QQ客服 官方微博 反馈留言

题名	光学设计智能化方法研究
作者	陈小刚
学位类别	博士
答辩日期	1993
授予单位	中国科学院上海光学精密机械研究所
导师	王之江
中文摘要	光学设计智能化研究的问题就是如何让计算机模拟人类专家的设计过程，自动完成光学系统的设计任务。在这个过程所面对的主要问题就是初始结构的生成问题和局部极小值问题。本文主要对这两个问题进行了深入研究，并以三片摄影物镜为例，给出了实验结果。本文完成的主要工作是建立了一个采用了人工智能技术的光学系统软件系统。本文首先完成了一个用阻尼最小二乘法进行偈差自动校正的透镜设计软件系统，它包括一个根据光学设计专家的多年经验所建立的一个初始结构自动生成系统。接着，本文从理论上对近年来重新挖掘出来的模拟退火法进行了较为系统地研究，对于得到客观实践证实的能获得最低能量状态的退火方案从数学上给予了论证，对仅用模拟退火法进行优化的缺陷进行了讨论。在此基础上，本文建立了一种用于解空间中搜寻全局极小的组合优化算法：首先采用 Monte Carlo 技术在解空间中确定一系列初始结构，接着用一个小型专家系统对其进行修改和筛选，然后结合模拟退火法和阻尼最小二乘法对它们进行优化，最后确定出全局极小。为此，本文还对人工智能技术进行了研究，建立了一个简单的规则基专家系统。这个产生式系统具有推理机，知识库，综合数据处理，知识库管理和解释器等五个子系统。该系统采用了置信度传播技术，可进行不精确推理。系统采用了反向链接规则解释技术，同时，系统还具有简单的学习功能。采用该组合优化算法获得了满意的结果，使得在微机上结构简单的光学系统进行全局优化得到了实现。本文还对软件工程，编程技术进行了较多的研究。包括窗口技术，菜单技术，三维图形显示技术，三维动画技术，键盘控制和打印机控制技术，以及程序内存驻留技术等。
英文摘要	The purpose of using Artificial Intelligence (AI) in optical system design is attempt to make computer to imitate the design process of human experts and finish the design automatically. In this process, the main problems to be solved are initial configuration problem and local minimum problem. In this paper, the two problems are deeply studied and an optical system design software, in which the AI technology was adopted, is finished. In this thesis Firstly, an optical system using software using Damped Least Squares (DLS) optimization technique and a system to automatically build initial configurations from the experiences of human experts are finished. Secondly, the Generalized Simulated Annealing (GSA) optimization technique which was excavated in recent years is theoretically and systematically studied, the annealing schedule which is practically effective and by which the lowest energy state may be attained is mathematically demonstrated, and the defect of using only GSA in optimization is discussed. On this basis, a combined optimization technique used to search for the global minimum in the object space is established. It takes the following optimization strategies. First, a group of initial configurations in the object space of thin lens model are constructed by using Monte Carlo random selection. Second, they are modified and sieved with a small expert system. Then, the optimization with the conjunction of DLS and GSA is performed. As the last, the configuration with lowest merit function is picked out as global minimum. Hence, a rule based small expert system used in optical system design is discussed. This production system consists of five major components: an inference engine, a knowledge base, a global data base, a knowledge base manager and an interpreter. In this system, certificate spread technology is used for inaccurate inference and backward-chaining rule interpreter is used. The system has primary ability of learning. Using the combined optimization technique, satisfactory results have been gained and the global optimization for simple optical system to PC computer is realized. In this thesis, software engineering and programming technology are also studied, which include the technologies of window, menu, 3-D graphics, 3-D carton, keyboard control and printer control, as well as memory resident program.
语种	中文
内容类型	学位论文
源URL	[http://ir.siom.ac.cn/handle/181231/15538]
专题	上海光学精密机械研究所_学位论文
推荐引用方式 GB/T 7714	陈小刚. 光学设计智能化方法研究[D]. 中国科学院上海光学精密机械研究所. 1993.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。