Modeling And Optimization Of Chemical Fault Diagnosis Based On Machine Learning And Swarm Intelligent Optimization Algorithm

Chemical fault diagnosis always occupies a crucial position in the chemical industry.Efficient chemical fault diagnosis technology can effectively reduce the economic loss of enterprises and countries.In this paper,weighted least squares support vector machine(WLSSVM),relevance vector machine(RVM)and kernel based extreme learning machine(KELM)are used to establish models,and intelligent optimization algorithms are used to optimize the parameters of the models.In addition,intelligent optimization algorithms are combined and improved to effectively improve the diagnostic ability of models.The paper presents some kinds of chemical fault diagnosis models based on intelligence optimization algorithm and machine learning,and the modes are applied to the Tennessee Eastman(TE)process.The main work and contributions of this paper are as follows:1.Based to the Tennessee Eastman process,appropriate input variables and output variables are selected as the variables to establish the model,principal component analysis(PCA)are adopted to reduce dimension and extract the features.Then the variables are standardized,and evaluation index are determined to compare models.2.Based on the traditional WLSSVM model,the particle swarm optimization(PSO)algorithm is used to optimize the parameters of the WLSSVM model because the WLSSVM model is not easy to determine the parameters of the model,then the PSO-WLSSVM fault diagnosis model is proposed and is applied to the TE process,the experimental results show that the PSO=WLSSVM model has better fault diagnosis capability and proves the validity of the PSO-WLSSVM model.3.Based on the traditional RVM model,because of the disadvantage that intelligence optimization algorithm is always easily trapped in local optimal solution,PSO and differential evolution(DE)algorithm are combined together,the DEPSO optimization algorithm is proposed and applied to the RVM model.So the DEPSO-RVM fault diagnosis model is established and applied to the TE process.The experimental results show that the DEPSO-RVM model is more excellent.Besides,the validity of DEPSO algorithm and DEPSO-RVM model is proved.4、Based on the traditional KELM model and shuffled frog leaping algorithm(SFLA),because of the problems that SFLA is easily trapped in local optimal solution,an adaptive mutation strategy is introduced and the SFLA algorithm is improved.The MSFLA algorithm is proposed and combined with KELM model,so the MSFLA-KELM model is established.The experimental results show that the diagnostic capability of the MSFLA-KELM model is further improved compared to SFLA-KELM model,and the results also prove the validity of the MSFLA algorithm and MSFLA-KELM model.