Research On Improved Bayesian Algorithm And Fault Diagnosis Of Rotary Kiln

Bayesian network is a probability graph model composed of structural algorithm,parameter algorithm,and inference algorithm.Bayesian network is a powerful tool in uncertain inference system.How to use bayesian algorithm to construct bayesian network quickly and accurately is the current research focus.To solve the problems of low search accuracy and easily falling into local optimum when using heuristic algorithm to improve bayesian algorithm,an improved structure learning algorithm is derived based on salp swarm algorithm.And an improved inference learning algorithm is derived based on particle swarm optimization.The working state of the cement rotary kiln will directly affect quality and energy consumption of the cement clinker.Therefore,bayesian algorithm is used to diagnose the rotary kiln.The structural algorithm is used to construct the structure of the rotary kiln network model.And the fault nodes in the model are diagnosed according to the inference algorithm.The specific research work is as follows:Firstly,the hybrid binary salp swarm-differential evolution algorithm is proposed based on salp swarm algorithm.The algorithm constructs the initial population based on the maximum support tree and the hill climbing algorithm.An adaptive scale factor is established to divide the population.The improved mutation operator and crossover operator are taken into salp search strategy and differential search strategy respectively to update different subswarms.At the stage of merging sub-populations,two-point mutation operator is used to increase population diversity.The convergence analysis of the proposed algorithm demonstrates that best structure can be obtained through the iterative search of population.Secondly,the simple immune particle swarm optimization is proposed based on particle swarm optimization.The algorithm uses immune strategy to extract excellent genes of the population.The simplified particle swarm optimization strategy is constructed to update the adaptive sub population.The improved marriage strategy is used to increase the communication between sub populations.The hybrid discrete particle swarm optimization-differential evolution is proposed based on simple immune particle swarm optimization.The algorithm used an adaptive reverse learning strategy to increase the diversity of the initial population.The differential mutation operator is introduced into the particle swarm optimization to establish an adaptive probability hierarchical search strategy to update the population.An adaptive mutation strategy is constructed to avoid the algorithm falling into local optimum.Finally,the improved bayesian algorithm is used to diagnose the fault of the rotary kiln.According to the process flow and expert knowledge of the rotary kiln,the fault characteristics of the rotary kiln are extracted as node variables.The structure of the rotary kiln network model is obtained by using the hybrid binary salp swarm-differential evolution,and the parameters of the rotary kiln network model is obtained by using the maximum likelihood method.In addition,the hybrid discrete particle swarm optimization-differential evolution algorithm is used to perform probabilistic inference learning on the fault nodes of the rotary kiln model.Then,the faults of the rotary kiln are diagnosed according to the inference results.