화학공학소재연구정보센터
Process Safety and Environmental Protection, Vol.131, 331-348, 2019
Development of prediction models for repair and maintenance-related accidents at oil refineries using artificial neural network, fuzzy system, genetic algorithm, and ant colony optimization algorithm
This study presents an accident causation model for repair and maintenance related accidents at oil refineries and proposes the best model for early accident prediction through the integration of artificial neural networks, fuzzy systems, and metaheuristic algorithms. The main factors affecting the occurrence of accidents were classified into six categories: external, internal, executive, behavioral, situational, and work features. These factors were regarded as the input variables. Then three other factors, namely accident type prediction, consequence type prediction, and population density were added to the input variables. Then, the collected data on accidents were reanalyzed by the predefined variables. They were then prepared, quantified and normalized to enter the networks. Regarding non-hybrid models, the perceptron neural network obtained the highest prediction accuracy of 90.9% and the highest rate of prediction precision of 96.19%. Regarding hybrid models, the neural-GA network obtained the highest prediction accuracy of 95.9% and a precision of 96.7%. Among all models, the neural-GA hybrid was the best prediction model. Thus, it is suggested to employ the neural-GA hybrid model to predict the accidents caused by repair and maintenance. (C) 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.