Study On Temperature Control Of Single Crystal Furnace

Monocrystalline silicon has excellent semiconductor characteristics and is the raw material of electronic devices,photovoltaic power generation and other industries.The growth of monocrystalline silicon mainly includes straight-pull method and zone melting method,among which the monocrystalline silicon grown by straight-pull method is the most popular in the market.The growth environment of monocrystalline silicon is very complex,which requires very high temperature and precise control.The current development of monocrystalline silicon towards large diameter and high quality requires higher temperature control in the furnace.However,the single crystal furnace used for the growth of monocrystalline silicon has the characteristics of large lag,and can not adapt to the growth of large diameter monocrystalline silicon.Therefore,in order to ensure the stable growth of monocrystalline silicon,the requirements for temperature control in the furnace are higher.In this paper,the neural network PID control algorithm for single crystal furnace of Anhui Yixin Semiconductor Co.,Ltd.is studied.On the basis of the original PID control,the neural network control is introduced and combined with it.A temperature control system for single crystal furnace based on the neural network PID control is proposed.The main contents of this paper are as follows:In this paper,the function of the mechanical part of the single crystal furnace and the process of silicon growth are introduced.Studies and analyses the theory of neural network,and introduces several main neural networks,focusing on several elements of neural network: activation function and learning rules.The working principle of traditional PID is expounded,and its advantages and disadvantages are introduced.On this basis,a neural network PID control algorithm is introduced.By analyzing the tracking characteristics,stability,response speed and dynamic characteristics of the system with the software of matlab,it is determined that the neural network PID control algorithm is more suitable for single crystal furnace heating system with large lag characteristics,and has better performance than the traditional PID control algorithm.The realization of neural network PID algorithm is designed.The communication between Schneider PLC and Matlab is built by Kingview,and the neural network is calculated by using the software of matlab.Then Kingview configuration software is used as the bridge between the transmission of Matlab and PLC.The collected dataare exchanged and transmitted through the OPC protocol,and the design of the whole algorithm is realized.Finally,the mathematical model of single crystal furnace heating system is identified by using Matlab tool,and the neural network PID control algorithm is tested by changing the position and speed of crucible in single crystal furnace.Overall,the expected control effect has been achieved,which basically verifies the feasibility and correctness of the optimized temperature control system of single crystal furnace proposed in this paper.