دانلود رایگان ترجمه مقاله پیش بینی غلظت هیدروژن در مهار حوادث شدید با شبکه عصبی فازی (نشریه الزویر 2015)

پیش بینی غلظت هیدروژن در مهار حوادث شدید با استفاده از شبکه عصبی فازی

Prediction of hydrogen concentration in containment during severe accidents using fuzzy neural network

Abstract

Recently, severe accidents in nuclear power plants (NPPs) have become a global concern. The aim of this paper is to predict the hydrogen buildup within containment resulting from severe accidents. The prediction was based on NPPs of an optimized power reactor 1,000. The increase in the hydrogen concentration in severe accidents is one of the major factors that threaten the integrity of the containment. A method using a fuzzy neural network (FNN) was applied to predict the hydrogen concentration in the containment. The FNN model was developed and verified based on simulation data acquired by simulating MAAP4 code for optimized power reactor 1,000. The FNN model is expected to assist operators to prevent a hydrogen explosion in severe accident situations and manage the accident properly because they are able to predict the changes in the trend of hydrogen concentration at the beginning of real accidents by using the developed FNN model.

1 Introduction

Recently, severe accidents in nuclear power plants (NPPs) have become a global concern. In the event of severe accidents, the major safety parameters of nuclear reactors change rapidly during the initial stages, leaving operators with insufficient time to devise an appropriate response. The efficient management of a serious accident requires observation of the key parameters during the very brief duration of initial events by establishing scenarios and initial events leading up to the accident. In particular, it is extremely important to determine safety-related parameters and critical information during the extremely short period following a loss of coolant accident (LOCA) and steam generator tube rupture (SGTR). This would enable verification of NPP status and determination of appropriate corrective action.

In case of severe accidents, the NPP operators are concerned about hydrogen explosion due to hydrogen accumulation in containment. Hydrogen is accumulated in containment by leakage from the primary pressure boundary Therefore, this work considered severe incidents that were caused by LOCAs, which were analyzed by using data from optimized power reactor 1,000 (OPR1000). The work aimed to predict the hydrogen concentration in the event of a severe accident. The increase in the hydrogen concentration is one of the factors threatening the integrity of the containment. The hydrogen inside the containment is generated by the radioactivation of water in the atmosphere, corrosion of the inner material of the containment by containment spray, and reaction of steam with the zirconium cladding. Maintaining the integrity of the containment by preventing the hydrogen within from exploding would require the local hydrogen concentration to be retained below 4%.

Therefore, in this study, various artificial intelligence (AI) methods were examined to predict changes in the hydrogen concentration. It was determined that a method using a fuzzy neural network (FNN) was the most suitable for predicting the hydrogen concentration. A number of AI techniques have been applied successfully to a variety of research fields of nuclear engineering, such as signal validation [1e3], plant diagnostics [4e7], event identification [8e10], and smart sensing (or function approximation) [11e13]. Many of the previous works used fuzzy inference systems (FISs) and neural networks (NNs). Jang and Sun [14] demonstrated the functional equivalence between NNs and FISs in cases when the activation functions of the NNs and the membership function of the FIS are the same.

An FNN is a data-based model that requires data for its development and verification. As data from real severe accidents do not exist, it is necessary to use numerical simulations to obtain the required data for the proposed model. The FNN model was verified based on the NPP simulation data acquired using MAAP4 code [15]. The successful management of NPPs as a result of the ability to rapidly predict safetycritical parameters during real accidents could lead to the safekeeping of NPPs.