Fault Diagnosis Of Nuclear Power Plant Research Articles

Enhancing interpretability in neural networks for nuclear power plant fault diagnosis: A comprehensive analysis and improvement approach

Deep neural networks, as applied in the field of nuclear power fault diagnosis, have garnered significant attention alongside advancements in artificial intelligence technology. However, the “black box” nature of deep learning models has raised concerns regarding their deployment in scenarios demanding high safety standards, such as nuclear power plants. In this paper, we innovatively propose the utilization of an explainable artificial intelligence method, grounded in game theory, to conduct a detailed analysis of the diagnostic behavior of neural network models applied to nuclear power plants. By leveraging SHAP, the decision-making processes of these opaque models are demystified, offering insights into how and why they arrive at their predictions. In this study, two of the most widely utilized neural network frameworks are applied across six representative fault diagnosis cases for a comprehensive analysis. Based on its analysis results, a novel SHAP-Enhanced Feature Selection for Efficient Neural Network Fault Diagnosis strategy is proposed, significantly reducing model complexity without sacrificing diagnostic performance. This research breaks through the limitations of data-driven models used in the field of nuclear power fault diagnosis, conducts an interpretable analysis of their behavior, and proposes an improved strategy based on the analysis results, contributing to the practical application of data-driven models in nuclear power.

Hierarchical FFT-LSTM-GCN based model for nuclear power plant fault diagnosis considering spatio-temporal features fusion

As safety-critical infrastructure, nuclear power plants (NPPs) require enhanced safety measures and risk minimization. To achieve this goal and to aid operator decision-making while reducing human error, various fault diagnosis (FD) methods have been proposed. Among these, deep learning-based approaches have demonstrated significant success in FD because of their ability to effectively extract information about machine degradation. However, most existing methods primarily focus on temporal features while neglecting spatial features. To leverage both temporal and spatial features effectively and achieve high diagnostic accuracy, we propose a hierarchical deep learning based model that comprises the fast Fourier transform (FFT), long short-term memory networks (LSTM) and graph convolutional networks (GCN). The application of FFT to sensor sequences effectively mitigates their volatility. The use of GCN enables automated extraction of intricate spatial features from multi-sensor data, while LSTM is adept at directly extracting temporal features from historical input data. To validate the proposed model, we conducted three experiments using data simulated by the personal computer transient analyzer (PCTRAN), and the results demonstrate that the diagnostic accuracy of the proposed hierarchical FFT-LSTM-GCN model surpasses that of any single model for NPP FD.

Open set recognition fault diagnosis framework based on convolutional prototype learning network for nuclear power plants

As an Open Set Recognition (OSR) problem, nuclear power plant fault diagnosis requires not only the correct classification of known faults, but also the effective identification of unknown faults, which cannot be satisfied by most previously developed models. In this study, a novel nuclear power plant OSR fault diagnosis framework based on Convolutional Prototype Learning (CPL) is proposed, in which CPL is used to extract discriminative fault features from raw nuclear power plant data. Besides, two classification methods, One-Class Support Vector Machine (OCSVM) method and Prototype Matching by Distance (PMD) method, are developed to complete the fault diagnosis of open space for the framework. To verify the feasibility and effectiveness of the proposed OSR framework, numerical experiments on 24 OSR tasks with high-dimensional and strong-nonlinear complex nuclear power plant simulation data are conducted, and the OSR performance evaluate by normalized accuracy and Youden's index, feature visualization and convergence rates are analyzed. Results show that compared with adopting traditional Convolutional Neural Network (CNN), the proposed CPL-based framework can significantly boost the OSR diagnostic performance on nuclear power plant fault diagnosis tasks, which benefits from the excellent ability of CPL to extract intra-class compact and inter-class separable feature representation.

ECOC-based integrated learning method for fault diagnosis in nuclear power plants

Abstract The fault diagnosis system of nuclear power plants plays an important role in ensuring the safety and economy of nuclear power plant operations. This paper first analyzes typical faults of nuclear power plants and their phenomena, and fault samples are obtained. A comprehensive study of the structure of the nuclear power plant system, its working mode and the association between each subsystem is carried out to analyze the monitoring parameters and fault characteristics and establish the fault data set. Secondly, an IFWA (Improved Fireworks Algorithm - Integrated Learning) algorithm is proposed to assess the severity of faults in the first circuit of a nuclear power plant. Finally, the fault diagnosis module is divided into three units according to the functional logic, i.e., condition monitoring unit, fault identification unit, and fault severity assessment unit. The results show that the diagnostic accuracy of the IFWA algorithm is 94.25% for SGTR in the single-fault diagnosis experiment and 96.25% for SGTR-LOCA in the multiple-fault diagnosis experiment. It shows that the IFWA algorithm proposed in this paper has the optimal performance capability when applied to nuclear power plant fault diagnosis and effectively assists managers in diagnosing faults and giving maintenance recommendations.

Generalization analysis and improvement of CNN-based nuclear power plant fault diagnosis model under varying power levels

The domain discrepancy caused by power level gap between training set (source domain) and test set (target domain) limits the generalization of data-driven models in practical nuclear power plant fault diagnosis applications. In this study, a highly fine-grained quantitative generalization description of Convolutional Neural Network (CNN) model is conducted with high-dimensional and strong-nonlinear complex nuclear power plant simulation data. Results show that with source domain of single power level, CNN suffers from over-fitting and poor domain discrepancy generalization; with source domain of multiple power levels, the sensitivity of CNN to minor domain discrepancy is greatly reduced and its generalization is significantly boosted. Besides, a novel Artificial Disturbance Method (ADM) based domain discrepancy generalization promotion framework is proposed in this study, which alleviates the over-fitting of CNN by adding disturbed training data to its training set. The feasibility and superiority of the framework is proven when Gaussian noise disturbance or uniformly distributed noise disturbance is adopted to generate disturbed training data. The ADM-based framework reduces the requirement for the number of power levels contained in source domain, this advantage endows it with strong practicability in actual nuclear power plant fault diagnosis tasks where the available source domain data are extremely limited.

A novel transfer CNN with spatiotemporal input for accurate nuclear power fault diagnosis under different operating conditions

Deep Neural Network (DNN) models, recognized for their exceptional feature extraction and end-to-end diagnostic capabilities, have gained considerable attention from researchers in the realm of nuclear power fault diagnosis. However, existing deep-learning based diagnostic technologies present certain challenges when applied to practical engineering scenarios. A significant issue is the limited adaptability of data-based machine learning diagnostic methods to dynamic and evolving operational conditions, leading to inconsistent accuracy under different operating parameters. To solve this problem, a spatiotemporal Convolutional Neural Network (CNN) model is proposed. A novel input scheme, considering both spatial and temporal aspects, is specially designed for nuclear power plant fault diagnosis and MLP convolutional layer are used to improve the feature extraction capability of the model. Finally, the performance of the model is analyzed through a series of experiments based on simulation data and data visualization. The results show that the proposed method outperforms traditional CNNs in diagnosing nuclear power system faults with higher accuracy. Furthermore, by employing a transfer learning strategy, the model exhibits improved speed for cross-operating condition fault diagnosis.

Fault Diagnosis of Nuclear Power Plant Based on Sparrow Search Algorithm Optimized CNN-LSTM Neural Network

Nuclear power is a type of clean and green energy; however, there is a risk of radioactive material leakage when accidents occur. When radioactive material leaks from nuclear power plants, it has a great impact on the environment and personnel safety. In order to enhance the safety of nuclear power plants and support the operator’s decisions under accidental circumstances, this paper proposes a fault diagnosis method for nuclear power plants based on the sparrow search algorithm (SSA) optimized by the CNN-LSTM network. Firstly, the convolutional neural network (CNN) was used to extract features from the data before they were then combined with the long short-term memory (LSTM) neural network to process time series data and form a CNN-LSTM model. Some of the parameters in the LSTM neural network need to be manually tuned based on experience, and the settings of these parameters have a great impact on the overall model results. Therefore, this paper selected the sparrow search algorithm with a strong search capability and fast convergence to automatically search for the hand-tuned parameters in the CNN-LSTM model, and finally obtain the SSA-CNN-LSTM model. This model can classify the types of accidents that occur in nuclear power plants to reduce the nuclear safety hazards caused by human error. The experimental data are from a personal computer transient analyzer (PCTRAN). The results show that the classification accuracy of the SSA-CNN-LSTM model for the nuclear power plant fault classification problem is as high as 98.24%, which is 4.80% and 3.14% higher compared with the LSTM neural network and CNN-LSTM model, respectively. The superiority of the sparrow search algorithm for optimizing model parameters and the feasibility and accuracy of the SSA-CNN-LSTM model for nuclear power plant fault diagnosis were verified.

Development and Validation of a Nuclear Power Plant Fault Diagnosis System Based on Deep Learning

As artificial intelligence technology has progressed, numerous businesses have used intelligent diagnostic technology. This study developed a deep LSTM neural network for a nuclear power plant to defect diagnostics. PCTRAN is used to accomplish data extraction for distinct faults and varied fault degrees of the PCTRAN code, and some essential nuclear parameters are chosen as feature quantities. The training, validation, and test sets are collected using random sampling at a ratio of 7:1:2, and the proper hyperparameters are selected to construct the deep LSTM neural network. The test findings indicate that the fault identification rate of the nuclear power plant fault diagnostic model based on a deep LSTM neural network is more than 99 percent, first validating the applicability of a deep LSTM neural network for a nuclear power plant fault-diagnosis model.

Transfer learning network for nuclear power plant fault diagnosis with unlabeled data under varying operating conditions

The performance of data-driven nuclear power plant fault diagnosis models trained by limited data cannot be guaranteed, the distribution discrepancy between training set (source domain) and test set (target domain) caused by varying operating conditions will seriously restrict their practical applications. To generalize the diagnostic knowledge learnt from labeled source domain to unlabeled target domain, a novel transfer learning method based on Maximum Mean Discrepancy (MMD) and Convolutional Neural Network (CNN) is proposed, which can reduce the domain discrepancy of the extracted features between source domain and target domain, by appending the MMD-based distribution discrepancy to the objective function of CNN. Numerical experiments with high-dimensional and strong-nonlinear complex nuclear power plant simulation data are conducted, results show that significant improvement in the diagnostic accuracy of target domain can be achieved on most transfer tasks. Besides, the influence of adopting different kernel functions (Linear kernel, Sigmoid kernel, Laplace kernel and Gaussian kernel) to calculate MMD is also studied, better transfer effect can be achieved when Gaussian kernel is used, including higher accuracy, faster convergence rate and less negative transfer. Overall, the proposed method is promising to expand the application scope of nuclear power plant intelligent fault diagnosis to varying operating conditions.

Transfer learning with limited labeled data for fault diagnosis in nuclear power plants

Numerous achievements have been made in intelligent fault diagnosis of nuclear power plants, under the assumption that the labeled training data is sufficient and in the same distribution as the test data, which is always contrary to actual situations and limits their practical applications. Therefore, a novel transfer learning framework with diverse transfer strategies is proposed on basis of a pre-trained CNN model in this study, to deal with the problem of limited labeled data in target tasks. The pre-trained CNN is obtained by ample data in source task, then its weights are transferred to the target models and fine-tuned. The results show that with proper transfer strategy, the proposed transfer learning framework significantly boosts the diagnostic performances, compared with training a brand-new CNN model from scratch with insufficient labeled target data. Consequently, the feasibility and superiority of transfer learning in nuclear power plant fault diagnosis with limited labeled data are proven.

Research on robustness of five typical data-driven fault diagnosis models for nuclear power plants

Data-driven models endowed with high flexibility and practicality are becoming the mainstream of nuclear power plant fault diagnosis. However, actual field data can be tainted by undesirable and unpredictable noise or disturbances, especially under fault conditions, which brings enormous challenges to performances of the data-driven models. Consequently, it is of great practical significance to study the robustness of data-driven models. In this study, data-driven models based on SVM, RF, XGBoost, FCNN and CNN are established respectively, and multiple test sets are generated pertinently, on which the robustness of the models is explored. Besides, a novel model training method named “Add Noisy Data” (AND) is proposed to improve the anti-noise ability of the models by adding noisy data to the training set. The results show that the robustness of the models varies greatly under different data taint modes, and the AND method can significantly improve the anti-noise robustness of RF model.

Advanced fault diagnosis method for nuclear power plant based on convolutional gated recurrent network and enhanced particle swarm optimization

A predictive approach to fault diagnosis in complex systems such as the Nuclear power plant (NPP) is becoming popular because of the efficiency and accuracy it presents. However, there is still a huge gap between the proposed fault diagnosis techniques and engineering applications. To further optimize the fault diagnosis route and encourage real-time application, this paper presents a highly accurate and adaptable fault diagnosis technique based on the convolutional gated recurrent unit (CGRU) and enhanced particle swarm optimization (EPSO). Stacking convolutional kernel and GRU results in a model that speedily extract the local characteristics and learn the time-series information. The EPSO is utilized to adaptively search for optimal hyper-parameters for the CGRU. Finally, the accuracy is evaluated on a dataset obtained from experiments, and comparative analysis of the proposed model with existing architectures and models are presented. Relevant research results that show the usefulness of the proposed model are also presented, which highlights the enhanced intelligence and information level achieved in the NPP fault diagnosis.

Knowledge base operator support system for nuclear power plant fault diagnosis

A high-tech, high-performance system such as the nuclear power plant needs a wide range of support for operators to efficiently operate the plant, interpret and manage the volume of information available, and detect and diagnose fault in a timely manner. Increasingly, application of artificial neural networks and its variants for fault detection and isolation has moved from toy examples to real-world systems. However, different network architectures respond to different data set in different ways, and the complex, dynamic, high background noise, overlapping patterns and non-linear characteristics of Nuclear Power Plants (NPP) requires a careful selection of a suitable neural network architecture that reflects these traits. This work presents a pilot scheme towards the development of a comprehensive knowledge base for the operator support system of the Chinese Qinshan II NPP, using Principal Component Analysis (PCA) and artificial neural networks. In this work, we utilize the PCA method for noise filtering in the pre-diagnostic stage, and evaluate the predictive/regression capability of two different recurrent neural networks – The Elman neural network and the Radial Basis Network – on a representative data from Qinshan II NPP. The process was validated using data from different fault scenarios simulated on a desktop Pressurized Water Reactor simulator, and the fault signatures were used as the input. The predictive outputs required are the location and sizes of the faults. The result shows that the Radial Basis network gives better prediction and diagnoses the faults faster than Elman neural network. Some of the important diagnostic results obtained from the networks are presented in this paper, and they serve as the pilot study for the development of knowledge base for the computerized NPP operator support system.

Analysis of Nuclear Power Plant Transient Recognition Method Validation Utility Strategy

Compare and analysis the transient identification in Nuclear Power Plant fault diagnosis method. According to how does these methods for nuclear power plant systems HTR effect, Further proposed to a problem that towards existing transient recognition of methods to verification effectiveness. Preliminary analysis of the three kinds of validation strategy, after a comparative demonstration, it is formed the preliminary research on the utility analysis framework. Utility to verify the framework for the Nuclear Power Plant fault diagnosis methods carry out intensive research provides a theoretical basis.

Research and design of distributed fault diagnosis system in nuclear power plant

Nuclear power plant (NPP) is a complex system with abundant operation data and various fault types. Moreover, in most cases, change of system parameters and prompt of the alarm system can not necessarily tell us directly what types the fault belongs to or where it lies in. When multiple faults occur, there is no one-to-one relationship between the fault symptom and the fault itself. Furthermore, the degree and response rate of the fault are different, so some faults are gradual and some are sudden. It is thus clear that for different faults, we need to conduct combined diagnosis with a variety of diagnostic methods to ensure accuracy and instantaneity in NPP fault diagnosis. According to the characteristic of the distributed function of equipment and the digital instrument and control (I&C) system, we studied and designed the distributed condition monitoring and fault diagnosis system in NPP. Based on the “disassemble-synthesizing” diagnostic idea, this paper proposed an intelligent diagnosis method which applied the fuzzy neural network (FNN) in doing local diagnosis and multi-source information fusion technology in global diagnosis. The simulation result showed that this method can quickly and accurately complete the tasks of diagnosing different levels of the single fault and different types of multiple faults.

Nuclear power plant fault diagnosis based on genetic-RBF neural network

It is necessary to develop an automatic fault diagnosis system to avoid a possible nuclear disaster caused by an inaccurate fault diagnosis in the nuclear power plant by the operator. Because Radial Basis Function Neural Network (RBFNN) has the characteristics of optimal approximation and global approximation. The mixed coding of binary system and decimal system is introduced to the structure and parameters of RBFNN, which is trained in course of the genetic optimization. Finally, a fault diagnosis system according to the frequent faults in condensation and feed water system of nuclear power plant is set up. As a result, Genetic-RBF Neural Network (GRBFNN) makes the neural network smaller in size and higher in generalization ability. The diagnosis speed and accuracy are also improved.

Application of data fusion method to fault diagnosis of nuclear power plant

The work condition of nuclear power plant (NPP) is very bad, which makes it has faults easily. In order to diagnose the faults real time, the fusion diagnosis system is built. The data fusion fault diagnosis system adopts data fusion method and divides the fault diagnosis into three levels, which are data fusion level, feature level and decision level. The feature level uses three parallel neural networks whose structures are the same. The purpose of using neural networks is mainly to get basic probability assignment (BPA) of D-S evidence theory, and the neural networks in feature level are used for local diagnosis. D-S evidence theory is adopted to integrate the local diagnosis results in decision level. The reactor coolant system is the study object and we choose 2# steam generator U-tubes break of the reactor coolant system as a diagnostic example. The experiments prove that the fusion diagnosis system can satisfy the fault diagnosis requirement of complicated system, and verify that the fusion fault diagnosis system can realize the fault diagnosis of NPP on line timely.

Application of genetic algorithms to fault diagnosis in nuclear power plants

A nuclear power plant (NPP) is a complex and highly reliable special system. Without expert knowledge, fault confirmation in the NPP can be prevented by illusive and real-time signals. A new method of fault diagnosis, based on genetic algorithms (GAs) has been developed to resolve this problem. This NPP fault diagnosis method combines GAs and classical probability with an expert knowledge base. By assessing the state of the NPP, the GA colony undergoes a transformation that produces an individual adapted to the NPP's condition. Experiments performed on the 950 MW full size simulator at the Beijing NPP simulation training center show that this method has comparative adaptability to diagnose signals and faults changed with time, imperfect expert knowledge, illusive signals and other phenomena.

Nuclear power plant fault diagnosis using neural networks with error estimation by series association

The accuracy of the diagnosis obtained from a nuclear power plant fault-diagnostic advisor using neural networks is addressed in this paper in order to ensure the credibility of the diagnosis. A new error estimation scheme called error estimation by series association provides a measure of the accuracy associated with the advisor's diagnoses. This error estimation is performed by a secondary neural network that is fed both the input features for and the outputs of the advisor. Our error estimation by series association outperforms previous error estimation techniques in providing more accurate confidence information with considerably reduced computational requirements. We demonstrate the extensive usability of our method by applying it to a complicated transient recognition problem of 33 transient scenarios. The simulated transient data at different severities consists of 25 distinct transients for the Duane Arnold Energy Center nuclear power station ranging from a main steam line break to anticipated transient without scram (ATWS) conditions. The fault-diagnostic advisor system with the secondary error prediction network is tested on the transients at various severity levels and degraded noise conditions. The results show that our error estimation scheme provides a useful measure of the validity of the advisor's output or diagnosis with considerable reduction in computational requirements over previous error estimation schemes.

Error Estimation by Series Association for Neural Network Systems

Estimation of confidence intervals for neural network outputs is important when the uncertainty of a neural network system must be addressed for safety or reliability. This paper presents a new approach for estimating confidence intervals, which can help users validate neural network outputs. The estimation of confidence intervals, called error estimation by series association, is performed by a supplementary neural network trained to predict the error of the main neural network using input features and the output of the main network. The accuracy of this approach is shown using a simple nonlinear mapping and more complicated, realistic nuclear power plant fault diagnosis problems. The results demonstrate that the approach performs confidence estimation successfully.