Researchon Chemical Accident Analysis Technologyfor Fault Tree Clustering

Frequent chemical accidents have seriously endangered the country's security and people's lives.Every time an accident occurs,security experts form an investigation report of the accident through extensive investigation and analysis,and store it in the form of fault tree,event tree,reliability block diagram,FME(C)A,etc.,forming a model library with a large number of accident causal analysis.However,the data integration of these expert experiences has not yet formed a set of automated,intelligent auxiliary analysis mechanisms,resulting that new accident analysis still relies on the experience of accident analysts,and no effective mechanism to verify the integrity and validity of the accident analysis.This paper aims at a large number of accident cases aggregated with the fault tree model,uses text mining technology to analyze the grammar and semantic information of fault tree nodes,and uses probabilistic graph models to analyze the accuracy and completeness of event evolution to form a fault tree model-based clustering method.The analysis system that automatically learns from historical cases and assists in new accidents is realized.The main work of this paper is as follows:1.Using web crawler technology to obtain a large number of chemical accident reports,build a basic vocabulary in the field,and adopt Word2 vec to train word vectors with domain features and generate fault tree node word vectors through techniques such as Chinese word segmentation,text preprocessing,and feature extraction.Establish an abstract interpretation relation model between abstract fault tree and case fault tree,and give a similarity calculation method between abstract node and case node,construct extended cutset of fault tree,support similarity calculation of fault tree structure,implement fault tree-based clustering analysis.2.Hidden Markov model prediction method is used to construct the Hidden Markov Model structural similarity matching technique is proposed to automatically match the structure of the case fault tree and the standard fault tree.Using the Viterbi algorithm to predict the sequence of the largest possible match,and can automatically test defect modes such as missing,redundant,and error modes of the fault tree,and realize the structural mapping problem between the case fault tree and the standard fault tree.Finally implement the clustering function of the fault tree model.3.Based on the above algorithm,a chemical accident analysis system based on fault tree model was designed and developed.It can realize automatic cluster analysis of a large number of case fault trees,achieve clustering based on abstract fault trees,and can also support the analysis of new accidents.It can support the structural integrity analysis,logic gate defect analysis of new accidents and recommendation of event evolution.