【作者】 梁新元；

【导师】 张勤；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2005， 博士

【摘要】 复杂系统是由大量相互作用的单元构成的系统。复杂性科学这一门被科学家誉为“21世纪的科学”,正逐渐为人们所理解和接受,并且有一批国际和国内著名学者加入到这一领域的研究中。1999年4月美国刊物Science出版了一个题目为“复杂系统”的专集,对复杂系统进行了专门研究,内容涉及物理、化学、生物、经济、生态、地理环境、气象、神经科学等方面。国内以钱学森教授为首的一批科学家致力于复杂系统的研究。复杂系统包括的子系统很多,它具有以下性质:整体性、系统性、开放性、复杂性、层次性、动态性、不确定性。复杂系统的故障诊断是人工智能技术应用于实际问题的重要领域之一。现在的系统非常复杂,使故障诊断问题复杂化。因此,研究如何能够切实保障现代工业复杂系统可靠、安全地运行,显得十分重要。复杂系统故障诊断具有层次性、时间性、延时性、相关性、模糊性、随机性、不确定性和放射性等特性。随着人工智能技术的不断发展,智能诊断技术开始在故障诊断领域发挥着越来越重要的作用。不确定性知识的表示和推理是人工智能研究的核心问题之一,非常重要,同时也是十分困难的,由于它很有现实意义,一直是国际上研究的热点。动态因果图是由张勤教授于1994年提出的不确定推理方法。因果图具有正向和反向及混合推理能力,由专家建立因果图,可以对复杂系统进行故障诊断。本文探讨动态因果图理论及因果图在复杂系统中的应用,对复杂系统因果图推理理论及算法进行研究,主要研究内容有:模糊因果图、因果图的动态推理问题、复杂系统因果图推理算法和因果图在复杂系统中的应用。针对复杂系统的模糊性、时间动态性、相关性等分别提出了模糊因果图理论和推理算法,带时间延迟的因果图迭代推理算法,复杂因果图正则分解和推理算法。针对复杂系统建模困难,提出了复杂因果图的合成方法。针对复杂系统因果图推理时间复杂度高的情况,提出了几种推理算法。另外,复杂系统的故障诊断是非常重要的问题,本文提出了一些新的故障诊断方法。最后还对故障树和因果图进行了全面比较,提出了它们之间相互转换的方法。本文针对复杂系统的模糊性,将模糊数学理论引入因果图,提出了模糊因果图理论和推理算法。论文重点研究了离散的单值模糊因果图和多值模糊因果图。提出了单值模糊因果图推理的两种算法。对事件概率取三角模糊数、梯形模糊数、正态模糊数和区间数的情况下,分别对单值模糊因果图推理进行了研究,并提出了各种情况下模糊概率的归一化方法。针对事件概率可取多种模糊数情况,提出了模糊因果图混合推理方法。提出了基于因果影响可能性分配的多值模糊因果图更多还原

【Abstract】 Complex System (CS) is a system composed of mass interacting units. The Complexity science which is honored as a 21st century science by scientists, is gradually understood and accepted by people, and a group of scholars both at home and abroad has conducted researches in the field. In April, 1994, the SCIENCE—an America journal has published a special issue entitled Complexity System, which focused on the complex system and its content covered physics, chemistry, biology, economics, ecology, geography, meteorology, neuroscience, etc. At home, a group of scientists, led by professor Qian Xue-sen, apply themselves to study Complex System. Complex System comprises many subsystems. It has the following characteristics: integrity, systematization, opening, complexity, hierarchy, dynamic, and uncertainty.Fault diagnosis of Complex System is an important application of artificial intelligence (AI) to practical problem. Nowadays the system is so complex, which complicates fault diagnoses. It becomes very important to study how to ensure that the modern industrial complex system runs safely and reliably. Fault diagnosis of complex system has such characteristics as hierarchy, timeliness, delay, relativity, fuzziness, randomicity, uncertainty, radioactivity, etc.With the persistent development of AI technology, Intelligent Diagnosis System (IDS) has played a more and more important role in the field of fault diagnosis. The representation and reasoning methodology dealing with uncertainty knowledge, which is an important, core and difficult problem of AI, has always been an international research hotspot because of its realistic significance. Dynamic Causality Diagram (DCD) is an uncertainty reasoning method proposed by professor Zhang Qin in 1994. The Causality Diagram (CD) which is constituted by experts, thanks to its reasoning capability (forward reasoning, backward reasoning and bi-directional reasoning), can thus diagnose faults in complex system.This paper discusses the Dynamic Causality Diagram and its application in complex system, focuses on the reasoning theory and algorithm of causality diagram of complex system (CDCS). Its content mainly involves fuzzy causality diagram (FCD), dynamic reasoning of CD, reasoning algorithm of CDCS, and the application CD in complex system. Based on the fuzziness, timeliness, dynamic, and relativity, this paper respectively puts forward the theory and algorithm of FCD, iterative reasoning更多还原