面向分子束外延系统的BP神经网络智能温度控制的研究

【作者】 吴国清；

【导师】 汪洋；

【作者基本信息】 上海师范大学 ， 工程硕士（专业学位）， 2022， 硕士

【摘要】 为了提高分子束外延(Molecular beam epitaxy,MBE)系统在超高真空环境中生长半导体薄膜材料的效果,本文对分子束外延系统中的生长温度控制进行研究。MBE技术是制备高精度探测器、智能芯片、半导体激光器等高性能器件所需新型材料的关键技术之一,而温度又是影响MBE外延层生长质量的关键因素。为了生长优质的外延材料,MBE系统中的温度控制起到尤为重要的作用。常见半导体薄膜材料需要的生长温度在200℃至 1200℃之间,但传统的温度控制算法会在温度切换时出现一些问题,如超调、延迟、误差等。以PID算法为代表的传统温度控制算法越来越难以满足当下MBE系统材料生长的需求。因此,本文通过对MBE系统的研究分析,以人工智能算法结合其温度控制算法的方法以满足更高的实际要求。本论文针对MBE系统的生长温度控制特点,提出了一种BP神经网络结合PID控制算法的MBE系统温度控制算法,简称基于BP-PID的MBE系统。BPPID算法在复杂的非线系统中有较好的表现,与MBE系统契合,可以解决其中PID控制存在的温度精度不足、响应速度慢等问题。本文首先介绍了MBE系统的基本组成结构和工作原理,对系统的加热过程、生长机理以及温度控制做进一步研究分析。再而对PID算法和BP神经网络算法进行介绍,并重点阐述了BP神经网络算法的实现过程,最后分析了两者相结合的BP-PID算法的工作原理。最后,在软件和硬件方面对基于BP-PID算法的MBE系统的进行论证。在软件方面,首先利用飞升曲线法计算出MBE系统的数学模型;其次搭建MBE系统的BP-PID模型,其中包括确定BP神经网络的各层结构和MBE系统的数据收集;再而使用KNN算法对预处理后的数据进行分类,可以更有效地训练和测试BP-PID模型,最后对测试结果进行分析。在硬件方面,由STM32F1控制器、MAX6675热电偶测温电路、计算机以及MBE系统加热装置中的欧陆温度表和电源系统可搭建硬件实验平台,验证了BP-PID算法的超调小、响应快等优点。通过BP-PID和PID算法的对比分析,得出BP-PID算法在MBE系统中具有更强的鲁棒性、更精确的温度控制、更短的响应时间、更快的响应速度。因此,BP-PID算法更适用于MBE系统,使半导体薄膜可以在最佳温度进行生长,保证了MBE生长材料的质量。更多还原

【Abstract】 In order to improve the effect of molecular beam epitaxy(MBE)system for growing semiconductor thin film materials in ultra-high vacuum environment,this paper studies the growth temperature control in molecular beam epitaxy system.MBE technology is one of the key technologies for the preparation of new materials for highprecision detectors,smart chips,semiconductor lasers and other high-performance devices,and temperature is a key factor affecting the growth quality of MBE epitaxial layers.In order to grow high-quality epitaxial materials,temperature control in the MBE system plays a particularly important role.Common semiconductor thin film materials require a growth temperature between 200℃ and 1200℃,but traditional temperature control algorithms will have some problems during temperature switching,such as overshoot,delay,and error.The traditional temperature control algorithm represented by the PID algorithm is more and more difficult to meet the requirements of the material growth of the current MBE system.Therefore,through the research and analysis of the MBE system,this paper combines the artificial intelligence algorithm with its temperature control algorithm to meet higher practical requirements.Aiming at the growth temperature control characteristics of MBE system,this paper proposes a temperature control algorithm of MBE system based on BP neural network combined with PID control algorithm,which is referred to as MBE system based on BP-PID.The BP-PID algorithm has a good performance in complex nonlinear systems,and it fits with the MBE system,which can solve the problems of insufficient temperature accuracy and slow response speed in PID control.This paper firstly introduces the basic structure and working principle of the MBE system,and further studies and analyzes the heating process,growth mechanism and temperature control of the system.Then the PID algorithm and the BP neural network algorithm are introduced,and the realization process of the BP neural network algorithm is emphasized.Finally,the working principle of the BP-PID algorithm combined with the two is analyzed.Finally,the MBE system based on BP-PID algorithm is demonstrated in terms of software and hardware.In terms of software,firstly,the mathematical model of the MBE system is calculated by the soaring curve method;secondly,the BP-PID model of the MBE system is built,which includes determining the structure of each layer of the BP neural network and the data collection of the MBE system.The preprocessed data is classified,which can train and test the BP-PID model more effectively,and finally analyze the test results.In terms of hardware,a hardware experiment platform can be built by the STM32F1 controller,MAX6675 thermocouple temperature measurement circuit,computer,and the Eurotherm thermometer and power supply system in the heating device of the MBE system,which verifies the small overshoot and fast response of the BP-PID algorithm.advantage.Through the comparative analysis of BP-PID and PID algorithm,it is concluded that BP-PID algorithm has stronger robustness,more accurate temperature control,shorter response time and faster response speed in MBE system.Therefore,the BP-PID algorithm is more suitable for the MBE system,so that the semiconductor thin film can be grown at the optimal temperature,which ensures the quality of the MBE grown material.更多还原