Data Modeling And Model Prediction Control Of Equal Diameter Phase Diameter Of CZ Single Crystal

The rapid rise of integrated circuits(ICs),digital analog system circuits,and solar photovoltaic industries has brought high demands on higher quality and larger size fo r silicon single crystals.In the actual production process,the main method for prepari ng high-quality silicon single crystals is the straight pull method based on visual calip er.The traditional double closed loop control system can control the growth process of single silicon crystal by heater power and pull speed.Therefore,this paper adopts the structure of constant tension speed control system based on visual diameter measurem ent,and controls the stable growth of crystal diameter through adjusting heating power to improve the quality of crystal.The process of heating power-crystal diameter has a series of characteristics such as nonlinear characteristics,time-varying characteristics,and large time lag characteristi cs.The effective control of crystal diameter is more effective than the model-based al-gorithm.The mechanism modeling of growth process is difficult and the solution is c omplex.In the aspect of model identification,the experimental data of heater power and c rystal diameter in equal diameter stage were selected for filtering processing,and the 1 ag order and model order were determined by fuzzy approximation method and deter minant ratio method.After the model structure is determined,support vector regression based on particle swarm optimization algorithm is used to identify model parameters.Si mulation.The simulation results show that the model of support vector regression base d on particle swarm optimization algorithm is more effective in suppressing the whole process of noise impact recognition results.Based on the experimental data,the identif ied lag order and model order are consistent with the empirical value.The support vect or regression model based on particle swarm optimization algorithm has been proved t o be very accurate.In terms of control,considering that in traditional model predictive control metho ds,the selection of prediction model cannot adequately describe the nonlinear process of the system,the dynamic linearized model of model-free adaptive control algorithm is selected as the prediction model.Combined with the idea of rolling optimization and feedback correction,a model-free adaptive predictive control method is proposed.Simu lation shows that the model-free adaptive predictive control algorithm can effectively c ontrol the nonlinear growth process,and the control speed and tracking accuracy are o bviously improved compared with other control methods.