]
[0064] in, These are the parameters corresponding to the sample data in the low-dimensional space. For random, fixed sample points, This is the Gaussian kernel function.
[0065] (S3-2) The set of health status features of the guide bearing extracted in step S2 As input, a linear t-SNE model constructed based on (S3-1) is used to capture a low-dimensional mapping representation of the feature set. For example... Figure 4 The figure shows the low-dimensional spatial mapping distribution of the vibration signal samples of the hydropower unit guide bearing after linear t-SNE dimensionality reduction in this embodiment. As can be seen from the figure, the low-dimensional spatial characterization of the signal samples obtained by linear t-SNE dimensionality reduction can realize the accurate identification of different health states of the guide bearing.
[0066] Step S4: Use a pattern classifier to complete the pattern recognition of the health status of the guide bearing.
[0067] (S4-1) Select the Softmax model as the equipment fault type classifier;
[0068] (S4-2) Using the low-dimensional mapping representation of the feature set obtained in step S3 as the input of the Softmax model, pattern recognition of the health status of the guide bearing is completed.
[0069] The following describes the hydropower unit guide bearing health management device provided in this application. The hydropower unit guide bearing health management device described below can be referred to in correspondence with the hydropower unit guide bearing health management method described above.
[0070] Figure 5 This is a schematic diagram of the health management device for the guide bearing of a hydroelectric generator provided in an embodiment of this application, as shown below. Figure 5 As shown, it includes:
[0071] The signal acquisition module is used to collect vibration signals from the guide bearing position during unit operation using online monitoring devices, which serve as the raw sample input for the health management model.
[0072] The feature extraction module is used to construct a deep fusion autoencoder network to extract a set of bearing health status features from the original samples obtained by the signal acquisition module.
[0073] The dimensionality reduction mapping module is used to establish a linear t-SNE model by taking the set of health status features extracted by the feature extraction module as input, and to capture the low-dimensional mapping of the feature set.
[0074] The pattern recognition module is used to perform pattern recognition of the health status of the guide bearing by using a pattern classifier based on the low-dimensional mapping representation output by the dimensionality reduction mapping module.
[0075] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.
[0076] Based on the methods in the above embodiments, this application provides an electronic device. Figure 6 A schematic diagram of the provided electronic device is shown, including: a processor 910, a communication interface 920, a memory 930, and a communication bus 940. The processor 910, communication interface 920, and memory 930 communicate with each other via the communication bus 940. The processor 910 can call logical instructions stored in the memory 930 to execute the methods described in the above embodiments.
[0077] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
[0078] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.
[0079] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.
[0080] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.
[0081] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.
[0082] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).
[0083] It is understood that the various numerical designations used in the embodiments of this application are merely for the convenience of description and are not intended to limit the scope of the embodiments of this application.
[0084] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A method for health management of guide bearings in hydropower units, characterized in that, include: Step S1: Collect vibration signals at the position of the guide bearing during unit operation as the raw sample input for the health management model; Step S2: Construct a deep fusion autoencoder network to extract the set of guide bearing health status features from the original samples; Step S3: Using the health status feature set as input, establish a linear t-SNE model to capture the low-dimensional mapping of the feature set; Step S4: Use a pattern classifier to complete the pattern recognition of the health status of the guide bearing.
2. The method for health management of guide bearings in hydroelectric generator units according to claim 1, characterized in that, Step S1 includes the following sub-steps: (S1-1) Based on the structural composition and layout of the hydropower unit, several acceleration sensors are arranged at its guide bearing. (S1-2) Using the deployed accelerometers, collect a set of vibration signal samples during the operation of the guide bearing. ,in , Indicates the sample dimension. Indicates the number of samples.
3. The method for health management of guide bearings in hydropower units according to claim 1, characterized in that, Step S2 includes the following sub-steps: (S2-1) Using vibration signal samples Using the input as input, a deep fusion autoencoder network DIAE is constructed, with the following number of hidden layers: ; (S2-2) Add random noise term to the DIAE layer. In order to eliminate the influence of noise components mixed in the original input; (S2-3) Integrate the Jacobian matrix into the 2nd... The loss function of the layer is used to suppress minor disturbances in the feature transfer process between different layers; The output of the top hidden layer of the DIAE constructed in (S2-4) is the extracted set of health status features of the guide bearing. .
4. The method for health management of guide bearings in hydropower units according to claim 3, characterized in that, The DIAE loss function is expressed as follows: ； in, and These represent the original input and the DIAE output, respectively. The loss function of the autoencoder. For random noise coefficients, It is the identity matrix. This is the underlying output of DIAE. As a regularization factor, This represents the Jacobian matrix.
5. The method for health management of guide bearings in hydropower units according to claim 1, characterized in that, Step S3 includes the following sub-steps: (S3-1) Establish a linear t-SNE model to realize high-dimensional spatial features. Representation in low-dimensional space The parameterized mapping is represented as follows: ； in, These are the parameters corresponding to the sample data in the low-dimensional space. For random, fixed sample points, The Gaussian kernel function; (S3-2) The set of health status features of the guide bearing extracted in step S2 Using the constructed linear t-SNE model as input, a low-dimensional mapping representation of the feature set is captured.
6. The method for health management of guide bearings in hydropower units according to claim 1, characterized in that, Step S4 includes the following sub-steps: (S4-1) Select the Softmax model as the equipment fault type classifier; (S4-2) Using the low-dimensional mapping representation of the feature set obtained in step S3 as the input of the Softmax model, pattern recognition of the health status of the guide bearing is completed.
7. A health management device for guide bearings of a hydroelectric generator, characterized in that, include: The signal acquisition module is used to collect vibration signals of the guide bearing position during unit operation using online monitoring devices, which serve as the raw sample input for the health management model. The feature extraction module is used to construct a deep fusion autoencoder network to extract a set of bearing health status features from the original samples obtained by the signal acquisition module. The dimension reduction mapping module is used to establish a linear t-SNE model with the set of health status features extracted by the feature extraction module as input, and capture the low-dimensional mapping of the feature set. The pattern recognition module is used to perform pattern recognition of the health status of the guide bearing by using a pattern classifier based on the low-dimensional mapping representation output by the dimensionality reduction mapping module.
8. An electronic device, characterized in that, include: At least one memory for storing computer programs; One or more processors are configured to execute a program stored in the memory, wherein when the program stored in the memory is executed, the processors are configured to perform the hydroelectric generator guide bearing health management method as described in any one of claims 1-6.
9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on an electronic device, the electronic device performs the hydroelectric generator guide bearing health management method as described in any one of claims 1-6.
10. A computer program product, characterized in that: When the computer program product is run on an electronic device, the electronic device performs the hydroelectric generator guide bearing health management method as described in any one of claims 1-6.