Study On The Safety Assessment In The Petro-chemical Industry On The Basis Of Artificial Neural Network

Petro-chemical industry is one of the key industries for normal operation of the national economy and it is the vital material basis for the social and economic development. However, due to the characteristics of extremely-complex and large-scale production, a petro-chemical plant is obviously an enterprice with high risk. It is of great importance to establish and perfect safety laws and regulations as well as management systems so as to make sure that the production can go on safely. As the key technology of the safety management. safety assessment techniques have been studied by many scholars both at home and abroad in the past decades.Based on the polestar of safety systematic engineering, this dissertation firstly discusses a few notions related to safety science briefly and their relationship as well. Then several widely-applied safety analysis and assessment approaches are concisely evaluated and their drawbacks are discussed at the same time.According to the systematic and scientific principles, from the point of human-machinery-environment model , a study was carried out on the establishment of safety assessment indices system. At the same time, quantification method of assessment indices is probed into.In the part that follows, the basic principles, frameworks, features and capacities of the Artificial Neural Network (ANN) are briefly introduced .The drawbacks in the traditional function-setting evaluation methods are discussed. Up to now, the problems of fixing and changing weight have never been solved perfectly. Based on the analysis of non-linear characteristics of the production system in the petro-chemical industry, the adaptability combining the safety assessment with the non-linear artificial neural network technology is discussed and ANN is believed to be a reasonable and powerful approach to the problems of safety assessment. Several ANN models are studied and Back Propagation Neural Network (BPNN) is found to be the most suitable network model in the assessment. On the basis of safety condition data from a large-scale domestic petroleum refinery company, a safety assessment network model structure is set up by means of MATLAB.