Principal component and deep learning algorithm-based main pump fault diagnosis method and system
By integrating principal component analysis and deep learning algorithms, multi-source data analysis and fault mode diagnosis of main pump equipment were realized. This solved the problem of insufficient reliability and accuracy of main pump condition monitoring and diagnosis in existing technologies, improved the intelligence level and reliability of nuclear power plant equipment operation, and reduced downtime losses due to failures.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NUCLEAR POWER INSTITUTE OF CHINA
- Filing Date
- 2023-12-28
- Publication Date
- 2026-06-12
AI Technical Summary
Existing methods for monitoring the status of main pumps mainly rely on manual methods and threshold alarms from set online sensors. They fail to achieve effective analysis and integration of multi-source monitoring sensor information, and cannot detect and warn of early abnormal conditions. The operational reliability of key equipment such as main pumps is poor, and fault diagnosis mainly relies on experience-based judgment, which is highly subjective, has low fault-finding efficiency, and is not real-time or reliable.
An integrated approach based on principal component analysis and deep learning algorithms is adopted to acquire multi-source real-time operating data of the main pump equipment for status monitoring and fault diagnosis. By combining time-frequency domain analysis and expert knowledge base, online status monitoring and fault diagnosis of the main pump equipment can be achieved, thereby improving the reliability and accuracy of monitoring and diagnosis.
It improves the reliability and accuracy of online status monitoring and fault diagnosis of main pumps, reduces the economic losses of power plants caused by abnormal shutdowns of main pump equipment, enhances the intelligence level and reliability of main pump equipment operation in nuclear power plants, and supports the construction of digital nuclear power plants with unmanned monitoring and minimal staffing.
Smart Images
Figure CN117662491B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of online condition monitoring technology for reactor coolant pumps in nuclear power plants, specifically to a method and system for diagnosing main pump faults based on principal component analysis and deep learning algorithms. Background Technology
[0002] The reactor coolant pump (referred to as the "main pump") is a crucial safety device for the reactor and primary coolant loop system in nuclear power plants. It ensures a continuous flow of heat generated by nuclear fission and decay to the secondary coolant loop system and forms part of the reactor coolant system's pressure boundary. Its operational status directly impacts the performance and safety of the nuclear reactor system. However, with increasing service life and the coupling effects between components, the main pump's mechanical components face potential performance degradation and safety issues, such as excessive vibration, bearing damage, mechanical wear, and shaft seal leakage. According to a statistical analysis of pressurized water reactor nuclear power plant accidents in the United States by Combustion Engineering, the average annual downtime for maintenance caused by main pump component failures is approximately 200.6 hours, contributing 1.67% of the total nuclear power plant downtime and triggering an average of 0.19 emergency shutdowns per year. Therefore, it is urgent to take measures to understand, monitor, and assess the health status of the main pump equipment to improve the overall availability of nuclear power equipment.
[0003] However, the operation and maintenance of main pump equipment in existing nuclear power plants mainly relies on manual methods. The status monitoring of key equipment in nuclear power plants mainly depends on the threshold alarms of set online sensors. Effective analysis and integration of multi-source monitoring and sensor information have not been achieved, making it impossible to detect and warn of early abnormal conditions. The operational reliability of key equipment such as main pumps is poor, and the diagnosis after a failure mainly relies on the judgment of experts based on experience, which is highly subjective, has low efficiency in troubleshooting, and the real-time performance, reliability and accuracy of online status monitoring and fault diagnosis of main pumps are poor.
[0004] In view of the above, this application is hereby submitted. Summary of the Invention
[0005] The technical problem to be solved by this invention is that existing main pump status monitoring methods mainly rely on manual methods and threshold alarms of set online sensors. They do not achieve effective analysis and integration of multi-source monitoring sensor information, cannot detect and warn of early abnormal states, have poor operational reliability of key equipment such as main pumps, and rely mainly on expert judgment based on experience after a fault occurs. This is highly subjective, has low fault diagnosis efficiency, and results in poor real-time performance, low reliability, and low accuracy of online status monitoring and fault diagnosis of main pumps.
[0006] The purpose of this invention is to provide a main pump fault diagnosis method and system based on principal component analysis and deep learning algorithms. This method can further improve the reliability and accuracy of online status monitoring and fault diagnosis of main pumps, enhance the intelligence level and reliability of main pump equipment operation in nuclear power plants, and reduce the economic losses of power plants caused by abnormal shutdowns due to main pump equipment failures. This has significant economic implications.
[0007] This invention is achieved through the following technical solution:
[0008] In a first aspect, the present invention provides a main pump fault diagnosis method based on principal component analysis and deep learning algorithms, the method comprising:
[0009] Acquire real-time operating data of the main pump equipment and store the real-time operating data of the main pump equipment;
[0010] Based on the real-time operating data of the main pump equipment, a cyclic monitoring method for the main pump operating status based on principal component analysis and multi-source data prediction is adopted to monitor the main pump status at the current moment and within a subsequent preset time period to obtain the monitoring status; the monitoring status includes normal status and abnormal status.
[0011] Based on the monitoring status and real-time operating data of the main pump equipment, the time-frequency domain analysis method and expert knowledge base are used, combined with the process parameters collected by the non-safety-grade DCS system, to diagnose and identify typical fault modes.
[0012] Furthermore, the real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the main pump equipment's own data interface;
[0013] The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system. The first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes signals such as temperature, flow rate, speed, current and pressure of the main pump equipment.
[0014] Furthermore, the main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction includes a main pump online status monitoring method based on principal component analysis, the steps of which are as follows:
[0015] Step A1: Obtain historical operating data under normal operating conditions of the main pump equipment to form an n×m dimensional data matrix, where n is the number of samples and m is the number of measurement points; calculate the mean and variance of the historical operating data and perform standardization processing to form standardized samples;
[0016] Step A2 involves calculating the covariance matrix of the standardized samples, performing eigenvector decomposition to obtain eigenvalues and eigenvectors, and using the cumulative percentage variance method to obtain the number of principal components t; calculating the projection of the eigenvectors onto the feature space to complete T.2 Calculation of control limits for the two statistical indicators, SPE;
[0017] Step A3, establish an offline principal component model and use actual equipment operation data for process monitoring: standardize the real-time operation data vector at time i to obtain vector x. i and vector x i Substituting into the principal component model, we obtain the estimated vector.
[0018] Step A4, calculate vector x i Statistic T 2 The system obtains real-time statistics from the SPE (Simultaneous Principal Component Model) and compares these statistics with the control limits determined by the offline principal component model to determine the monitoring status of the main pump equipment. If the real-time statistics exceed the control limits, it indicates that the monitored main pump equipment is malfunctioning; otherwise, it indicates that the monitored main pump equipment is operating normally.
[0019] Furthermore, the main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction also includes multi-source parameter prediction based on autoregressive moving average, the steps of which are as follows:
[0020] Step B1: Preprocess the historical operation data to obtain preprocessed historical operation data;
[0021] Step B2: Perform a stationarity test on the preprocessed historical running data, and calculate the autocorrelation function and partial autocorrelation function of the preprocessed historical running data;
[0022] Step B3: Combine the stationarity test to select and determine the relevant parameters in the autoregressive moving average model;
[0023] Step B4 involves inputting real-time running data into a trained autoregressive moving average-based model for prediction, thereby achieving real-time data prediction results.
[0024] Furthermore, the main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction also includes multi-source parameter prediction based on a recurrent neural network (GRU), with the following steps:
[0025] Step C1: Divide the dataset formed by historical running data into training set and test set, and use MinMaxScaler estimator to normalize and preprocess the dataset to obtain processed training data and test data.
[0026] Step C2: Input the processed training data into the recurrent neural network for training. After the loss function reaches the preset standard, save the existing weights and biases in the recurrent neural network.
[0027] Step C3: Input the processed test data into the recurrent neural network for feedforward calculation to obtain the predicted output; and use the MinMaxScaler estimator to inverse normalize the predicted output to obtain the prediction result.
[0028] Furthermore, based on the monitoring status and real-time operating data of the main pump equipment, time-frequency domain analysis and an expert knowledge base are used, combined with process parameters collected by the non-safety-grade DCS system, to diagnose and identify typical fault modes, including:
[0029] Step D1: Based on the monitoring status, analyze the fault mechanisms of rotor bearing system imbalance fault, rubbing fault and base loosening fault, and construct simulation dynamic models under different fault modes.
[0030] Step D2: Solve the simulation dynamic model under different fault modes to calculate the excitation response under typical faults such as vibration and obtain the simulation results;
[0031] Step D3 involves analyzing the time-domain waveform, shaft trajectory, and vibration characteristic values of the simulation calculation results of vibration, etc., to establish a fault expert knowledge base for the main pump equipment under typical fault modes.
[0032] Step D4 involves preprocessing the real-time operating signal to obtain its time-domain waveform, shaft center trajectory, and spectrum, and analyzing its characteristic frequencies to obtain the relevant spectrum of the real-time operating signal. The relevant spectrum of the real-time operating signal and the data characteristics of the second data collected by the non-safety-level DCS system at the same time are then compared and analyzed with the fault expert knowledge base to achieve the diagnosis and identification of typical fault modes of the main pump.
[0033] Among them, the process parameters collected by the non-safety-grade DCS system include high-pressure leakage flow, shaft seal outlet pressure, bearing temperature, motor current and speed, etc.
[0034] Secondly, the present invention provides a main pump fault diagnosis system based on principal component analysis and deep learning algorithms, which uses the aforementioned main pump fault diagnosis method based on principal component analysis and deep learning algorithms; the system includes:
[0035] The acquisition unit is used to acquire and store the real-time operating data of the main pump equipment.
[0036] The status monitoring unit is used to monitor the status of the main pump equipment at the current moment and within a subsequent preset time length based on the real-time operating data of the main pump equipment and adopts a cyclical monitoring method for the main pump operating status based on principal component analysis and multi-source data prediction. The monitored status includes normal status and abnormal status.
[0037] The fault diagnosis unit is used to diagnose and identify typical fault modes based on the monitoring status and real-time operating data of the main pump equipment, using time-frequency domain analysis and expert knowledge base, and in conjunction with process parameters collected by the non-safety-grade DCS system.
[0038] Furthermore, the real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the main pump equipment's own data interface;
[0039] The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system. The first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes signals such as temperature, flow rate, speed, current and pressure of the main pump equipment.
[0040] Thirdly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described main pump fault diagnosis method based on principal component analysis and deep learning algorithms.
[0041] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described main pump fault diagnosis method based on principal component analysis and deep learning algorithms.
[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0043] This invention relates to a main pump fault diagnosis method and system based on principal component analysis and deep learning algorithms. It proposes an integrated data-driven algorithm framework to realize online status monitoring of main pump equipment. This invention improves the reliability and accuracy of online status monitoring and fault diagnosis of main pumps, enhances the intelligence level and reliability of main pump equipment operation in nuclear power plants, and has significant economic implications in reducing the economic losses of power plants caused by abnormal shutdowns due to main pump equipment failures.
[0044] (1) In view of the problems of multi-source sensing monitoring of main pump equipment and the temporal development of monitoring data, this invention proposes a main pump operation status cyclic monitoring method based on principal component analysis and multi-source data prediction, which can realize the main pump status monitoring at the current moment and within a specific time length thereafter.
[0045] (2) Combining expert knowledge of the typical failure mechanism of the main pump with time-frequency domain analysis method, it is possible to distinguish and identify the typical failure mode after the main pump is in abnormal operation, thereby providing more accurate technical support for the subsequent operation and maintenance decision of the equipment.
[0046] (3) The proposed integrated framework model based on the data-driven method accurately matches the monitoring and diagnostic needs of the main pump equipment at different times. Compared with a single monitoring and diagnostic model, this invention can achieve early warning of abnormal conditions of the main pump. The method of this invention can further improve the reliability and accuracy of online status monitoring and fault diagnosis of the main pump, and has important economic significance for improving the reliability of the operation of the main pump equipment in nuclear power plants and reducing the economic losses of power plants caused by abnormal failures of the main pump equipment. Attached Figure Description
[0047] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0048] Figure 1 This is a flowchart of the main pump fault diagnosis method based on principal component analysis and deep learning algorithms of the present invention.
[0049] Figure 2 This is a detailed flowchart of the main pump fault diagnosis method based on principal component analysis and deep learning algorithms of the present invention;
[0050] Figure 3 This is a flowchart of the online status monitoring method for the main pump based on principal component analysis according to the present invention.
[0051] Figure 4 This is a flowchart of the multi-source parameter prediction process based on Autoregressive Moving Average (ARIMA) of this invention.
[0052] Figure 5 This is a flowchart of the multi-source parameter prediction process based on the recurrent neural network GRU of this invention;
[0053] Figure 6 This is a flowchart illustrating the fault diagnosis process based on time-frequency domain analysis and an expert knowledge base, as presented in this invention.
[0054] Figure 7 This is a block diagram of the main pump fault diagnosis system based on principal component analysis and deep learning algorithms according to the present invention. Detailed Implementation
[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0056] Existing main pump status monitoring methods rely primarily on manual methods and threshold alarms from set online sensors. They fail to achieve effective analysis and integration of multi-source monitoring sensor information, making it impossible to detect and warn of early abnormal conditions. This results in poor operational reliability of key equipment such as the main pump. Furthermore, fault diagnosis relies mainly on expert judgment based on experience, leading to strong subjectivity, low fault diagnosis efficiency, and poor real-time performance, reliability, and accuracy of online status monitoring and fault diagnosis for the main pump.
[0057] This invention designs a main pump fault diagnosis method and system based on principal component analysis and deep learning algorithms. Specifically, it proposes an integrated data-driven algorithm framework to realize online status monitoring of main pump equipment. This can further improve the reliability and accuracy of online status monitoring and fault diagnosis of main pumps, enhance the intelligence level and reliability of main pump equipment operation in nuclear power plants, reduce the possibility of economic losses to power plants caused by abnormal shutdowns of key equipment such as main pumps, and lay the foundation for the subsequent implementation of digital nuclear power plants and smart nuclear power plants with "unmanned monitoring and minimal staffing".
[0058] Specifically, this invention focuses on the main pump equipment of nuclear power plants, designing an integrated, data-driven process framework for online status monitoring of the main pump equipment. The framework integrates dimensionality reduction methods such as Principal Component Analysis (PCA), data prediction methods such as Autoregressive Integrated Moving Average (ARIMA) and Gated Recurrent Unit (GRU), expert knowledge, and time-frequency domain analysis methods. Dimensionality reduction methods such as PCA enable effective analysis of multi-source monitoring data of the main pump, achieving real-time monitoring of the main pump's operating status. Simultaneously, based on prediction models such as ARIMA, parameter constraints of process signals are extracted, abnormal operating behaviors are captured promptly, and the false alarm rate is reduced, laying the foundation for subsequent main pump fault diagnosis based on expert knowledge and time-frequency domain analysis methods. It is worth emphasizing that the research work in this invention has both basic scientific research value and practical application prospects. From a scientific perspective, this invention relates to the interdisciplinary field of expert knowledge on typical main pump fault mechanisms and artificial intelligence technology, focusing on the abnormal monitoring and fault mode differentiation of typical main pump fault modes. From a technical application perspective, it has the following characteristics and advantages:
[0059] (1) In view of the problems of multi-source sensing monitoring of main pump equipment and the temporal development of monitoring data, this invention proposes a main pump operation status cyclic monitoring method based on principal component analysis and multi-source data prediction, which can realize the main pump status monitoring at the current moment and within a specific time length thereafter.
[0060] (2) Combining expert knowledge of the typical failure mechanism of the main pump with time-frequency domain analysis method, it is possible to distinguish and identify the typical failure mode after the main pump is in abnormal operation, thereby providing more accurate technical support for the subsequent operation and maintenance decision of the equipment.
[0061] (3) The proposed integrated framework model based on the data-driven method accurately matches the monitoring and diagnostic needs of the main pump equipment at different times. Compared with a single monitoring and diagnostic model, this invention can achieve early warning of abnormal conditions of the main pump. The method of this invention can further improve the reliability and accuracy of online status monitoring and fault diagnosis of the main pump, and has important economic significance for improving the reliability of the operation of the main pump equipment in nuclear power plants and reducing the economic losses of power plants caused by abnormal failures of the main pump equipment.
[0062] Example 1
[0063] like Figure 1 As shown, the present invention provides a main pump fault diagnosis method based on principal component analysis and deep learning algorithms, the method comprising:
[0064] Step 1: Obtain the real-time operating data of the main pump equipment and store the real-time operating data of the main pump equipment;
[0065] In this embodiment, the real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the data interface of the main pump equipment itself.
[0066] The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system. The first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes signals such as temperature, flow rate, speed, current and pressure of the main pump equipment.
[0067] In this embodiment, the real-time operating data of the main pump equipment is saved to the database by the system's data storage software for use by other modules.
[0068] Step 2: Based on the real-time operating data of the main pump equipment, the main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction is adopted to monitor the main pump status at the current moment and within the subsequent preset time length to obtain the monitoring status; the monitoring status includes normal status and abnormal status.
[0069] Step 2, the main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction, includes online status monitoring and multi-source data prediction.
[0070] (1) Online status monitoring method for main pump based on principal component analysis (PCA): Real-time operating data of main pump equipment is used to achieve abnormal detection and fault alarm of equipment by combining PCA, filtering and feature extraction.
[0071] (2) Multi-source parameter prediction: Auto Regressive Integrated Moving Average (ARIMA) and artificial neural networks are used to predict the trend of equipment monitoring parameters within the normal threshold, providing important engineering reference for the transformation of equipment from planned maintenance to preventive maintenance.
[0072] (1) Online status monitoring method for main pumps based on principal component analysis. For example... Figure 3 As shown, Figure 3 This is a flowchart of the online status monitoring method for main pumps based on principal component analysis; the steps of the online status monitoring method for main pumps based on principal component analysis are as follows:
[0073] Step A1 involves acquiring historical operating data under normal operating conditions of the main pump equipment to form an n×m dimensional data matrix, where n is the number of samples and m is the number of measuring points. Since the process parameters measured by the sensors used at the measuring points include temperature, pressure, flow rate, etc., the acquired historical operating data set is multi-source heterogeneous data. To avoid the adverse effects caused by different data magnitudes and dimensions, the mean and variance of the historical operating data are calculated and standardized to form standardized samples.
[0074] Step A2 involves calculating the covariance matrix of the standardized samples, performing eigenvector decomposition to obtain eigenvalues and eigenvectors, and using the cumulative percentage variance method to obtain the number of principal components t; calculating the projection of the eigenvectors onto the feature space to complete T. 2 Calculation of control limits for the two statistical indicators, SPE;
[0075] Step A3, establish an offline principal component model and use actual equipment operation data for process monitoring: standardize the real-time operation data vector at time i to obtain vector x. i and vector x i Substituting into the principal component model, we obtain the estimated vector.
[0076] Step A4, calculate vector x i Statistic T 2 The system obtains real-time statistics from the SPE (Simultaneous Principal Component Model) and compares these statistics with the control limits determined by the offline principal component model to determine the monitoring status of the main pump equipment. If the real-time statistics exceed the control limits, it indicates that the monitored main pump equipment is malfunctioning; otherwise, it indicates that the monitored main pump equipment is operating normally.
[0077] (2) Multi-source parameter prediction
[0078] (2-1) Multi-source parameter prediction based on autoregressive moving average. For example... Figure 4 As shown, Figure 4 The flowchart for multi-source parameter prediction based on Autoregressive Moving Average (ARIMA) is as follows:
[0079] Step B1: Preprocess the historical operation data to obtain preprocessed historical operation data;
[0080] Step B2: Perform a stationarity test on the preprocessed historical running data, and calculate the autocorrelation function and partial autocorrelation function of the preprocessed historical running data;
[0081] Step B3: Combine the stationarity test selection to determine the relevant parameters (p,i,q) in the autoregressive moving average model;
[0082] Step B4 involves inputting real-time running data into a trained autoregressive moving average-based model for prediction, thereby achieving real-time data prediction results.
[0083] (2-2) Multi-source parameter prediction based on recurrent neural network GRU, such as Figure 5 As shown, Figure 5 The flowchart for multi-source parameter prediction based on the recurrent neural network GRU is as follows:
[0084] Step C1: Divide the dataset formed by historical running data into training set and test set, and use MinMaxScaler estimator to normalize and preprocess the dataset to obtain processed training data and test data.
[0085] Step C2: Input the processed training data into the recurrent neural network for training. After the loss function reaches the preset standard, save the existing weights and biases in the recurrent neural network.
[0086] Step C3: Input the processed test data into the recurrent neural network for feedforward calculation to obtain the predicted output; and use the MinMaxScaler estimator to inverse normalize the predicted output to obtain the prediction result.
[0087] Step 3: Based on the monitoring status and real-time operating data of the main pump equipment, use time-frequency domain analysis and expert knowledge base, combined with process parameters collected by the non-safety-grade DCS system, to diagnose and identify typical fault modes.
[0088] like Figure 6 As shown, Figure 6 The flowchart for fault diagnosis based on time-frequency domain analysis and expert knowledge base is as follows: Step 3 specifically includes:
[0089] Step D1: Based on the monitoring status, analyze the fault mechanisms of rotor bearing system imbalance fault, rubbing fault and base loosening fault, and construct simulation dynamic models under different fault modes.
[0090] Step D2: Solve the simulation dynamic model under different fault modes to calculate the excitation response under typical faults such as vibration and obtain the simulation results;
[0091] Step D3 involves analyzing the time-domain waveform, shaft trajectory, and vibration characteristic values of the simulation calculation results of vibration, etc., to establish a fault expert knowledge base for the main pump equipment under typical fault modes.
[0092] Step D4 involves preprocessing the real-time operating signal to obtain its time-domain waveform, shaft center trajectory, and spectrum, and analyzing its characteristic frequencies to obtain the relevant spectrum of the real-time operating signal. The relevant spectrum of the real-time operating signal and the data characteristics of the second data collected by the non-safety-level DCS system at the same time are then compared and analyzed with the fault expert knowledge base to achieve the diagnosis and identification of typical fault modes of the main pump.
[0093] Among them, the process parameters collected by the non-safety-grade DCS system include high-pressure leakage flow, shaft seal outlet pressure, bearing temperature, motor current and speed, etc.
[0094] In step 3 of the above technical solution, the typical fault modes of the main pump equipment are identified by utilizing expert knowledge of the main pump equipment, combined with time-frequency domain analysis methods, shaft trajectory models and other algorithms, so as to determine the type, location and cause of the equipment fault.
[0095] This invention employs an integrated framework for online monitoring of the main pump's status, including a main pump online status monitoring method based on Principal Component Analysis (PCA), multi-source parameter prediction based on autoregressive moving average, multi-source parameter prediction based on recurrent neural network (GRU), and a time-frequency domain analysis method. PCA technology is used to effectively monitor vibration signals, and the ARIMA model and GRU algorithm are used to predict the trends of various main pump signals within normal thresholds, reducing the false alarm rate. Finally, the diagnosis and analysis of typical main pump failure modes are achieved based on an equipment expert knowledge system and time-frequency domain analysis technology.
[0096] This invention can improve the accuracy of fault diagnosis and trend prediction results, thereby further improving the reliability of nuclear power plant main pump equipment operation, reducing the economic losses of power plants caused by abnormal failures of main pump equipment, and improving the economic efficiency of power plants. It has important economic significance and broad application prospects.
[0097] Example 2
[0098] like Figure 7As shown, the difference between this embodiment and Embodiment 1 is that this embodiment further provides a main pump fault diagnosis system based on principal component analysis and deep learning algorithms. This system uses the main pump fault diagnosis method based on principal component analysis and deep learning algorithms from Embodiment 1. The system includes:
[0099] The acquisition unit is used to acquire and store the real-time operating data of the main pump equipment.
[0100] The status monitoring unit is used to monitor the status of the main pump equipment at the current moment and within a subsequent preset time length based on the real-time operating data of the main pump equipment and adopts a cyclical monitoring method for the main pump operating status based on principal component analysis and multi-source data prediction. The monitored status includes normal status and abnormal status.
[0101] The fault diagnosis unit is used to diagnose and identify typical fault modes based on the monitoring status and real-time operating data of the main pump equipment, using time-frequency domain analysis and expert knowledge base, and in conjunction with process parameters collected by the non-safety-grade DCS system.
[0102] As a further implementation, the real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the data interface of the main pump equipment itself;
[0103] The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system. The first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes signals such as temperature, flow rate, speed, current and pressure of the main pump equipment.
[0104] The execution process of each unit can be carried out according to the main pump fault diagnosis method based on principal component and deep learning algorithm in Example 1, and will not be described in detail in this example.
[0105] Meanwhile, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the main pump fault diagnosis method based on principal component analysis and deep learning algorithm of Embodiment 1.
[0106] Meanwhile, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the main pump fault diagnosis method based on principal component analysis and deep learning algorithm of Embodiment 1.
[0107] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0108] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0111] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A main pump fault diagnosis method based on principal component analysis and deep learning algorithms, characterized in that, The method includes: Acquire real-time operating data of the main pump equipment and store the real-time operating data of the main pump equipment; Based on the real-time operating data of the main pump equipment, a cyclic monitoring method for the main pump operating status based on principal component analysis and multi-source data prediction is adopted to monitor the main pump status at the current moment and within a subsequent preset time period to obtain the monitoring status; the monitoring status includes normal status and abnormal status. Based on the monitoring status and the real-time operating data of the main pump equipment, the time-frequency domain analysis method and expert knowledge base are used, combined with the process parameters collected by the non-safety-level DCS system, to diagnose and identify typical fault modes. The main pump operation status cyclic monitoring method based on principal component analysis and multi-source data prediction includes a main pump online status monitoring method based on principal component analysis, the steps of which are as follows: Step A1: Obtain historical operating data under normal operating conditions of the main pump equipment to form an n×m dimensional data matrix, where n is the number of samples and m is the number of measurement points; calculate the mean and variance of the historical operating data and perform standardization processing to form standardized samples; Step A2 involves calculating the covariance matrix of the standardized samples, performing eigenvector decomposition to obtain eigenvalues and eigenvectors, and using the cumulative percentage variance method to obtain the number of principal components t; calculating the projection of the eigenvectors onto the feature space to complete T. 2 Calculation of control limits for the two statistical indicators, SPE; Step A3, establish an offline principal component model and use actual equipment operation data for process monitoring: standardize the real-time operation data vector at time i to obtain a vector. and the vector Substituting into the principal component model, we obtain the estimated vector. ; Step A4, calculate the vector Statistic T 2 The real-time statistics are obtained by combining the SPE with the control limits determined by the offline principal component model to determine the monitoring status of the main pump equipment. If the real-time statistics exceed the control limits, it indicates that the monitored main pump equipment is abnormal; otherwise, it indicates that the monitored main pump equipment is operating normally. The main pump operation status cyclic monitoring method based on principal component analysis and multi-source data prediction also includes multi-source parameter prediction based on autoregressive moving average, with the following steps: Step B1: Preprocess the historical operation data to obtain preprocessed historical operation data; Step B2: Perform a stationarity test on the preprocessed historical operating data, and calculate the autocorrelation function and partial autocorrelation function of the preprocessed historical operating data; Step B3: Combine the stationarity test to select and determine the relevant parameters in the autoregressive moving average model; Step B4: Input the real-time running data into the trained autoregressive moving average model for prediction to achieve real-time data prediction results.
2. The main pump fault diagnosis method based on principal component analysis and deep learning algorithm according to claim 1, characterized in that, The real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the data interface of the main pump equipment itself. The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system; the first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes the temperature, flow rate, speed, current and pressure signals of the main pump equipment.
3. The main pump fault diagnosis method based on principal component analysis and deep learning algorithm according to claim 1, characterized in that, The main pump operation status cyclic monitoring method based on principal component analysis and multi-source data prediction also includes multi-source parameter prediction based on recurrent neural networks, with the following steps: Step C1: Divide the dataset formed by historical running data into training set and test set, normalize and preprocess the dataset to obtain processed training data and test data. Step C2: Input the processed training data into the recurrent neural network for training. Once the loss function reaches the preset standard, save the existing weights and biases in the recurrent neural network. Step C3: Input the processed test data into the recurrent neural network for feedforward calculation to obtain the prediction output; and perform inverse normalization on the prediction output to obtain the prediction result.
4. The main pump fault diagnosis method based on principal component analysis and deep learning algorithm according to claim 1, characterized in that, Based on the monitoring status and real-time operating data of the main pump equipment, time-frequency domain analysis and an expert knowledge base are used, combined with process parameters collected by the non-safety-grade DCS system, to diagnose and identify typical fault modes, including: Step D1: Based on the monitoring status, analyze the fault mechanisms of rotor bearing system imbalance fault, rubbing fault and base loosening fault, and construct simulation dynamic models under different fault modes. Step D2: Solve the simulation dynamics model under different fault modes, calculate the excitation response under typical faults, and obtain the simulation results; Step D3: By analyzing the time-domain waveform, shaft center trajectory, and vibration characteristic values of the simulation results, a fault expert knowledge base for the main pump equipment under typical fault modes is established. Step D4: By preprocessing the real-time operating signal, the time-domain waveform, shaft center trajectory, and spectrum of the real-time operating signal are obtained, and its characteristic frequencies are analyzed to obtain the relevant spectrum of the real-time operating signal; the relevant spectrum of the real-time operating signal and the data characteristics of the second data collected by the non-safety level DCS system at the same time are compared and analyzed with the fault expert knowledge base to realize the diagnosis and identification of typical fault modes of the main pump. Among them, the process parameters collected by the non-safety-grade DCS system include high-pressure leakage flow, shaft seal outlet pressure, bearing temperature, motor current and speed.
5. A main pump fault diagnosis system based on principal component analysis and deep learning algorithms, characterized in that, The system uses the main pump fault diagnosis method based on principal component analysis and deep learning algorithms as described in any one of claims 1 to 4; the system includes: The acquisition unit is used to acquire real-time operating data of the main pump equipment and store the real-time operating data of the main pump equipment. The status monitoring unit is used to monitor the status of the main pump equipment at the current moment and within a subsequent preset time period based on the real-time operating data of the main pump equipment and adopt a main pump operating status cyclic monitoring method based on principal component analysis and multi-source data prediction, so as to obtain the monitoring status; the monitoring status includes normal status and abnormal status. The fault diagnosis unit is used to diagnose and identify typical fault modes based on the monitoring status and the real-time operating data of the main pump equipment, using time-frequency domain analysis and an expert knowledge base, and in conjunction with process parameters collected by the non-safety-level DCS system.
6. The main pump fault diagnosis system based on principal component analysis and deep learning algorithms according to claim 5, characterized in that, The real-time operating data of the main pump equipment refers to the real-time operating data of the sensors collected using the data interface of the main pump equipment itself. The real-time operating data of the main pump equipment includes first data acquired by a dedicated high-frequency acquisition device and second data acquired by a non-safety-grade DCS system; the first data includes the original digital signals of vibration and shaft displacement of the main pump equipment, and the second data includes the temperature, flow rate, speed, current and pressure signals of the main pump equipment.
7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the main pump fault diagnosis method based on principal component analysis and deep learning algorithms as described in any one of claims 1 to 4.
8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the main pump fault diagnosis method based on principal component analysis and deep learning algorithms as described in any one of claims 1 to 4.