Accident identification in nuclear power plants using hidden Markov models

The identification of the type of accident during the early stages of an ac cident in a nuclear power plant is crucial for the selection of the appropr iate response actions. A plant accident can be identified by its time-depen dent patterns, related to the principal variables. The Hidden Markov Model (HMM) can be applied to accident identification, which is a spatial and tem poral pattern-recognition problem. The HMM is created for each accident fro m a set of training data by the maximum-likelihood estimation method, which uses an algorithm that employs both forward and backward chaining, and a B aum-Welch re-estimation algorithm. The accident identification is decided b y calculating which model has the highest probability for the given test da ta. The optimal path for each model at the given observation is found by th e Viterbi algorithm, and the probability of the optimal path is then calcul ated. The system uses a left-to-right HMM, including six states and 22 inpu t variables, to classify eight types of accidents and a normal stale. The s imulation results show that the proposed system identifies the accident typ es correctly. It is also shown that the identification is performed well fo r incomplete input observations caused by sensor faults or by the malfuncti oning of certain equipment. (C) 1999 Elsevier Science Ltd. All rights reser ved.