Study On The Multi-Agent Based Model Of Dangerous Chemicals Risk Assessment

Along with the understanding of the hurt of dangerous chemicals on humans and the environment, the concept of "Risk Assessment" has been introduced into the area of chemicals management, in order to put forward the management, prevention and disposal measures on the basis of qualitative and quantitative analysis of the nature of dangerous chemicals. In short, "prevention is better than cure" is of great significance.This research is based on the dangerous chemicals risk assessment, the objective is to get the better reflection of the experts' subjective evaluation and the evaluated objects' objective nature, and make the decision-making managers be supported better by the evaluation results.As one of the core conceptions of Artificial Intelligence, Agent is introduced into the whole process of risk assessment in this paper. Using the characters of coordination and auto-learning, the function of dangerous chemicals risk assessment is realized. A Multi-Agent based model of dangerous chemicals risk assessment is established; its work flow and the realization of the combination of subjective experience and objective evaluation are explained.In this paper, the detailed process of risk assessment, including the selection of index system and the establishment of weights are studied. While selecting the evaluation methods, taking into account the inconsistency of results using different evaluation methods and the one-sidedness of just using one evaluation method, this paper presents a selection method based on fuzzy clustering analysis to evaluate the compatible method set, reflected in the Multi-Agent system, the simplified acquaintance-model can be used to the evaluation method to achieve the communication and collaboration between Agent. Finally, a specific application is given.Another focus of this paper is how to deal with the results of different methods in order to get a more comprehensive result. On the basis of analysis of various combination methods, this paper proposes a method of using fuzzy Borda to get the comprehensive and integrated conclusion from the single evaluation methods. The conclusion is adjusted to meet the needs of different user preferences, and the adjusting rules are given in order to achieve the feasibility and reproducibility of the evaluation process.