Predictive Control Method Of Grinding-classification Process For Model Mismatch Problem

The grinding-classification process is the key link of beneficiation operation.The output and fineness of the product directly affect the economic and technical indexes of the concentrator.However,due to the problems of variable ore properties,multi-variable coupling and strong interference in the grinding and classification process,model-based optimization control for the grinding and classification process cannot run effectively for a long time because of model mismatch.Therefore,this paper takes the one-stage grinding and classification process of a concentrator in Hunan province as the research object to study the predictive control method with model mismatch problem.The main research work and innovation of this paper are as follows:(1)The mechanism model of each module in the grinding and classification process was studied according to the principle of mass balance,and the nonlinear state space model was constructed by connecting each module in sequence according to the actual process flowsheet.The corresponding parameters of the model were identified by using the industrial data.The accuracy of the model was then verified by the actual production data.(2)An adaptive DMC control method for fuzzy grinding classification process was proposed considering of the simplicity of DMC algorithm with small calculation load.Firstly,based on the mechanism model of the grinding classification process,a four-input and four-output transfer function model was built as the prediction model.Considering that when using DMC algorithm to solve the model mismatch problem,it is difficult to achieve both robustness and quick response to interfere when selecting the calibration parameters,an adaptive fuzzy DMC control method was put forward.Simulation results show that control effect obtained by using the adaptive DMC algorithm,is better than that of the DMC without adaptive function.(3)Considering the inherent nonlinear,constrained and unmeasured state variables of the grinding and classification process,a grinding and classification process control strategy based on NMPC-PI was proposed.The NMPC controller and PI controller with constraints were designed to control the water filling to the pump pool in the grinding process by switch between the NMPC and PI.A particle filter state observer was introduced to estimate the system state and disturbance to solve the model mismatch problem in the NMPC-PI control process.And a firefly algorithm was used to improve the particle diversity reduction problem in the particle filter resampling process.Compared with the unimproved particle filter,the simulation results show that the improved particle filter has higher estimation accuracy.In addition,step response experiment results t show the effectiveness of the proposed control method.