Research On Improved Bayesian Algorithm And Parameter Model Of Cement Grate Cooler Heat Transfer System

Grate cooler is the key equipment in process of cement production.It is responsible for cooling the high temperature clinker,improving calcination conditions and recycling heat.However,it is difficult to establish an accurate model for grate cooler heat transfer system because of its complex working conditions,multiple and mutual coupling parameters,and few internal measurement and control points.Bayesian network combines probability theory with graph theory,has powerful reasoning ability and convenient decision-making mechanism,and has obvious advantages in dealing with uncertainties.Therefore,the Bayesian network is applied to the grate cooler heat transfer system,and the fault diagnosis and state prediction of the grate cooler’s key parameters grate pressure are studied.The specific research work are as follows:Firstly,the research background and significance of the project are briefly introduced,the research status at home and abroad of Bayesian structure learning algorithm,Bayesian prediction algorithm,the fault diagnosis and prediction of grate cooler is analyzed,and the main research contents are elaborated.Secondly,Aiming at the problem that hill-climbing and greedy algorithm need to search more and have poor ability in global optimization,an improved batch Bayesian structure algorithm is constructed.Simulation comparison shows that this algorithm has the characteristics of not relying on prior knowledge,high accuracy and high efficiency.And aiming at the shortcomings of incremental maintenance performance of batch algorithm,an incremental learning algorithm of Bayesian network was proposed.Simulation comparison shows that the algorithm can incrementally maintain the network and save space and time to a certain extent.It lays the foundation for the establishment of the fault diagnosis model of grate cooler.Thirdly,we proposed improved multi-population genetic algorithm which optimizes hidden Markov model.This algorithm improves the traditional multi-population algorithm from the design of crossover,mutation,immigration operators and so on,and optimizes the parameters of hidden Markov model by using the improved algorithm.The research results show that the improved algorithm can converge to the global optimum and improveconvergence accuracy and speed.It lays the foundation for the establishment of technological parameter prediction model of grate cooler.Finally,the principle of the grate cooler heat transfer system is analyzed,eight important process parameters are selected and the actual data are quantified.On the one hand,improved algorithms are used to establish the static and incremental model of parameters of the grate cooler heat transfer system,and EM algorithm is used to learn the parameters of the model.We use the joint tree reasoning algorithm to diagnose the grate pressure,find out the cause of the fault and take measures.Experiments show that the model has high diagnostic accuracy.On the other hand,the improved algorithm is used to establish the prediction model of the grate pressure,analyze and compare the actual data state and the prediction state.Experiments show that the model can predict grate pressure accurately and effectively.