Research On Thermal Field Temperature Modeling And Control Method Of CZ Silicon Single Crystal In Constant-Diameter Growth Stage

The development of large-scale integrated circuit and solar photovoltaic industry puts high requirements for quality and size of silicon single crystal.Czochralski is the effective method for high quality growth.Traditional control system structure based on the visual measurement for crystal diameter achieves the CZ silicon single crystal growth process control by heater and pulling rate.However,coupling complexity between key variables is not easy to realize decoupling control.Basing on constant pulling rate with visual measurement,how to adjust heating power to control thermal field temperature is the focus of this paper.The thermal transfer process of heating power-thermal field has the characteristics of nonlinear,time-varying and large lag.The model based algorithm is more effective for controlling of thermal field temperature.The mechanism modeling of growth process is difficult and complex to solve.This paper adopt data-driven modeling ideas.It is difficult to get accurate model by once identification for long time growth process.In consideration,this paper proposes the strategy of segments identification and segments control.In model identification,the experimental data of sufficient excitation are selected to be segmented and filtered.Fuzzy approximation and determinant ratio method are used to identify the lag order and model order.After determining model structure,for simple segment one,once improved ALO algorithm is used to identify the model parameters;for complex segment two,a new T-S fuzzy identification method combined with twice improved ALO algorithm are used.Simulation shows that the improved ALO algorithm have better global optimization capability;the new T-S fuzzy method is inhibitory to the influence of noise on the identification result than the traditional method.For the identification of Silicon Single Crystal data,both the lag order and the model order identified are consistent with the practice;Based on improved ALO algorithm and new T-S fuzzy identification method,the segments model with high precision is obtained.In controlling,considering the instability of noise parameter and the complexity of the generalized predictive control,improved generalized predictive control algorithm is selected to realize segments model control,meanwhile,the model 2 parameters are modified online.In view of the uninstability caused by the model switching process,the strategy that calculating the control variable of being switched model in advance and limit for controlled variable are proposed.Simulation shows that improved generalized predictive control algorithm can effectively control the segments process,and the control strategy of model switching ensures a smooth transition of the switching process.