If a loss-of-coolant accident (LOCA) happens in nuclear power plants (NPPs), core cooling capability is maintained and abnormal states are mitigated by various safety-related systems and facilities in NPP. However, in the event that safety injection systems (SISs) among these systems do not function in time, core cooling capability can be lost on account of delayed emergency core coolant injection, and eventually the risk that reactor core is uncovered and damaged can occur. Hence, a technique to predict the time for SIS recovery is considered to be needed to prevent core uncovery and reactor vessel (RV) failure in the LOCA circumstance when SISs do not normally work. In this study, the corresponding time is defined as golden time. As a technique that predicts golden time, deep fuzzy neural networks (DFNNs) [1-3] with rule-dropout is utilized in the study. Briefly, the rule-dropout DFNN, a kind of artificial intelligence technique, is the method that syllogistic fuzzy reasoning through multiconnected fuzzy neural network (FNN) modules is simplified and the fuzzy rule number in every single FNN module is individually adjusted to efficiently improve its inference capability by its multiple modules. Simulated data on the postulated LOCAs in which safety injection does not normally actuated in optimized pressurized reactor 1000 (OPR1000) were applied to the rule-dropout DFNN.

If safety injection systems (SISs) do not work in the event of a loss-of-coolant accident (LOCA), the accident can progress to a severe accident in which the reactor core is exposed and the reactor vessel fails. Therefore, it is considered that a technology that provides recoverable maximum time for SIS actuation is necessary to prevent this progression. In this study, the corresponding time was defined as the golden time. To achieve the objective of accurately predicting the golden time, the prediction was performed using the deep fuzzy neural network (DFNN) with rule-dropout. The DFNN with rule-dropout has an architecture in which many of the fuzzy neural networks (FNNs) are connected and is a method in which the fuzzy rule numbers, which are directly related to the number of nodes in the FNN that affect inference performance, are properly adjusted by a genetic algorithm. The golden time prediction performance of the DFNN model with rule-dropout was better than that of the support vector regression model. By using the prediction result through the proposed DFNN with rule-dropout, it is expected to prevent the aggravation of the accidents by providing the maximum remaining time for SIS recovery, which failed in the LOCA situation.

A serious threat to the integrity of the reactor core, reactor coolant system, or containment is incurred if proper and essential actions to mitigate accidents cannot be taken owing to insufficient information about the internal states of the nuclear power plant (NPP). Therefore, this study was carried out to develop a model capable of mitigating the risk of severe accidents by accurately predicting the internal states of an NPP containment. A deep fuzzy neural network (DFNN) is a method in which syllogistic fuzzy reasoning is relatively efficient and inference capability is enhanced. In this study, the internal states of an NPP containment, hydrogen concentration and pressure, are predicted using a rule-dropout DFNN, as little NPP information is available under the circumstances of severe accidents. In addition, the performance of the proposed rule-dropout DFNN model is compared with that of other fuzzy neural network variations to verify the enhancement in the accuracy of the DFNN. The developed rule-dropout DFNN model is expected to be capable of providing accident monitoring information in advance for accident mitigation, as its prediction error for the hydrogen concentration and pressure in the containment is low. (C) 2021 Elsevier Ltd. All rights reserved.

When a design basis accident occurs in nuclear power plants (NPPs), signals to protect the NPPs generate, safety systems operate, and an accident is alleviated. However, if the safety systems, particularly engineering safety features (ESF), normally operate not, the accident can progress to a severe accident circumstance since the integrity of a reactor core gets worse by loss of its cooling capability. In a severe accident, since a large number of radioactive gases and fission products are released from the reactor core, the reliability of the instrument signals is poor, and then available signals are limited. Hence, it is impossible to take appropriate actions to mitigate the accident. In this study, therefore, a deep fuzzy neural network (DFNN) model was developed that provides information on the integrity of containment through limited information in such accident situations to support successful actions and mitigation. As the containment is the final barrier of defense-in-depth in NPPs, it is important to maintain its integrity. The causes of structural failure of the containment during the severe accident include steam and hydrogen explosion, over-pressurization, and so on [1,2]. In this study, hydrogen (H2) concentration and pressure in the containment, which are regarded as variables for internal states in the containment, were predicted since the circumstance that a threat to the containment occurs due to a degradation in the integrity of the reactor core by loss of coolant accident (LOCA) is mainly considered. The DFNN, as an artificial intelligence methodology used to predict the containment states, is based on a FNN method. The DFNN deeply stacks its FNN modules configured to improve reasoning capability and is a method simplifying syllogistic fuzzy reasoning. The data are numerical data acquired using the modular accident analysis program (MAAP) code [3]. To simulate a severe accident by LOCAs and steam generator tube rupture (SGTR), it is assumed that ESF does not work. In this paper, the prediction results of the containment states using the proposed DFNN model are described. Therefore, the effectiveness of the DFNN, used to monitor the containment states under a severe accident circumstance in NPPs, can be checked.

In the event that any event such as a transient going beyond normal operating condition happens in nuclear power plants (NPPs), accurately recognizing and identifying it is essential to establish necessary actions for early mitigating an undesired state under such a circumstance. Especially, initial identification of events, such as a design basis accident (DBA) circumstance in the NPPs, can be one of the critical requisites to prevent from progression to a severe accident. However, correct identification of accident occurrence locations or types may not be easily done on account of monitoring of too many instrumentation signals related to an accident. Therefore, this study is performed to develop models accurately identifying 9 events in initial time after an accident occurrence, and accordingly artificial intelligence (AI) techniques were used to make the models. Among various machine learning methods based on artificial neural network (ANN) structures as AI techniques, long-short term memory (LSTM) [1] and gated recurrent unit (GRU) [2], which are with the recurrent neural network structure, were utilized in the study. The main reason why these methods were applied is that recurrent neural network structure has an advantage that information in previous steps in its network is relatively well transferred to current and next steps than other methods with typical feedforward network structure (e.g. deep neural network (DNN) or convolutional neural network (CNN) [3]). In addition, since event identification model using the DNN was developed and its result was compared with that of support vector machine (SVM) model in the previous study [4], in an attempt to check performance on various event identification by newly applied methods in the study, the models were developed using the LSTM and GRU. Thus, identification results for 9 initial events of the LSTM and GRU models are shown in this paper. Furthermore, ongoing work on event identification through clustering using an unsupervised learning method, as another AI technique, is briefly indicated in the paper.

The pipe bends and elbows in nuclear power plants (NPPs) are vulnerable to degradation mechanisms and can cause wall-thinning defects. As it is difficult to detect both the defects generated inside the wall-thinned pipes and the preliminary signs, the wall-thinning defects should be accurately estimated to maintain the integrity of NPPs. This paper proposes a deep fuzzy neural network (DFNN) method and estimates the collapse moment of wall-thinned pipe bends and elbows. The proposed model has a simplified structure in which the fuzzy neural network module is repeatedly connected, and it is optimized using the least squares method and genetic algorithm. Numerical data obtained through simulations on the pipe bends and elbows with extrados, intrados, and crown defects were applied to the DFNN model to estimate the collapse moment. The acquired databases were divided into training, optimization, and test datasets and used to train and verify the estimation model. Consequently, the relative root mean square (RMS) errors of the estimated collapse moment at all the defect locations were within 0.25% for the test data. Such a low RMS error indicates that the DFNN model is accurate in estimating the collapse moment for wall-thinned pipe bends and elbows.

Many studies that suggested operator support systems for nuclear power plants (NPPs) have been being carried out. Several operator support systems were designed for effective actions and mitigation in an abnormal state or an accident circumstance. Among them, the systems showed its capability for tasks such as fault detection, diagnosis of an abnormal state or an accident, and prediction of safety-related factors in NPPs by deploying artificial intelligence algorithms. Furthermore, with the introduction of machine learning methods from conventional support vector machines [1], fuzzy neural networks [2] to state-of-the-art deep learning methods with feedforward deep neural network (DNN) [3] or recurrent neural network (RNN) [4] architectures, many methods were able to be applied to various NPP factors, and therefore their performances were shown and being advanced. In an effort to predict a safety-critical NPP factor, long-short term memory (LSTM) neural network [5], of which structure is based on RNNs’, was used to predict reactor vessel (RV) water level under postulated severe accident circumstances of the NPPs. The LSTM was utilized in the study due to the fact that it is well known for its better stability for time series prediction in largescale networks, and less vulnerable to vanishing gradient problem than typical RNNs. For application to the LSTM and establishment of a prediction model, modular accident analysis program (MAAP) code [6] was used to obtain simulated data. The data were comprised of time-dependent values of variables gained by simulating the severe accident circumstances originated from postulated loss-ofcoolant accidents (LOCAs) and steam generator tube rupture (SGTR). In this paper, prediction performance of an established LSTM model is presented when limited instrumentation signals from the aforementioned simulated data were applied. In addition, prediction performance of two deep learning methods for a NPP factor can be assessed by comparing the proposed LSTM model with the DNN model designed in previous study [7].

Fukushima accident was worse by instrument inability. Eventually, the accident was not mitigated and keeping the integrity of reactor was failed since the operators was not able to quickly understand the situation and take necessary actions. Therefore, in this study, reactor vessel (RV) water level considered as one of the parameters to keep the integrity of reactor is predicted in loss of coolant accident (LOCA) situation using the deep neural network (DNN) method. This is in an effort to provide supporting information under the severe circumstance. The simulation data obtained from modular accident analysis program (MAAP) are applied to the DNN method to check the prediction performance of the RV water level. The prediction performance of RV water level using the proposed DNN model is presented as root mean square error (RMSE). Although the data of several circumstances among a variety of LOCAs are applied, good prediction performance is shown using the proposed DNN method.

Acquiring instrumentation signals generated from nuclear power plants (NPPs) is essential to maintain nuclear reactor integrity or to mitigate an abnormal state under normal operating conditions or severe accident circumstances. However, various safety-critical instrumentation signals from NPPs cannot be accurately measured on account of instrument degradation or failure under severe accident circumstances. Reactor vessel (RV) water level, which is an accident monitoring variable directly related to reactor cooling and prevention of core exposure, was predicted by applying a few signals to deep neural networks (DNNs) during severe accidents in NPPs. Signal data were obtained by simulating the postulated loss-of-coolant accidents at hot- and cold-legs, and steam generator tube rupture using modular accident analysis program code as actual NPP accidents rarely happen. To optimize the DNN model for RV water level prediction, a genetic algorithm was used to select the numbers of hidden layers and nodes. The proposed DNN model had a small root mean square error for RV water level prediction, and performed better than the cascaded fuzzy neural network model of the previous study. Consequently, the DNN model is considered to perform well enough to provide supporting information on the RV water level to operators.

Nuclear power plants (NPPs) are composed of very large complex systems. During transient occurrences in NPPs, operators determine the transients of the NPP through information acquired from various measuring instruments. A support vector machine (SVM) based on serial and parallel connections, termed as a multiconnected SVM, is introduced in this paper. The loss of coolant accidents (LOCAs) was identified and their break sizes are estimated using the multiconnected SVM model. The optimal parameter values of the multiconnected SVM models are obtained using a genetic algorithm. In this paper, the modular accident analysis program code was used to simulate the severe accidents occurring due to a variety of design basis accidents. The proposed algorithm uses the short time-integrated simulated sensor signals just after the reactor trip. The results show that the multiconnected SVM model can identify LOCAs and estimate their break sizes accurately. It is expected that the LOCA identification and the accurate estimation of the break size are useful for NPP operators when they try to manage severe accidents.

Residual stress is a critical element in determining the integrity of parts and the lifetime of welded structures. It is necessary to estimate the residual stress of a welding zone because residual stress is a major reason for the generation of primary water stress corrosion cracking in nuclear power plants. That is, it is necessary to estimate the distribution of the residual stress in welding of dissimilar metals under manifold welding conditions. In this study, a cascaded support vector regression (CSVR) model was presented to estimate the residual stress of a welding zone. The CSVR model was serially and consecutively structured in terms of SVR modules. Using numerical data obtained from finite element analysis by a subtractive clustering method, learning data that explained the characteristic behavior of the residual stress of a welding zone were selected to optimize the proposed model. The results suggest that the CSVR model yielded a better estimation performance when compared with a classic SVR model.