基于改进BP神经网络算法的激光晶体生长控制研究  

Research on Laser Crystal Growth Control Based on Improved BP Neural Network Algorithm

作　　者：李建鸿 纪文刚[1] 宋星 储承贵 LI Jian-hong;JI Wen-gang;SONG Xing;CHU Cheng-gui(College of Information Engineering,Beijing Institute of Petrochemical Technology,Beijing 102617,China;College of Information Science and Technology,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]北京石油化工学院信息工程学院,北京102617 [2]北京化工大学信息科学与技术学院,北京100029

出　　处：《人工晶体学报》2019年第8期1438-1444,共7页Journal of Synthetic Crystals

摘　　要：针对激光晶体生长后期晶体生长炉温度上漂导致晶体不能维持等径生长的问题,采用一种基于自适应学习率优化算法改进BP神经网络与经典PID控制技术相结合的方法应用到晶体生长控制过程中,通过改进的BP神经网络的自学习以及调整加权系数,实现一种由BP神经网络整定的最佳PID控制。通过MATLAB/Simulink仿真对比表明,改进的BP神经网络PID控制算法较传统PID控制方法具有较好的控制性能和鲁棒性,能够有效克服晶体生长炉的温漂问题,更好地保持晶体等径生长,提高了控制精度。In order to solve the problem of the crystal growth furnace temperature drifts in the late stage of laser crystal growth which cannot maintain equal diameter growth, a method based on adaptive learning rate optimization algorithm to improve BP neural network and traditional PID control technology is applied to the crystal growth control process. Through the self-learning of the improved BP neural network and the adjustment of the weighting coefficients, an optimal PID control is set by the BP neural network. The simulation comparison of MATLAB/Simulink shows that the improved BP neural network PID control algorithm has better control performance and robustness than the traditional PID control method, which can effectively overcome the temperature drift problem of the crystal growth furnace, better maintain the equal diameter growth of the crystal and improve the control precision.

关 键 词：激光晶体生长 自适应学习率 BP神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]