Research On Improved Bayesian Network Algorithm And Fault Diagnosis Of Grate Cooler

Bayesian network is a powerful probabilistic graphical model tool in the field of uncertainty,which is built by structure learning algorithm,parameter learning algorithm and probabilistic inference algorithm.The current research hotspot of the Bayesian network is to establish efficient and accurate Bayesian network.Considering these algorithms have low efficiency and tend to be trapped by local optimizations,two improved Bayesian network algorithms are proposed,they are improved structure learning algorithm and inference algorithm.The grate cooler is used for cooling clinker and heat recovery,which is the key equipment in cement production.Due to the process parameters of the grate cooler system are disordered and complicated,it is difficult to diagnose the fault of the grate cooler.Therefore,this paper applies the improved Bayesian network algorithm to build the grate cooler network model for fault diagnosis.The work is described in detail as follows:First of all,an improved wolf pack algorithm is proposed,which is used to optimize the order of nodes.The algorithm first uses the depth first search strategy to generate the node topological sort of the largest supporting tree,and based on this,the initial population was constructed.Then the dynamic mutation and optimal crossover operators are proposed to construct the improved predator behavior,which is suitable for optimizing order of nodes.The adaptive dynamic factor is used to balance the local and global optimization ability of the algorithm,and the optimal order of nodes is obtained through iteration.Combined with the K2 algorithm,IWPN gets the optimal structure of the Bayesian network.Secondly,a Bayesian network inference algorithm of hybrid strategies and wolf pack is proposed,which uses the Tent chaotic mapping method to increase diversity of the initial population.An improved information transfer update strategy is proposed to update the population.Using the balance selection mechanism to balance the local and global search and prevent the algorithm from falling into the local optimum,and then the Bayesian network obtain the most probable explanation after iteration.Finally,the improved Bayesian network algorithm is used to diagnose the fault of the grate cooler.According to the operation mechanism of the grate cooler system,this paper extracts the important process parameters of system and quantifies data of the parameters.The structure of the grate cooler diagnosis model is constructed by the improved wolf pack structure algorithm.The parameters of the grate cooler diagnosis model is got by the maximum likelihood estimation method,and then using the hybrid strategies improved wolf pack inference algorithm to perform probabilistic reasoning on the faulty nodes of the grate cooler system network model to make fault diagnosis.