【作者】 段鹰；

【导师】 易树平；

【作者基本信息】 重庆大学 ， 机械工程， 2009， 博士

【摘要】 当今世界政治经济环境多变,科学技术高速发展,企业必须适应变化,科学并正确地作出决策,才能生存发展。同时,决策活动也是贯穿管理过程始终的一项基本活动,企业各项管理职能中都存在着如何科学合理决策的问题。论文以系统方法论作为指导,研究复杂生产系统决策中的若干关键技术,涉及面广,包括复杂问题诊断、战术层优化决策、战略层群体决策、人机智能系统的技术支撑等多个层面。生产经营系统是个复杂大系统,其决策问题的解决涉及多个学科的交叉和融合。论文概述系统方法论与系统科学体系,综述了复杂系统诊断与决策、综合集成方法等研究现状,讨论了生产经营系统的决策流程,明确了论文要解决的问题。针对复杂生产系统的决策难点,本文提出的核心思路是:从系统方法与相关知识工程两个角度,发挥群体智慧,定性定量综合集成,并提出了相应决策的总体思路,构建了决策过程的框架模型。问题识别与诊断是做好决策的基础。论文在对生产系统“问题”进行界定和分类的基础上,建立了问题识别层次模型。总结了面对不同“问题结构”和不同利益群体的“问题”挖掘方法。分别对半结构的和无结构的“问题诊断技术”进行讨论,尤其针对“软系统方法论模型”所存在的不足,提出了方法上的修正,给出了分析框架,设计了分析步骤与方法。对两类问题的两个相应案例进行处理,收到良好的效果。随着人们认知水平的提高,对一些具有不确定性、无结构的决策问题可以由定性决策转为定量决策。在生产经营系统的战术决策层面,解决了因时滞性因素的存在,难以构建定量化模型以进行优化决策的问题,完成了以下工作:①针对生产系统的时滞性因素,建立了一个基于大系统“关联预测法”的递阶模型,分散了系统的复杂性;②针对时滞性生产系统,提出了基于“条件最优化原理”的动态规划方法,用微分动态规划方法解决了动态规划的“维数灾”,并证明了算法的收敛性;③论文用上述方法解决某工厂多过程的最优协调和生产调度问题,仿真结果收到良好的节能效果,证明了本文提出方法的有效性。对于战略性决策的复杂问题,往往需要多人共同参与决策。论文进一步研究了复杂生产系统群体决策问题。从分析群体决策特点入手,给出了群体决策的过程模型;研究了群体决策有效性的影响因素,从决策个体的决策行为、角色、群体研讨规则等角度提出了相应对策;研究了群体决策中的知识交互与协同策略,讨论了决策中的克星循环现象,通过工程实践,总结了消除克星循环的一些方法。最后,论文研究了群体决策沟通的技术环境与手段的影响,通过设计试验与统计分析,得出了提高效率的有效途径,给出了相应的数据。考虑到生产经营系统在进行决策的各个阶段,都必须得到知识的支撑,问题诊断和战略决策中往往还离不开人―机智能系统,论文最后研究了工程领域中的自然语言理解与诊断推理技术。针对这两个问题:①研究提出了一种相关对象组合匹配的分层表达模型,在词形的特征标注中采用常量、特征常量与变量相结合的方法,便于在匹配搜索时分层地加以处理,兼顾了规则的准确性和覆盖面;②基于此模型,提出了切分组合与规则调用的方法、应用翻译规则获取译文流畅性的方案;③对规则匹配时的搜索,提出一种采用闭环消除法解耦的递阶智能搜索方法,证明了方法的严密性,避免了搜索中大量的回溯、内存占用和“组合爆炸”问题。最后对本文提出的理论和方法进行了综合应用研究。对一家工业集团公司进行技术系统的战略规划,设计了全部战略决策过程。整个过程通过合理的系统分解,采用各种群体研讨方式,综合使用定性定量方法,有效地解决了跨行业知识传递、整合的问题,达成了共识,给出了各类产品的发展前景,给企业指明了发展方向。更多还原

【Abstract】 Today, in the world’s ever-changing political and economic environment, with the rapid development of science and technology, enterprises must adapt to change themselves and make scientific and right decision in order to survive and develop. The decision-making activities has always been a basic activity throughout the management process. The problem of how to make rational decision exists in all of the enterprise various management functions. This paper takes systematic methodology as a guide to study some key technologies in complex production system decision-making. The research involves a wide range of diagnosis of complex problem, tactical optimization decision-making, group decision-making on strategic level, technical support for man-machine intelligent system and so on.Production-operation system is a large complex system that involves the intersection and integration of multiple disciplines. Therefore this paper summarizes the system methodology and the system of systematic science, and specially the status quo of diagnosing & decision-making, comprehensive integration approach for a complex system. On this basis, it gives the decision-making process of production-operation system, and highlights the main issues that should be studied. The core idea of this paper is that playing collective wisdom, exerting integrated qualitative and quantitative factors from the two angles of system approach and related knowledge work. It proposes general ideas of corresponding decision-making and builds a framework for decision-making process.Problem identification and diagnosis is the basis of good decision-making. On the basis of defination and classification of the‘problem’in production system, it establishes hierarchical model of problem identification. It sums up the method of excavating the‘problem’when facing different "problem structure" and the different interest groups. The paper discusses the semi-structured and unstructured‘problem diagnostic techniques’respectively. In particular, for the shortcomings of "Soft Systems Methodology Model", it proposes methods amendments, gives the analytical model and designs analysis steps and methods. Handling two corresponding types of cases to the two types of problems has received good results.As the level of cognitive rising, some uncertain type, non-structural decision-making problems can be changed from qualitative to quantitative decision-making. On the tactical decision-making level of production and management system, it solves the difficult of build quantitative models to optimize the decision-making due to the presence of time-delay factor. It creatively completed the following work:①for time-delay factor of the production system, it establishes a hierarchical model based on‘related prediction’of large scale system and disperses complexity of the system.②To solve problems of time-delay production system, it proposes dynamic programming based on“principle of optimality conditions”. It uses differential dynamic programming to solve the "curse of dimensionality" of dynamic programming and proves convergence of the algorithm.③The paper uses it to solve the problems of optimal coordination and production scheduling among multi-process in a factory. The simulation shows that a great energy-saving effect is achieved, and the validity of the method is proved.Complex strategic decision-making problems often require a few persons participating in decisions-making together. The paper studies group decision-making problems of complex production systems; and gives group decision-making process models based on the analysis of characteristics of group decision-making. Then it studies the factors that influence effectiveness of group decision-making, and puts forward corresponding countermeasures from the viewpoint of decision-maker’s behavior, role, and rules of group decision-making. It also studies knowledge interaction and collaboration strategies during group decision-making; and discusses killer-loop exists in decision-making. Through the engineering practice, it summarizes some of the ways to eliminate killer-loop. Finally, the paper studies impact of group decision-making communication’s technical environment and means. Experiment and statistical analysis results in effective ways to improve efficiency, which gives corresponding data.Considering every phase of decision-making within the production-operation system, knowledge is a must-be brace. Problem diagnosis and strategic decision could not be made without man-machine system. Thus, this paper studies the natural language understanding and diagnostic reasoning techniques in the field of engineering. Aims to those two problems:①this study proposes a layered expressed model of related objects with combinatorial matches. The constant, the characteristic constant and the variable was unified within the morphology characteristic labeling, in order to handle hierarchically when match searching which considering the rules of accuracy and the coverage;②based on this model, the paper proposes the method of segmentation- combination and rule scheduling, and the plan of access to affluent translation by translation rules;③when the rule’s match searching, it proposes a hierarchical intelligence researching method using a closed loop to eliminate the law decoupling which has proven to be rigorous. This method has avoided the massive recollections and the memory occupation during the search, as well as the problem of“combination explosion”.Finally, it makes comprehensive applied research on the theories and methods in this paper. It makes a technology systems strategic planning for an industry group and designs the whole process of the strategic decision-making. Through reasonable system disassembling, variety of group discussion ways, qualitative and quantitative methods, the paper effectively solves the cross-industry knowledge transfer, integration problems and reached agreement. It gives the prospects of the development of various products and points out the direction of development.更多还原