Multi-strategy fusion nuclear reactor system fault diagnosis method and system
Through a multi-strategy fusion nuclear reactor system fault diagnosis method, combined with statistics, knowledge rules and machine learning, effective diagnosis of minor anomalies, moderate faults and severe faults in the reactor system is achieved, thereby improving the safety and reliability of the system.
Patent Information
- Application Number
- CN202510673097.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-23
- Publication Date
- 2025-09-23
AI Technical Summary
Existing nuclear reactor system fault diagnosis methods have low diagnostic efficiency under complex operating conditions, are difficult to adapt to engineering practices, and are difficult to effectively identify early system operation anomalies.
A multi-strategy fusion approach, including statistical analysis, knowledge-based rule methods, and machine learning, is adopted. Combined with the preprocessing of key operating parameters of the reactor system, minor anomalies, moderate faults, and severe faults are diagnosed through stability and consistency analysis, fault tree analysis, and machine learning models.
It has achieved effective diagnosis of various abnormalities in the reactor system, helped operators understand the system status in a timely manner, and improved the safety and reliability of the system.
Smart Images
Figure CN120687955A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nuclear reactor engineering, and in particular to a multi-strategy fusion nuclear reactor system fault diagnosis method and system. Background Art
[0002] A nuclear reactor system is a complex system, and its components and equipment may malfunction during operation. If an anomaly or failure goes undetected, it can lead to the reactor's inability to continue stable operation, or even to a nuclear safety accident, resulting in serious consequences. Fault diagnosis of reactor systems is an effective means of reducing operational risks. By monitoring and diagnosing the operating status of reactor systems and equipment, providing timely warnings of anomalies, and preventing potential failures, it is crucial to further improve the safety and reliability of reactor system operations.
[0003] Since the 1970s, extensive research has been conducted both domestically and internationally on fault diagnosis methods for nuclear reactor systems. Numerous system fault diagnosis methods have been developed, including those based on physical models, knowledge (such as those based on fuzzy reasoning, expert systems, symbolic directed graphs, and case studies), and data (such as machine learning based on artificial neural networks, deep learning based on convolutional neural networks, and radial basis function networks based on extreme learning machines).
[0004] The model-based approach has problems such as insufficient physical model accuracy and limited applicability to complex system diagnosis. The knowledge-based approach has problems such as difficulty in knowledge acquisition and low efficiency in complex fault diagnosis. The data-based approach has problems such as limited diagnostic capabilities due to lack of sample data and poor interpretability of diagnostic results.
[0005] In general, most of the nuclear reactor system diagnostic methods currently under development are limited by the limitations of the methods and are difficult to adapt to the complex working conditions of the actual project. They have not yet been effectively applied in the project. At the same time, they can only achieve relatively effective diagnosis of accident conditions and it is still difficult to effectively identify early operational abnormalities of the system. Summary of the Invention
[0006] The purpose of the present invention is to provide a multi-strategy fusion nuclear reactor system fault diagnosis method and system, which adopts the multi-strategy fusion method to conduct in-depth analysis of the key operating parameters of the reactor system, realize effective diagnosis of various minor abnormalities, moderate faults and serious faults or accidents of the reactor system, help operating personnel intuitively understand the system operation status, timely judge the faults existing in the nuclear reactor system, provide support for the safe and reliable operation of the system, and solve the technical problem of how to effectively identify the early operation abnormalities of the system.
[0007] The present invention is implemented through the following technical solution: a multi-strategy fusion nuclear reactor system fault diagnosis method, comprising the following steps:
[0008] Obtain key operating parameters of the reactor system and build the core operating parameter set of the system;
[0009] Preprocessing the core operating parameter set of the system to obtain structured operating parameters;
[0010] According to the degree of deviation between the structured operating parameters and the reference values, the corresponding diagnostic strategy is used to perform fault diagnosis;
[0011] Among them, if the structured operating parameter is within the preset fluctuation range of the benchmark value, the statistical analysis method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset fluctuation range of the benchmark value and within the preset deviation value, the knowledge rule-based method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset deviation value of the benchmark value, the machine learning method is used to diagnose the fault of the structured operating parameter.
[0012] According to a preferred embodiment, the preprocessing includes performing missing value processing, outlier extraction and noise reduction processing on the numerical parameters in the system core operating parameter set.
[0013] According to a preferred embodiment, the positive fluctuation and negative fluctuation of the preset fluctuation range are respectively set as the calculated envelope values of the key operating parameters of the reactor system of a type of transient state, wherein the positive fluctuation is the positive transient fluctuation boundary of the normal design reference value, and the negative fluctuation is the reverse transient fluctuation boundary of the normal design reference value.
[0014] According to a preferred embodiment, the method of using statistical analysis to perform fault diagnosis on structured operating parameters includes:
[0015] Stability analysis and consistency analysis are performed on the structured operating parameters respectively, and abnormality determination is performed based on the stability analysis results and the consistency analysis results.
[0016] According to a preferred embodiment, the stability analysis comprises the following steps:
[0017] performing normalization processing on the structured operating parameters;
[0018] Extracting time domain features of the normalized structured operating parameters, wherein the time domain features include K-order central moments and variances;
[0019] Based on the key operating parameters of the reactor system under normal conditions, the upper and lower envelopes of each time domain characteristic are constructed as the stability assessment baseline;
[0020] By comparing the real-time time domain characteristics with the stability assessment baseline, it is determined whether the stability is abnormal and the stability analysis results are obtained;
[0021] According to a preferred embodiment, the consistency analysis comprises the following steps:
[0022] Calculating correlation coefficients between key operating parameters of the reactor system under normal operating conditions and normalized structured operating parameters, wherein the correlation coefficients include a Pearson correlation coefficient and a Spearman correlation coefficient;
[0023] Obtain a consistency assessment baseline based on the historical key operating parameters of the reactor system;
[0024] The consistency analysis results are obtained by comparing the correlation coefficient with the consistency assessment baseline to determine whether the consistency is abnormal.
[0025] According to a preferred embodiment, the positive and negative deviation values of the preset deviation value are set as the calculated envelope values of the key operating parameters of the reactor system in the first and second transient states.
[0026] According to a preferred embodiment, after the preliminary determination of a medium fault, the method further comprises converting the structured operating parameters into explanatory data, and performing fault diagnosis on the explanatory data using a knowledge rule-based method.
[0027] According to a preferred embodiment, the method of using a knowledge rule-based method to perform fault diagnosis on structured operating parameters includes:
[0028] Conduct a test-based extended failure mode analysis on the reactor system, establish a positive correlation between failure modes and instrument parameters, and obtain the potential causes of various failure phenomena;
[0029] Conduct fault tree analysis on various fault phenomena to obtain a logical reasoning tree structure of specific fault phenomena and potential fault causes, and construct diagnostic rules based on expert knowledge;
[0030] Fault diagnosis is performed according to the expert knowledge-based diagnosis rules.
[0031] The present invention also provides a multi-strategy fusion nuclear reactor system fault diagnosis system, which is applied to the nuclear reactor system fault diagnosis method described above. The system includes:
[0032] Parameter acquisition module, used to obtain key operating parameters of the reactor system and build the core operating parameter set of the system;
[0033] A preprocessing module, configured to preprocess the system core operating parameter set to obtain structured operating parameters;
[0034] A fault diagnosis module is used to diagnose faults using corresponding diagnostic strategies based on the degree of deviation between the structured operating parameters and the reference values;
[0035] Among them, if the structured operating parameter is within the preset fluctuation range of the benchmark value, the statistical analysis method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset fluctuation range of the benchmark value and within the preset deviation value, the knowledge rule-based method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset deviation value of the benchmark value, the machine learning method is used to diagnose the fault of the structured operating parameter.
[0036] The technical solution of a multi-strategy fusion nuclear reactor system fault diagnosis method and system provided by the present invention has at least the following advantages and beneficial effects: the present invention adopts a multi-strategy fusion method to conduct in-depth analysis of system operation status data, and realizes effective diagnosis of various minor abnormalities, moderate faults and serious faults or accidents in the reactor system, helping operating personnel to intuitively understand the system operation status, timely judge the faults existing in the nuclear reactor system, and provide support for the safe and reliable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 A schematic flow chart of a multi-strategy fusion nuclear reactor system fault diagnosis method provided in Example 1 of the present invention;
[0038] Figure 2 This is a flow chart of the minor abnormality diagnosis algorithm provided in Example 1 of the present invention. DETAILED DESCRIPTION
[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0040] Example 1
[0041] In order to achieve effective identification of early-stage system operation anomalies, the present invention proposes a multi-strategy fusion method for nuclear reactor system fault diagnosis.
[0042] The multi-strategy fusion nuclear reactor system fault diagnosis method of this embodiment is as follows: Figure 1 As shown, the following steps are included:
[0043] Step 1: Build the core operating parameter set of the system;
[0044] In this embodiment, the key operating parameters that have been monitored in the nuclear reactor system and can reflect the operating status of the system are first obtained, and then the system core operating parameter set is constructed based on the obtained key operating parameters; in some preferred implementations, the key operating parameters include reactor power, system temperature, system pressure, flow, ambient temperature and humidity and dosage, operating status data of important equipment, etc.
[0045] Step 2: Preprocessing the system core operating parameter set;
[0046] In this embodiment, the preprocessing includes performing missing value processing, outlier extraction and noise reduction on the numerical parameters in the system core operating parameter set to obtain high-quality structured operating parameters, providing a data basis for subsequent fault diagnosis.
[0047] Step 3: Based on the degree of deviation between the structured operating parameters and the reference values, a corresponding diagnostic strategy is used to perform fault diagnosis;
[0048] In this embodiment, according to the degree of deviation between the structured operating parameters and the reference values, the faults can be preliminarily divided into three types: slight abnormality, moderate fault, severe fault or accident.
[0049] Specifically, if the structured operating parameter is within the preset fluctuation range of the benchmark value, it is preliminarily determined to be a slight abnormality, such as a decrease in pump performance or a decrease in heat exchange capacity of the heat exchanger, and then a statistical analysis method is used to perform fault diagnosis on the structured operating parameter. In this embodiment, the positive fluctuation and negative fluctuation of the preset fluctuation range are respectively set as a type of transient envelope value of the calculation of the key operating parameters of the reactor system. Among them, the positive fluctuation is the positive transient fluctuation boundary of the normal design benchmark value, and the negative fluctuation is the reverse transient fluctuation boundary of the normal design benchmark value.
[0050] If the structured operating parameter is outside the preset fluctuation range of the reference value and within the preset deviation value, it is preliminarily determined to be a moderate fault, such as failure of an auxiliary system pump or valve, container leakage, etc., and then a knowledge rule-based method is used to perform fault diagnosis on the structured operating parameter. In this embodiment, the positive and negative deviation values of the preset deviation value are set as the calculated envelope values of the key operating parameters of the reactor system for Class I transient and Class II transient conditions.
[0051] If the structured operating parameter is outside the preset deviation value of the benchmark value, it is preliminarily determined to be a serious fault or accident, such as loss of coolant flow, rupture of the evaporator heat transfer tube, loss of water supply, etc., and then a machine learning method is used to diagnose the fault of the structured operating parameter.
[0052] The following is a further description of each fault diagnosis method:
[0053] For statistical analysis, see Figure 2 As shown, the fault diagnosis of the structured operating parameters in this embodiment specifically includes: performing stability analysis and consistency analysis on the structured operating parameters respectively, and making an abnormality determination based on the stability analysis results and the consistency analysis results.
[0054] In some preferred embodiments, the stability analysis includes the following steps: (1) normalizing the structured operating parameters to obtain normalized structured operating parameters; (2) extracting time domain features of the normalized structured operating parameters using methods such as wavelet transform and principal component analysis, wherein the time domain features include K-order central moments and variances; (3) constructing upper and lower envelopes of each time domain feature as a stability assessment baseline based on the key operating parameters of the reactor system under normal operating conditions; (4) determining whether the stability is abnormal by comparing the real-time time domain features with the stability assessment baseline to obtain a stability analysis result;
[0055] The consistency analysis includes the following steps: (1) calculating the correlation coefficient between the key operating parameters of the reactor system under normal operating conditions and the structured operating parameters after normalization, wherein the correlation coefficient includes the Pearson correlation coefficient and the Spearman correlation coefficient, so as to intuitively characterize the operating status of the reactor system; (2) obtaining a consistency assessment baseline based on the historical key operating parameters of the reactor system; (3) judging whether the consistency is abnormal by comparing the correlation coefficient with the consistency assessment baseline, and obtaining the consistency analysis result.
[0056] Regarding the knowledge rule-based method, this embodiment specifically includes the following steps for fault diagnosis of structured operating parameters: (1) conducting a test extended failure mode analysis (FMECA) on the reactor system, establishing a positive correlation between the failure mode and the instrument parameters, and obtaining various types of fault phenomena, such as potential fault causes such as deviation of measuring instrument parameters; (2) conducting a fault tree analysis on the various types of fault phenomena, obtaining a logical reasoning tree structure of specific fault phenomena and potential fault causes, and constructing expert knowledge-based diagnostic rules to build a nuclear reactor system fault diagnosis model based on expert knowledge; (3) converting the structured operating parameters into standardized and unified explanatory data such as high power, high reactor outlet temperature, and low coolant flow; (4) diagnosing faults on the explanatory data according to the expert knowledge-based diagnostic rules. In some preferred embodiments, the expression of the expert knowledge-based nuclear reactor system fault diagnosis model is as follows:
[0057]
[0058] In the above formula, RelevanceScore i,j Indicates the correlation score between the i-th parameter and the j-th fault type, reflecting the knowledge of fault type analysis, OccurrenceRatej Indicates the frequency level of the jth fault type, with a value of 1-9, 1 means extremely rare, 9 means extremely frequent, j Indicates the severity level of the jth fault type, with a value of 1-9, 1 means slight impact, 9 is extremely serious, j It represents the difficulty level of detecting the j-th fault type, with a value of 1-9, where 1 means very easy to detect and 9 means very difficult to detect. The denominator 9 is used to normalize the product of the three 1-9 scores. m ) indicates fault x m The confidence level reflects the degree of certainty of the fault and reflects the knowledge of parameter monitoring and analysis. i Indicates the deviation of the i-th parameter, normalized to [0,1], Priority(F m ) indicates fault F m Priority to determine the diagnosis order, reflecting the operation and maintenance decision-making knowledge, Frequency (F m ) indicates fault F m The historical frequency of occurrence is in the range of [0,1], Complexity (F m ) indicates fault F m The diagnostic complexity of the fault tree is measured, where a larger value indicates a more complex fault, such as based on the fault tree depth, number of rules, etc.
[0059] Regarding the method based on machine learning, this embodiment specifically includes the following steps for fault diagnosis of structured operating parameters: (1) performing wavelet denoising and principal component analysis on the structured operating parameters to extract time domain, frequency domain and principal component features to form a low-dimensional feature vector; (2) constructing an integrated learning model that integrates artificial neural networks, support vector machines and random forests, and using the SMOTE algorithm to oversample and balance samples, and using five-fold cross-validation to optimize the parameters of the integrated learning model; (3) inputting the low-dimensional feature vector in real time, and outputting the fault type through the integrated learning model after training. In some preferred embodiments, the expression of the integrated learning module is as follows:
[0060]
[0061] Model k (x)=Update(Model k ,x new ,y new )
[0062] In the above formula, EnsembleScore is the final output score of the ensemble learning model, which is obtained by fusing the results of multiple models. K represents the number of ensemble learning models. For example, when neural network, random forest, and support vector machine are selected, K = 3, w k Indicates the weight of the kth model, reflecting the importance of the model in the integration. k (x) represents the output of the kth model for input x, where x is the input data, such as a parameter of a nuclear reactor. Indicates the historical accuracy of the k1th model, which is used to predict the correct proportion of the beam model. The value range is [0,1]. It represents the robustness of the k1th model, which is calculated by testing the data with noise or abnormality, and its value range is [0,1]. Used to normalize weights, Model k represents the k-th model itself, x new Represents new input data, such as a newly monitored parameter sequence of a nuclear reactor, y new Indicates the label of the new input data, such as the known fault type, Update() indicates incremental learning to update the model, using x new and y new Model k Update to improve the model's adaptability to new data.
[0063] In summary, the present invention adopts a multi-strategy fusion method to conduct in-depth analysis of system operation status data, and realizes effective diagnosis of various minor anomalies, moderate faults and serious faults or accidents of the reactor system, helping operators to intuitively understand the system operation status, timely judge the faults existing in the nuclear reactor system, and provide support for the safe and reliable operation of the system.
[0064] Example 2
[0065] Based on the technical solution provided in Example 1, this embodiment provides a system applied to the nuclear reactor system fault diagnosis method provided in Example 1. The system includes a parameter acquisition module, a preprocessing module, and a fault diagnosis module.
[0066] Specifically, the parameter acquisition module is used to obtain key operating parameters of the reactor system and construct a core operating parameter set of the system; the preprocessing module is used to preprocess the core operating parameter set of the system to obtain structured operating parameters; the fault diagnosis module is used to perform fault diagnosis using corresponding diagnostic strategies based on the degree of deviation between the structured operating parameters and the benchmark values; wherein, if the structured operating parameters are within the preset fluctuation range of the benchmark values, a statistical analysis method is used to perform fault diagnosis on the structured operating parameters; if the structured operating parameters are outside the preset fluctuation range of the benchmark values and within the preset deviation values, a knowledge rule-based method is used to perform fault diagnosis on the structured operating parameters; if the structured operating parameters are outside the preset deviation values of the benchmark values, a machine learning method is used to perform fault diagnosis on the structured operating parameters.
[0067] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.
Claims
1. A multi-strategy fusion method for nuclear reactor system fault diagnosis, characterized in that: The steps include: Obtain key operating parameters of the reactor system and build the core operating parameter set of the system; Preprocessing the core operating parameter set of the system to obtain structured operating parameters; According to the degree of deviation between the structured operating parameters and the reference values, the corresponding diagnostic strategy is used to perform fault diagnosis; Among them, if the structured operating parameter is within the preset fluctuation range of the benchmark value, the statistical analysis method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset fluctuation range of the benchmark value and within the preset deviation value, the knowledge rule-based method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset deviation value of the benchmark value, the machine learning method is used to diagnose the fault of the structured operating parameter.
2. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 1, characterized in that: The preprocessing includes performing missing value processing, outlier extraction and noise reduction on the numerical parameters in the system core operating parameter set.
3. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 1, characterized in that: The positive fluctuation and negative fluctuation of the preset fluctuation range are respectively set as the calculation envelope values of the key operating parameters of the reactor system of a type of transient state, wherein the positive fluctuation is the positive transient fluctuation boundary of the normal design reference value, and the negative fluctuation is the reverse transient fluctuation boundary of the normal design reference value.
4. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 3, characterized in that: The method of using statistical analysis to perform fault diagnosis on structured operating parameters includes: Stability analysis and consistency analysis are performed on the structured operating parameters respectively, and abnormality determination is performed based on the stability analysis results and the consistency analysis results.
5. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 4, characterized in that: The stability analysis includes the following steps: performing normalization processing on the structured operating parameters; Extracting time domain features of the normalized structured operating parameters, wherein the time domain features include K-order central moments and variances; Based on the key operating parameters of the reactor system under normal conditions, the upper and lower envelopes of each time domain characteristic are constructed as the stability assessment baseline; By comparing the real-time time domain characteristics with the stability assessment baseline, it is determined whether the stability is abnormal and the stability analysis results are obtained.
6. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 5, characterized in that: The consistency analysis includes the following steps: Calculating correlation coefficients between key operating parameters of the reactor system under normal operating conditions and normalized structured operating parameters, wherein the correlation coefficients include a Pearson correlation coefficient and a Spearman correlation coefficient; Obtain a consistency assessment baseline based on the historical key operating parameters of the reactor system; The consistency analysis results are obtained by comparing the correlation coefficient with the consistency assessment baseline to determine whether the consistency is abnormal.
7. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 1, characterized in that: The positive and negative deviation values of the preset deviation value are set as the calculated envelope values of the key operating parameters of the reactor system for type I transient and type II transient.
8. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 7, characterized in that: After the preliminary determination is made that the fault is moderate, the method further includes converting the structured operating parameters into explanatory data, and performing fault diagnosis on the explanatory data using a knowledge rule-based method.
9. The multi-strategy fusion nuclear reactor system fault diagnosis method according to claim 8, characterized in that: The method of using knowledge rule-based method to perform fault diagnosis on structured operating parameters includes: Conduct a test-based extended failure mode analysis on the reactor system, establish a positive correlation between failure modes and instrument parameters, and obtain the potential causes of various failure phenomena; Conduct fault tree analysis on various fault phenomena to obtain a logical reasoning tree structure of specific fault phenomena and potential fault causes, and construct diagnostic rules based on expert knowledge; Fault diagnosis is performed according to the expert knowledge-based diagnosis rules.
10. A multi-strategy fusion nuclear reactor system fault diagnosis system, characterized in that: Applied to the nuclear reactor system fault diagnosis method according to any one of claims 1 to 9, the system comprises: Parameter acquisition module, used to obtain key operating parameters of the reactor system and build the core operating parameter set of the system; A preprocessing module, configured to preprocess the system core operating parameter set to obtain structured operating parameters; A fault diagnosis module is used to diagnose faults using corresponding diagnostic strategies based on the degree of deviation between the structured operating parameters and the reference values; Among them, if the structured operating parameter is within the preset fluctuation range of the benchmark value, the statistical analysis method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset fluctuation range of the benchmark value and within the preset deviation value, the knowledge rule-based method is used to diagnose the fault of the structured operating parameter; if the structured operating parameter is outside the preset deviation value of the benchmark value, the machine learning method is used to diagnose the fault of the structured operating parameter.