基于贝叶斯算法的批间控制器设计.doc

基于贝叶斯算法的批间控制器设计

1.73万字
我自己原创的毕业论文，仅在本站独家提交，大家放心使用

摘要 随着社会经济的持续发展，半导体制造业也在迅猛增长。虽然国内外已有关于混合制程的相关研究成果，但是有关批间控制的研究，目前还是主要针对单一机台上的单一产品；然而在实际的生产中，由于生产需求的变化和技术的更新，必定会增加许多新产品，同时淘汰许多较旧产品，再考虑到半导体行业较其他行业更高的投资，对设备的高投入就导致其高利用率，因此对同一产品同一工序而言，两个相邻的批次在不同的机台上生产的事件完全有可能发生。这就是所谓的高度混合制程，这种高度混合的生产模式为批间控制的设计带来很大困难。传统的建模方法(如传递函数)和控制算法(如PID控制)等，都无法直接适用于这种混合产品生产的制程。批间控制(Run-to-Run Control，RtR Control)是将统计过程控制和工程过程控制相结合，并且在将其中的统计过程控制增添了反馈与前馈机制。批间控制越来越广泛地应用于半导体制程，也是半导体制造业中目前发展极为活跃的研究方向。
本论文在了解批间控制器基本原理及结构的基础上，从控制算法和方法两个方面对其进行分类。此外，比较了不同批间控制器并且说明了各自的优缺点。在对贝叶斯统计原理及半导体高度混合制程的生产模式深入了解的基础上，本论文提出了一种基于ANOVA模型和贝叶斯算法的批间控制器设计方案。该方案先使用先前观察结果递推计算出一种概率密度函数，然后就可以通过本论文中所推导的相关公式得出机台与产品的相对状态。该方案不需要求取观察矩阵的逆，也就大大减少了用求取广义逆矩阵方法估计系统状态的传统方案的巨大计算量。
本论文利用MATLAB仿真工具进行2机台3产品的仿真研究。此外，本论文使用IMA(1,1)模型反映实际生产的扰动。与基于线程的EWMA进行的对比中，本论文所提出的方案在输入阶跃扰动的情况下，只有一次大的波动并且很迅速地就返回到目标值，也就证明了本论文所设计方案的合理性和正确性。
关键字 贝叶斯统计理论；高混合制程；批间控制；方差分析


Design of Run to Run Controller using Bayesian Algorithm
Abstract With the rapid development of the social economy, the semiconductor manufacturing industry is also in a rapid growth. Although the research results in China and abroad about the mixing process exist, but the research is always for a single product single tool at present; however, in the actual productive process, due to the changes in demand and the update of technology,many new products will be introduced and many of the old ones to be eliminated.What’s more,Semiconductor trade and investments are higher than other industries, the high investments in equipment have led to its high utilization rate.As a result,it’s entirely possible for the same products in the same process produced in different tools.The process is called the high-mix manufacturing process, the mixed mode of production for the design of control between runs to brings great difficulties.Traditional modeling theory (such as state space representation, transfer function and so on) and the control algorithm (such as PID), the process cannot be directly applied to such processes.Run-to-Run Control combines the advantages of statistical process control and process control,and join feed-forward and feed-back mechanism to the statistical process control.As a result,it has become a hot topic in the semiconductor manufacturing industry in recent years .
In this paper,we learn the basic structure and principle of RtR control.Then,we classify if from two aspects of the control algorithm and control method.The basis of the classification is also introduced. At the same time,we expound the advantages and disadvantages of each category and make the overall comparison.This paper proposed a method based on ANOVA model and Bayesian algorithm . Based on the theory of RtR control, this paper also explain the Bayesian statistic theory, of which the Bayesian algorithm is the key point. Based on the mixing process, using the relevant knowledge of Bayesian statistics theory and ANOVA model, we focus on principles first and illustrate the general steps of using the algorithm.
In this paper,we use the MATLAB simulation.Based on existing theoretical results, we also take the disturbance of IMA(1,1) into consideration,in order to represent the real disturbance in the practical manufacturing process. The result of the waveform analysis of 2 tools, 3 products , compare to the thread EWMA, has proved the rationality and correctness of the proposed method in this paper.

Key words Bayesian statistical theory; high-mix manufacturing process; run to run control; ANOVA


目录
第一章 绪 论 1
1.1 研究背景与意义 1
1.2 国内外发展现状 2
1.2.1批间控制发展历程 2
1.2.2 批间控制研究现状 3
1.2.3 批间控制未来研究方向 4
1.3 存在问题 4
1.4 本文内容 6
第二章 批间控制算法理论 7
2.1 批间控制基本原理 7
2.1.1批间控制基本特性 7
2.1.2批间控制基本步骤 8
2.2 批间控制器基本控制结构 9
2.2.1批间控制器基本思想 9
2.2.2 批间控制器基本结构 9
2.3 EWMA批间控制器分类 10
2.3.1 从控制算法进行分类 10
2.3.2 从控制方法进行分类 15
第三章 基于贝叶斯算法的批间控制器设计 18
3.1 贝叶斯统计原理 18
3.1.1 贝叶斯算法起源 18
3.1.2 贝叶斯算法理论基础 18
3.2 问题提出及解..