【作者】 陈庭贵；

【导师】 肖人彬；

【作者基本信息】 华中科技大学 ， 机械设计及理论， 2009， 博士

【摘要】 经济的全球化,激烈的市场竞争促使企业在更短的时间内开发出高质量、低成本的新产品;同时也激发了人们对产品开发过程的重视,涌现出了多种产品开发过程优化、管理方法。其中美国学者Steward提出的设计结构矩阵方法具有一定的代表性,该方法力图通过矩阵的形式来描述和分析产品开发过程,减少过程的复杂性以及开发活动中出现的反复与迭代,使整个开发活动系统化、逻辑化和理性化。已有应用实践业已证明该方法是一个有效的产品开发过程优化与管理工具。本文借鉴国内外相关研究成果,以设计结构矩阵为研究工具,将人工免疫网络系统、反馈控制理论引入其中,探讨设计结构矩阵方法在复杂产品设计与开发领域应用的普适性。首先,研究了基于多输入多输出任务重叠设计方法。根据产品开发过程中串行任务间可能存在的重叠情况,考虑上游任务信息进展度与下游任务信息敏感度对任务执行方式的影响,提出基于任务返工迭代的重叠模型,并结合设计结构矩阵方法将其应用到普遍多输入多输出产品开发过程中。而且,对影响重叠效果的重要参数进行了分析、讨论,说明采用重叠执行方式存在一定的风险性。其次,研究了基于资源均衡策略的耦合集求解方法。考虑到并行迭代模型——工作转移矩阵未考虑具体资源约束下的资源分配问题,难以对产品开发过程进行有效的定量分析,建立了一种基于资源均衡策略的耦合集求解模型,旨在减少由于资源分配不均引起资源闲置的状况,而后结合遗传算法对模型求解,寻找较优的资源分配方案。再次,研究了动态环境中基于反馈控制理论的设计迭代分析方法。讨论了动态环境中产品开发过程的不确定因素以及其对开发过程产生的影响,通过引入资源监测控制矩阵对过程进行实时反馈调节,一方面使迭代过程在期望的时间内完成,另一方面实现对迭代过程的动态监测与控制作用。最后,研究了基于人工免疫网络算法的多项目调度问题求解方法。从目前研究较少的多项目多模态调度问题入手,通过对问题本质的分析,发现该问题具有明显的多峰特性,而人工免疫网络算法是一种局部搜索与全局搜索相结合的方法,对求解多峰优化问题具有显著的优势,从而提出基于人工免疫网络算法的多项目调度求解方法。该方法通过引入设计结构矩阵对任务间存在的时序约束关系进行简化,进而使用基于启发式规则的资源竞争优先级建立任务调度顺序,而后通过人工免疫算法中的克隆选择、阴性选择和网络压缩操作实现解的局部搜索和全局搜索相结合,从而保证了算法具有很强的搜索能力,避免了可能存在的组合爆炸问题。此外,基于上述理论方法研究成果,通过门座起重机设计与开发实例说明了其可行性和有效性。更多还原

【Abstract】 Enterprises must develop their new products with high quality and low prices within a shorter time due to economic globalization and fierce market competition. At the same time, more and more people take cognizance of the importance of product development process. Up to now, many optimization theories and methods of product development process have been proposed. Among these methods, design structure matrix (DSM) presented by American scholar Steward is a representative one. Design structure matrix offers matrix form to describe and analyze product development process in order to reduce the complexity of process and minimize the iterative trial-and-error process, and make the whole development process more systematic, logical and rational. In addition, design structure matrix method has been proven by a number of researchers as a useful tool in optimization and management of product development process.Using the relevant theoretical research achievement for references, the design structure matrix is deeply researched in this dissertation, in which artificial immune network system and feedback control theory are introduced. The applications of design structure matrix in complex product design and development are also discussed.Firstly, the method of task overlapping based on multiple input and output is studied. For overlapping instances among sequential tasks existing in concurrent product development process, the model to overlap product development with rework and iteration is presented based on both information evolution of upstream task and information sensitivity of downstream task. And then, it is expanded to apply to product development process with multiple input and output using design structure matrix. Furthermore, some important parameters such as information evolution and sensitivity in it are discussed and impacts on overlapping caused by them are also revealed. The result shows that there possibly exist potential risks using overlapping strategy.Secondly, the method for solving coupled task set based on resource leveling is investigated. The parallel iteration model based on work transformation matrix (WTM) has not yet been fully considering the specific resource constraints, and it is difficult for product development process to carry out effective quantitative analysis. In view of this point, a model for solving coupled task set based on resource leveling strategy is proposed in order to reduce the unequal allocation of resources caused by idle resources, thereby it can efficiently improve the resource utilization and shorten the iteration time of product development. Furthermore, the genetic algorithm (GA) is applied to the proposed model for searching for the optimal solution.Thirdly, the method of analysis of design iteration based on feedback control theory in a dynamic environment is discussed. The uncertain factors such as task durations, output branches of tasks and resource allocations, existing in product development, are analyzed and their impacts on development process are also discussed. A control and monitoring matrix is introduced to timely adjust process based on feedback control theory. In doing so, on the one hand, the iteration process is able to be completed within the desired lead time; on the other hand, the iteration process can be monitored and controlled.Finally, the method for solving multi-mode resource constrained multi-project scheduling problem based on artificial immune network algorithm is explored. Because there are fewer researches on multi-mode and multi-project scheduling problem, we take it as an object of study. Through analysis, we can easily find that this problem has an obvious multimodal characteristic. Artificial immune network algorithm specially designed for multimodal function optimization proposed by Castro and Timmis has been demonstrated to have powerful multimodal searching ability as well as good stabilization, which includes both local searching and global searching strategy. So, it is very suitable to solve multi-mode and multi-project scheduling problem. This proposed method simplifies information constraints among tasks based on design structure matrix. And then, task scheduling orders are set up based on priority of resource competition using heuristic rules. Operations including clonal selection, negative selection and network suppression are used to realize the local searching and global searching which will assure that the algorithm has a powerful searching ability and also avoid the possible combinatorial explosion.In addition, using the methods mentioned above, the design and development of portal crane is successfully accomplished. The case study indicates that the proposed methods are feasible and effective.更多还原