节点文献

产品设计资源受限项目调度规划与算法

Project Planning and Scheduling with Constrained Resource for Product Design

【作者】 陈旺

【导师】 滕弘飞;

【作者基本信息】 大连理工大学 , 机械设计及理论, 2010, 博士

【摘要】 本文以装甲车辆传动装置设计为例,研究产品设计资源受限项目调度规划方法和优化调度问题求解算法。美国航空航天局NASA在2009年8月的科学技术报告中指出,研究资源受限项目调度问题(RCPSP)对于带工期、人力和设备约束的军工产品研制具有重要意义。本文问题在理论上属于NP-hard问题,具有理论和应用价值。本文针对某科研所委托“传动装置设计项目调度系统”项目的工程实际需求,在国家自然科学基金资助下,以传动装置设计为例,提出一种产品设计流程调度规划方法,给出设计资源受限项目优化调度问题的两种求解算法,以最终实现工程实用为目的。本文主要工作如下:(1)提出一种基于工程语义网-设计结构矩阵(简称ESW-DSM)的设计流程规划方法。在产品结构树基础上,基于工程语义,构建产品结构、任务和设计团队(有限资源)工程语义网,进而基于DSM给出设计流程规划,最后根据设计团队有限资源,给出该受限项目调度规划方案。最后,以该传动装置的设计流程调度规划实例对本文方法进行了验证。(2)提出一种基于群智能的散射搜索算法(简称SISS),利用带偏好的最大最小蚂蚁算法为散射搜索法提供多样性解,避免早熟。同时引入向前向后改进法,对当前最优解进行局部搜索,使算法快速收敛,用于求解单模式资源受限项目优化调度问题。分别通过单模式RCPSP问题集PSPLIB和装甲车辆传动装置设计实例问题求解验证,表明SISS算法可以快速收敛到工程满意解,提高计算效率。(3)给出一种带分布估计的广义差异演化算法(简称GDE-EDA),基于广义差异演化算法(GDE),利用分布估计算法(EDA)分析问题解空间的全局信息和设计变量间的内在联系,对最优解的分布进行预测,指导全局搜索过程,可以更快的获得多样性更好的Pareto解集,用于求解多目标多模式设计资源受限项目优化调度问题。经标准多目标测试函数(考题)与实例数值仿真实验表明,本文算法与NSGA-Ⅱ(非支配排序遗传算法)和GDE3广义差异演化算法)相比,具有较好的收敛性和多样性。(4)在上述研究成果的基础上,基于工作流技术研制了传动装置设计项目调度原型系统(TDPS)。综上所述,本文以装甲车辆传动装置设计为应用背景,在理论上,给出了一种产品设计资源受限项目调度规划方法(ESW-DSM)和优化调度问题求解的两种算法(SISS和GDE-EDA),并研制了该传动装置设计项目调度原型系统(TDPS),可望推广应用于工程车辆或工程机械产品设计。

【Abstract】 Taking transmission design of armored vehicles as an example, this dissertation does research on Resource-Constrained Project Scheduling Problem (RCPSP) for product design and the solving algorithms. In the scientific and technical reports presented in August 2009, NASA points out that research of RCPSP problem is important for the military product development with band schedules, manpower and equipment constraints. This problem is a NP-hard problem in theory, which has theoretical and practical value.To satisfy the actual engineering requirement of "transmission design project scheduling system" project commissioned by a research institute, the dissertation explores a planning and scheduling method for product design, and proposes two solving algorithms for project scheduling problem with constrained resource. Our research is supported by the National Natural Science Foundation. Our contribution is as follows:(1) We present a planning method for design process based on Engineering Semantic Web-Design Structure Matrix (ESW-DSM). According to product structure tree, we utilize engineering semantic to construct engineering semantic web of product structure, task and design team with constrained resource. Then we give design process planning on the basis of DSM and we give the constrained project scheduling plan according to the constrained resource of the design team. Finally, the proposed method is verified by the design process planning of a transmission.(2) We propose a Swarm Intelligent based Scatter Search method (SISS). In the SISS, we make use of Biased MAX-MIN Ant System to provide diverse solutions for the scatter search method avoiding prematurity. We also introduce Forward-Backward Improvement (FBI) strategy to make the algorithm quickly converge by searching local optimal solution. The FBI strategy can used to solve Single-mode RCPSP optimization problem. Through verifying on Single-mode RCPSP problem set PSPLIB and the design problem of transmission of the armored vehicles, we find SISS algorithm can quickly converge to the project satisfactory solution and improve computation efficiency.(3) We propose a Generalized Differential Evolution algorithm with Estimation of Distribution Algorithm (GDE-EDA). Based on generalized evolutionary algorithm, GDE-EDA utilizes the estimation of distribution algorithm to analyze the global information of the problem solution space and the intrinsic link between the design variables, predicts the distribution of the optimal solution and guides the global search process. The proposed method can faster obtain Pareto solution set of better diversity to solve the multi-mode multi-objective RCPSP optimization problem. Comparing the experimental results of with NSGA-II and GDE3 on multi-objective functions (benchmark) and actual example, the proposed algorithm has better convergence and diversity.(4) According to the above research, we developed "transmission design project scheduling system (TDPS)" based on workflow technology.Our work takes the transmission design of armored vehicles as application background. We give a ESW-DSM based complex product design process planning methods, propose two algorithms (SISS and GDE-EDA) for the scheduling of the product design; and develop transmission design project scheduling system (TDPS). Our work is expected to be applied to the design of engineering vehicles and engineering machinery product.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络