Nuclear equipment reliability data mining method
Through the integrated data mining method of empirical modal decomposition, clustering algorithm and quantum genetic algorithm optimization, the problem that traditional analysis methods are difficult to comprehensively and accurately evaluate the reliability of nuclear-level equipment and timely discover potential faults is solved, and efficient and accurate real-time reliability evaluation and fault prediction are achieved, reducing operation and maintenance costs.
Patent Information
- Application Number
- CN202411780562.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-05-06
AI Technical Summary
Traditional nuclear-grade equipment reliability analysis methods are difficult to comprehensively and accurately evaluate equipment reliability, and cannot detect potential fault hazards in a timely manner, resulting in high maintenance costs and inefficient efficiency.
The historical working condition data of the nuclear-level equipment is obtained through the operating condition signal acquisition device, integrated empirical modal decomposition and clustering algorithm preprocessing, labeling the data sets, and building a data mining model through machine learning algorithms, and optimizing the model using quantum genetic algorithms to achieve real-time reliability evaluation.
It improves the accuracy and efficiency of data processing, enhances the performance of the data mining model, realizes real-time evaluation of the reliability of nuclear-level equipment, promptly discovers potential failure risks, reduces operation and maintenance costs, and improves equipment safety and stability.
Smart Images
Figure CN119939142A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of nuclear energy engineering and data mining technology, and in particular to a nuclear-grade equipment reliability data mining method. Background Art
[0002] In the field of nuclear power generation and nuclear energy application, the reliability of nuclear-grade equipment is crucial. These devices usually operate under extreme conditions, including high temperature, high pressure and strong radiation environment, so their performance and stability are directly related to the safe operation of nuclear power plants and the effective use of nuclear energy. In order to improve the reliability and safety of nuclear-grade equipment, its operating conditions must be continuously monitored and analyzed.
[0003] Traditional nuclear equipment reliability analysis methods mainly rely on expert experience and regular maintenance inspections. However, this method has many limitations. First, expert experience may be limited by personal cognitive level and experience accumulation, making it difficult to fully and accurately evaluate the reliability of equipment. Second, although regular maintenance inspections can detect some obvious faults and problems, they often fail to detect potential fault hazards in a timely manner, resulting in high maintenance costs and low efficiency. Summary of the invention
[0004] The purpose of the present invention is to provide a nuclear-grade equipment reliability data mining method, which aims to solve the many limitations of traditional nuclear-grade equipment reliability analysis methods, such as difficulty in comprehensively and accurately evaluating the reliability of nuclear-grade equipment and failure to promptly discover potential fault hazards, resulting in high maintenance costs and low efficiency.
[0005] The present invention is achieved through the following technical solutions:
[0006] A method for nuclear equipment reliability data mining comprises the following steps:
[0007] Obtain historical operating data of target nuclear-grade equipment through operating signal acquisition equipment to obtain a historical operating signal data set;
[0008] By integrating empirical mode decomposition and clustering algorithms, the historical operating condition signal data set is preprocessed, and the preprocessed historical operating condition signal data set is labeled to obtain a labeled data set;
[0009] Based on the labeled data set, a data mining model is constructed through a machine learning algorithm, and the data mining model is optimized through a quantum genetic algorithm to obtain a data mining optimization model;
[0010] The real-time operating signal data of the target nuclear-grade equipment is collected through the operating signal acquisition equipment, the real-time operating signal data is input into the data mining optimization model, and the data mining results of the target nuclear-grade equipment are output.
[0011] Optionally, the specific process of acquiring the historical operating condition data of the target nuclear-grade equipment through the operating condition signal acquisition device and obtaining the historical operating condition signal data set is:
[0012] The operating condition data of the target nuclear-grade equipment are collected in real time or at regular intervals by deploying the operating condition signal collection equipment on the target nuclear-grade equipment; the collected operating condition data are sorted in chronological order to form a historical operating condition signal data set containing time series information.
[0013] Optionally, the specific process of preprocessing the historical operating condition signal data set by integrating empirical mode decomposition and clustering algorithm is as follows:
[0014] Perform integrated empirical mode decomposition on the historical operating condition signal data set to obtain the corresponding intrinsic mode data set;
[0015] Perform feature extraction on the intrinsic mode data set to obtain a feature vector data set;
[0016] The feature vector data set is clustered by a clustering algorithm to obtain a clustered data set.
[0017] Optionally, the specific process of labeling the preprocessed historical operating condition signal data set to obtain the labeled data set is:
[0018] The clustering data set is displayed through a human-computer interaction interface; based on professional knowledge or expert experience, the state of the clustering data set is identified and corresponding labels are set to form a labeled data set.
[0019] Optionally, the specific process of constructing the operating condition data mining model through a machine learning algorithm based on the labeled data set is:
[0020] The initial data mining model is constructed based on the support vector machine algorithm; the initial data mining model is trained by labeling the data set to obtain the data mining model.
[0021] Optionally, the specific process of optimizing the data mining model by using the quantum genetic algorithm to obtain the data mining optimization model is:
[0022] Define the input parameters of the quantum genetic algorithm, including the white noise amplitude of the integrated empirical mode decomposition algorithm, the cluster center of the clustering algorithm, and the kernel function parameters of the support vector machine;
[0023] The fitness function of the quantum genetic algorithm is set as the state recognition or classification accuracy of the data mining model;
[0024] The input parameters are optimized and iterated through quantum genetic algorithm to obtain the optimal input parameters, and the input parameters are updated with the optimal input parameters to form a data mining optimization model.
[0025] Optionally, the specific process of optimizing and iterating the input parameters by using the quantum genetic algorithm is:
[0026] The white noise amplitude, cluster center and kernel function parameters are quantum-encoded to form chromosomes, and an initial population consisting of several chromosomes is randomly generated;
[0027] The initial population is measured to form a binary population, and the fitness of the binary population is evaluated to obtain a fitness evaluation result;
[0028] Based on the fitness evaluation results, through quantum gate evolution operation and iterative calculation, the solution of the population converges to the optimal solution, and the optimal white noise amplitude, clustering center and kernel function parameters are obtained.
[0029] Optionally, the specific process of the fitness evaluation is:
[0030] The binary population is converted into a decimal population, the parameter values in the decimal population are brought into the data mining model, and the fitness function value corresponding to each parameter is calculated.
[0031] Optionally, the clustering algorithm is a K-means clustering algorithm.
[0032] Optionally, the data mining results are compared with the actual operating conditions of the target nuclear-grade equipment to verify the data mining results; based on the verification results, the data mining optimization model is adjusted and optimized.
[0033] The technical solution of the present invention has at least the following advantages and beneficial effects:
[0034] Improve the accuracy and efficiency of data processing: By integrating empirical mode decomposition (EEMD) and clustering algorithms to pre-process historical operating condition signal data sets, it is possible to effectively decompose complex signal components and identify key features in the signal, thereby improving the accuracy of data processing; at the same time, the application of clustering algorithms helps to classify data of similar operating conditions, simplifying the subsequent data analysis process and improving processing efficiency.
[0035] Enhance the performance of data mining models: Using machine learning algorithms to build data mining models and optimizing them through quantum genetic algorithms (QGA) can significantly enhance the model's prediction and classification capabilities. By simulating the characteristics of quantum systems, it is possible to quickly search for the optimal solution on a global scale, thus avoiding the problem that traditional optimization methods may fall into local optimality, making data mining models more accurate and efficient.
[0036] Real-time reliability assessment: By collecting the operating signal data of the target nuclear-grade equipment in real time and inputting it into the optimized data mining model, real-time assessment of the equipment reliability can be achieved. This not only improves the timeliness of the assessment, but also enables timely discovery of potential equipment failure risks, providing strong support for equipment maintenance and management.
[0037] Improve the safety and stability of nuclear-grade equipment: Through data mining methods, it is possible to deeply explore the hidden information and patterns in the equipment operating data, provide a scientific basis for the preventive maintenance and fault prediction of the equipment, and help take measures in advance to avoid safety accidents caused by equipment failures, thereby significantly improving the safety and stability of nuclear-grade equipment.
[0038] Reduce operation and maintenance costs: Through accurate data mining and real-time reliability assessment, equipment maintenance plans can be formulated and implemented more scientifically, avoiding unnecessary maintenance and replacement costs; at the same time, timely detection and handling of equipment failures also reduces production losses caused by equipment downtime, thereby reducing overall operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 Schematic diagram of the process of the nuclear equipment reliability data mining method according to Embodiment 1 of the present invention;
[0040] Figure 2 Schematic diagram of the process of nuclear-grade equipment reliability data mining method according to embodiment 2 of the present invention. DETAILED DESCRIPTION
[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 here can be arranged and designed in various different configurations.
[0042] Example 1
[0043] Reference Figure 1 , a nuclear-grade equipment reliability data mining method, comprising the steps of:
[0044] Step 1: Obtain historical operating data of target nuclear-grade equipment through operating signal acquisition equipment to obtain a historical operating signal data set.
[0045] In this embodiment, the historical operating data of the target nuclear-grade equipment is obtained through the operating signal acquisition device, and the specific process of obtaining the historical operating signal data set is as follows:
[0046] The operating condition signal acquisition equipment deployed on the target nuclear-grade equipment can be sensors, data recorders, etc., to collect the operating condition data of the target nuclear-grade equipment in real time or at regular intervals; the collected operating condition data are sorted in chronological order and stored in appropriate storage media to form a historical operating condition signal data set containing time series information. The operating condition data may include but are not limited to signals such as temperature, pressure, flow, and vibration.
[0047] Step 2: Preprocess the historical operating condition signal data set by integrating empirical mode decomposition (EEMD) and clustering algorithm, and mark the preprocessed historical operating condition signal data set to obtain a marked data set.
[0048] In this embodiment, the specific process of preprocessing the historical operating condition signal data set by integrating empirical mode decomposition and clustering algorithm is as follows:
[0049] The historical operating signal data set is subjected to integrated empirical mode decomposition to obtain the corresponding intrinsic mode data set, that is, the complex nonlinear and non-stationary historical operating signal is decomposed into several relatively simple, locally stable modal components. In order to overcome the modal aliasing problem in traditional empirical mode decomposition (EMD), multiple new signals are constructed by adding white noise of different amplitudes to the original signal multiple times; empirical mode decomposition is performed on each signal with white noise added to obtain a series of intrinsic mode components and residuals; a series of intrinsic mode components are averaged to eliminate the influence of white noise and obtain the intrinsic mode data set.
[0050] The feature vector data set is obtained by extracting features from the intrinsic mode data set. Features are key features that can reflect the state of the target nuclear-grade equipment, including time domain features, frequency domain features, and time-frequency domain features; among them, time domain features include mean, variance, peak, peak-to-peak value, root mean square value, etc.; frequency domain features are obtained by Fourier transform or fast Fourier transform to obtain frequency spectrum, and then extract spectrum features, such as main frequency, secondary frequency, spectrum energy, etc.; time-frequency features are extracted by wavelet transform, short-time Fourier transform and other methods.
[0051] The feature vector data set is clustered by a clustering algorithm to obtain a clustered data set, that is, the feature vector data set is divided into multiple clusters, each cluster reflects a device state of the target nuclear-level device. Select a clustering algorithm that meets the requirements, such as K-means, hierarchical clustering, DBSCAN, etc.
[0052] In this embodiment, the specific process of marking the preprocessed historical operating condition signal data set to obtain the marked data set is as follows:
[0053] The clustering data set is displayed through a human-computer interaction interface; based on professional knowledge or expert experience, the state of the clustering data set is identified and corresponding labels are set to form a labeled data set.
[0054] In this embodiment, a human-computer interaction interface such as data visualization software, a dashboard or a customized GUI interface can graphically represent time series data, such as a line graph, a scatter plot, etc., so that experts can intuitively observe data features. According to the professional knowledge such as the working principle, failure mode, and operation experience of the target nuclear-grade equipment, the clustered data is state identified. For example, experts can identify that some clusters represent that the equipment is in normal operation, and some clusters can represent a certain abnormal or faulty state of the equipment. Define a clear and unambiguous label for each identified state, for example, "normal operation", "slight vibration abnormality", "severe wear", etc. These labels will be used for subsequent data mining model training and verification. During the labeling process, some data may need to be re-clustered or the labels adjusted. This usually occurs when experts have doubts about the clustering results or discover new data features. The labeled data set can be continuously optimized in an iterative manner until satisfactory accuracy and consistency are achieved.
[0055] Step 3: Based on the labeled data set, a data mining model is constructed through a machine learning algorithm, and the data mining model is optimized through a quantum genetic algorithm to obtain a data mining optimization model.
[0056] In this embodiment, according to the data characteristics and task requirements of the target nuclear-level equipment, a support vector machine (SVM) is selected as the algorithm for building a data mining model. The initial data mining model is constructed based on the support vector machine algorithm; the labeled data set is used as training data, and the relevant parameters of the initial data mining model are configured, such as the kernel function type, penalty parameters, and kernel parameters; the initial data mining model is trained by the labeled data set to obtain a data mining model. The parameters of the data mining model are encoded as a quantum bit string, each quantum bit represents a possible value of a parameter; an initial population containing multiple quantum individuals is randomly generated, each quantum individual represents a possible combination of data mining model parameters; the fitness of each quantum individual in the population is evaluated using a labeled data set, that is, the classification accuracy or loss function value of the SVM model corresponding to each quantum individual on the training data is calculated; according to the fitness value, the quantum individual with higher fitness is selected as the parent generation for subsequent crossover and mutation operations; the parent quantum individual is crossovered to generate child quantum individuals, and the child quantum individuals are mutated to increase the diversity of the population; the parent quantum individuals with lower fitness are replaced with the child quantum individuals to form a new population; after multiple iterations, until a predetermined number of iterations is reached or the fitness value is no longer significantly improved; the optimal parameter combination of the data mining model is decoded from the optimal quantum individual; the data mining model is retrained using the decoded optimal parameter combination to obtain a data mining optimization model.
[0057] Step 4. Collect the real-time operating signal data of the target nuclear-grade equipment through the operating signal acquisition equipment, input the real-time operating signal data into the data mining optimization model, and output the data mining results of the target nuclear-grade equipment. The real-time operating signal data of the target nuclear-grade equipment includes various sensor signals such as temperature, pressure, flow, vibration, etc. The collected real-time operating signal data is preprocessed as necessary, such as denoising, filtering, normalization, etc., to ensure the quality and consistency of the data. The preprocessed real-time operating signal data is input into the data mining optimization model obtained after optimization by the quantum genetic algorithm. The data mining optimization model calculates and analyzes the input real-time data and outputs the data mining results of the target nuclear-grade equipment. The data mining results include the operating status of the equipment, fault warning, remaining life prediction, etc.
[0058] In this embodiment, the data mining results are compared with the actual operating conditions of the target nuclear-grade equipment to verify the accuracy and reliability of the data mining results, which can be achieved by comparing the difference between the predicted value of the model and the actual observed value; according to the verification results, the data mining optimization model is adjusted and optimized. If there is a large deviation between the predicted results of the model and the actual working conditions, the parameters or structure of the model need to be readjusted to improve the prediction accuracy and generalization ability of the model. The real-time operating signal data of the target nuclear-grade equipment is continuously monitored, and the data mining model is regularly updated to adapt to changes in equipment status and new operating conditions.
[0059] Example 2
[0060] Based on Example 1, refer to Figure 2 In this embodiment, the vibration analog signal of the nuclear-grade equipment is collected in real time by a vibration sensor, the collected vibration analog signal is amplified by a charge amplifier, the amplified vibration analog signal is converted into a corresponding digital signal by a vibration signal collection device, and the digital signal corresponding to the vibration analog signal is used as the input of the data mining optimization model.
[0061] In this embodiment, the data mining model is optimized by the quantum genetic algorithm, and the specific process of obtaining the data mining optimization model is as follows:
[0062] Define the input parameters of the quantum genetic algorithm, which include the white noise amplitude of the integrated empirical mode decomposition algorithm (EEMD), the cluster centers of the K-means clustering algorithm, and the kernel function parameters of the support vector machine; the cluster centers of the K-means clustering algorithm include the number of cluster centers and the initial position parameters; the kernel function parameters of the support vector machine include the C parameter (penalty parameter) and the gamma parameter (kernel parameter).
[0063] The fitness function of the quantum genetic algorithm is set as the state recognition or classification accuracy of the data mining model, and the classification accuracy is the classification accuracy;
[0064] The input parameters are optimized and iterated through quantum genetic algorithm to obtain the optimal input parameters, and the input parameters are updated with the optimal input parameters to form a data mining optimization model.
[0065] In this embodiment, the specific process of optimizing and iterating the input parameters by using the quantum genetic algorithm is as follows:
[0066] The white noise amplitude, cluster center and kernel function parameters are quantum encoded to form chromosomes. Each parameter can be mapped to one or more quantum bits, and the possible values of the parameter are represented by quantum states (such as the superposition state of |0> and |1>); and an initial population consisting of several chromosomes is randomly generated; the population size can be determined according to the complexity of the problem and computing resources. The state recognition or classification accuracy of the data mining model is used as the fitness function, which can be obtained by inputting the labeled data set into the data mining model and calculating the classification accuracy of the model output.
[0067] The initial population is measured to form a binary population, and the fitness of the binary population is evaluated to obtain the fitness evaluation result, that is, the parameter values represented in binary are brought into the data mining model, the binary population is converted into a decimal population, and the parameter values in the decimal population are brought into the data mining model for training and testing, and the fitness function value corresponding to each parameter combination is calculated.
[0068] Based on the fitness evaluation results, chromosomes with higher fitness are selected as parents, and quantum gate evolution operations such as quantum rotating gates and quantum cross gates are used to generate offspring chromosomes, aiming to guide the population to evolve in a direction with higher fitness; mutation operations are performed on offspring chromosomes to increase the diversity of the population, and mutation operations can randomly change the state of certain quantum bits; offspring chromosomes are used to replace parent chromosomes with lower fitness to form a new population, and after iterative calculations, the population solution converges to the optimal solution, that is, until the predetermined number of iterations is reached or the fitness value is no longer significantly improved, and the optimal white noise amplitude, cluster center and kernel function parameters are obtained. After reaching the stopping condition, the optimal input parameter combination (i.e., the white noise amplitude of EEMD, the cluster center of K-means and the kernel function parameter of SVM) is decoded from the optimal chromosome, and the data mining model is retrained using the decoded optimal parameter combination to obtain a data mining optimization model.
[0069] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for mining nuclear equipment reliability data, characterized in that: Includes steps: Obtain historical operating data of target nuclear-grade equipment through operating signal acquisition equipment to obtain a historical operating signal data set; By integrating empirical mode decomposition and clustering algorithms, the historical operating condition signal data set is preprocessed, and the preprocessed historical operating condition signal data set is labeled to obtain a labeled data set; Based on the labeled data set, a data mining model is constructed through a machine learning algorithm, and the data mining model is optimized through a quantum genetic algorithm to obtain a data mining optimization model; The real-time operating signal data of the target nuclear-grade equipment is collected through the operating signal acquisition equipment, the real-time operating signal data is input into the data mining optimization model, and the data mining results of the target nuclear-grade equipment are output.
2. The method for nuclear equipment reliability data mining according to claim 1, characterized in that: The specific process of obtaining the historical operating condition data of the target nuclear-grade equipment through the operating condition signal acquisition device and obtaining the historical operating condition signal data set is as follows: The operating condition data of the target nuclear-grade equipment are collected in real time or at regular intervals by deploying the operating condition signal collection equipment on the target nuclear-grade equipment; the collected operating condition data are sorted in chronological order to form a historical operating condition signal data set containing time series information.
3. The nuclear equipment reliability data mining method according to claim 1, characterized in that: The specific process of preprocessing the historical operating condition signal data set by integrating empirical mode decomposition and clustering algorithm is as follows: Perform integrated empirical mode decomposition on the historical operating condition signal data set to obtain the corresponding intrinsic mode data set; Perform feature extraction on the intrinsic mode data set to obtain a feature vector data set; The feature vector data set is clustered by a clustering algorithm to obtain a clustered data set.
4. The method for nuclear equipment reliability data mining according to claim 3, characterized in that: The specific process of labeling the preprocessed historical operating condition signal data set to obtain the labeled data set is as follows: The clustering data set is displayed through a human-computer interaction interface; based on professional knowledge or expert experience, the state of the clustering data set is identified and corresponding labels are set to form a labeled data set.
5. The method for nuclear equipment reliability data mining according to claim 1, characterized in that: The specific process of constructing the working condition data mining model through the machine learning algorithm based on the labeled data set is as follows: The initial data mining model is constructed based on the support vector machine algorithm; the initial data mining model is trained by labeling the data set to obtain the data mining model.
6. The method for nuclear equipment reliability data mining according to claim 1, characterized in that: The specific process of optimizing the data mining model by the quantum genetic algorithm to obtain the data mining optimization model is as follows: Define the input parameters of the quantum genetic algorithm, including the white noise amplitude of the integrated empirical mode decomposition algorithm, the cluster center of the clustering algorithm, and the kernel function parameters of the support vector machine; The fitness function of the quantum genetic algorithm is set as the state recognition or classification accuracy of the data mining model; The input parameters are optimized and iterated through quantum genetic algorithm to obtain the optimal input parameters, and the input parameters are updated with the optimal input parameters to form a data mining optimization model.
7. The method for mining nuclear equipment reliability data according to claim 6, characterized in that: The specific process of optimizing and iterating the input parameters by the quantum genetic algorithm is as follows: The white noise amplitude, cluster center and kernel function parameters are quantum-encoded to form chromosomes, and an initial population consisting of several chromosomes is randomly generated; The initial population is measured to form a binary population, and the fitness of the binary population is evaluated to obtain a fitness evaluation result; Based on the fitness evaluation results, through quantum gate evolution operation and iterative calculation, the solution of the population converges to the optimal solution, and the optimal white noise amplitude, clustering center and kernel function parameters are obtained.
8. The method for nuclear equipment reliability data mining according to claim 7, characterized in that: The specific process of fitness evaluation is as follows: The binary population is converted into a decimal population, the parameter values in the decimal population are brought into the data mining model, and the fitness function value corresponding to each parameter is calculated.
9. The method for nuclear equipment reliability data mining according to claim 7, characterized in that: The clustering algorithm is the K-means clustering algorithm.
10. The method for nuclear equipment reliability data mining according to claim 1, characterized in that: Compare the data mining results with the actual operating conditions of the target nuclear-grade equipment to verify the data mining results; adjust and optimize the data mining optimization model based on the verification results.