Abstract: Temperature control of the silicon rod surface in the reduction furnace of polycrystalline silicon requires that the surface temperature of the silicon rod has to be maintained at the optimal reaction temperature and the fluctuation should not be violent. Due to the change of the electrical structure of the silicon rod as a result of the vapor deposition reaction in the production of polycrystalline silicon, there exists the characters of interconnection and nonlinear. This paper adopts the double closed-loop control system with model-free adaptive fuzzy controller as the major controller and PI as the auxiliary controller. It uses the structure of model-free adaptive controller and carries out on-line revision of its parameters by means of fuzzy rules, which not only ensures the convergence and stability of the temperature control system, but also enhances the rapid dynamic performance of the system toward temperature control. The simulations show the availability of this control method.

Key words: model-free adaptive, fuzzy control system, parameter self-tuning, silicon rod temperature

摘要： 多晶硅还原炉硅棒表面温度控制既要保证硅棒表面在最佳反应温度，又不要使其波动较大。由于多晶硅生产过程中的气相沉积反应使硅棒电气结构发生变化，因而存在很强的耦合作用和非线性特性。提出了以无模型自适应模糊控制器为主控制器，以PI为副控制器的双闭环结构的控制系统，采用基于无模型自适应控制器的结构，根据无模型自适应控制器的参数学习因子和步长序列对系统动态性能影响趋势，使用模糊自适应方法，通过模糊规则在线修正无模型自适应控制器的学习因子和步长序列两个参数，使温度控制系统既能保证收敛稳定性，又能提高系统对于温度控制的动态性能。仿真结果表明了该控制方法的有效性。

关键词: 无模型自适应, 模糊控制, 参数自整定, 硅棒温度