Study On Optimization Algorithm Of Fault Diagnosis Based On Improved Bayesian Network

Bayesian Network is the most effective ideal model in the field of uncertain knowledge expression and reasoning,it has a great advantage in solving the problems caused by complex uncertain factors,and it is widely used in data mining,image processing,industrial fault diagnosis and machine learning.Industrial fault diagnosis is of great significance to industrial production.The development of artificial intelligence provides an intelligent diagnostic method for fault diagnosis and has been successfully applied.This project plans to focus on the fault diagnosis optimization algorithm based on bayesian network,and verify the feasibility of the algorithm in actual fault diagnosis based on data collected from jidong cement plant.The specific research contents are as follows:Firstly,this paper proposes an improved ant colony optimization algorithm aiming at the defect of large space in search space of ant colony algorithm and easy to fall into local optimum.The proposed algorithm based on mutual information to determine the initial network first,isolated node processing at the same time,and by changing the pheromone update methods,and to join the penalty function,reduce the redundancy number of edges of the algorithm and makes the algorithm more tend to score better results.Finally,the performance of the algorithm is verified by simulation.Secondly,based on Tabu Search algorithm and improved algorithm research,in order to solve the problem that Tabu Search algorithm is too dependent on initial solution and slow convergence speed,an improved Tabu Search optimization algorithm is proposed.Firstly,the initial solution is determined by calculating the mutual information and relative entropy.Then,the neighborhood solution set is generated by adding edges,subtracting edges and rotating edges,so as to avoid circulation through the centralized and diversified search strategy.Finally,the optimal bayesian network structure is output according to the scoring function.The accuracy and execution time of the algorithm are obtained by simulation.Finally,analysis of the process parameters of rotary kiln,and according to thehigh degree of correlation coefficient to select relevant parameters as variables,bayesian network node will collect data of cement quantitative screening,using two kinds of improved algorithm to build fault diagnosis bayesian model structure of cement rotary kiln.The classical Maximum Likelihood Estimation method and Variable Elimination method are used for parameter learning and diagnostic reasoning.According to the results of data experiment,the two improved bayesian network structure learning algorithms have high accuracy and efficiency in the fault diagnosis of cement rotary kiln.