【作者】 熊凯；

【导师】 熊体凡；

【作者基本信息】 华中科技大学 ， 机械工程， 2019， 硕士

【摘要】 随着市场经济快速发展以及经济全球化趋势愈演愈烈,企业所面临的市场竞争压力越来越大。激烈的市场竞争促使企业更加重视产品开发过程,企业希望通过项目管理的技术,对项目任务和资源进行合理的调度与分配,力求用最短的时间和最低的成本开发出更高质量的新产品。在产品开发项目调度中,由于任务数量的增加和多技能开发员工的存在,导致任务调度和人员配置问题中的可行方案急剧加大,因此,如何在扩大的可行域中进行时间、成本等多个目标方案的优选,成为企业项目管理者面临的一个决策难题。本文以软件开发行业为背景,在借鉴国内外研究成果基础上,针对多技能人员约束下的多目标产品开发项目调度问题进行了数学建模,并对其智能求解算法进行了研究,然后采用标准数值案例对所提算法进行了计算仿真实验,验证了算法的有效性,最后利用软件产品开发实践案例验证了所研究理论和方法的实用性。本文研究成果主要包括以下几方面:首先,本文对多技能人员约束下的产品开发项目调度问题进行了分析,并将其表示为一类多技能资源受限项目调度问题,然后详细介绍了本文研究的多技能资源受限项目调度问题的约束条件和优化目标,最后针对多目标多技能资源受限项目调度问题进行了形式化描述,建立了数学模型。其次,本文围绕多目标的多技能资源受限项目调度问题的求解方法进行了深入研究,并提出了两种新的智能优化算法:1)借鉴传统多目标果蝇优化算法,针对关键算子进行了改进,并引入模糊支配概念,设计了基于模糊支配排序的果蝇优化算法;2)为了验证智能优化算法结合强化学习机制在求解复杂问题时是否能够表现出良好的性能,本文在基于模糊支配排序果蝇优化算法基础上,通过采用Sarsa学习机制来控制算法的进化方式,从而提高算法的搜索效率,最后提出了基于Sarsa-学习的模糊支配排序果蝇优化算法对所研究问题进行求解。然后,本文采用已有研究中的两种具有代表性的算法作为对比对象,以验证所提算法的有效性。随后使用所有算法对标准数值案例进行了求解,最后对实验结果进行了统计分析。结果表明,与其他算法相比,本文所提出算法在超体积、平均理想距离以及CPU计算时间等性能度量指标方面都有不错的表现。最后,在以上理论和研究方法上,本文通过某基于AR的机车产品螺纹装配管理平台的产品开发项目实例说明了本文所提算法的实用性。更多还原

【Abstract】 With the rapid development of the market economy and the growing trend of economic globalization,the market competition pressure faced by enterprises is increasing.The fierce market competition has prompted enterprises to pay more attention to the product development process.By using project management technology,the company will rationally dispatch and allocate project tasks and resources,and strive to develop new products with higher quality in the shortest time and lowest cost.In the product development project scheduling,due to the increase in the number of tasks and the existence of multi-skilled development staff,the feasible solutions in task scheduling and staffing problems have increased dramatically.Therefore,how to optimize the schemes of multiple goals such as time and cost in the feasible feasible domain becomes a decision-making problem faced by enterprise project managers.In this paper,based on the background of software development industry and the research results at home and abroad,the multi-objective product development project scheduling problem with multi-skilled personnel constraints is modeled,and the intelligent optimization algorithms for solving this problem is studied.Then the standard numerical examples are used to simulate the simulation of the proposed algorithms,and the effectiveness of the algorithms is verified.Finally,the practicality of the research theory and methods is verified by the software product development practice case.The research results of this paper mainly include the following aspects:Firstly,this paper analyzes the problem of product development project scheduling under the constraints of multi-skilled personnel,and represents it as a multi-skill resource-constrained project scheduling problem.Then the constraints and optimization objectives of the multi-skill resource-constrained project scheduling problem studied in this paper are introduced in detail.Finally,the multi-objective multi-skill resource-constrained project scheduling problem is formulated,and the mathematical model is established.Secondly,this paper studies the solution approaches for solving the multi-objective and multi-skill resource-constrained project scheduling problem,and proposes two new intelligent optimization algorithms: 1)Based on the traditional multi-objective fruit fly optimization algorithm,the key operator is improved and the concept of fuzzy dominance is introduced.Then,a fruit fly optimization algorithm based on fuzzy dominance sorting is designed.2)In order to verify whether the intelligent optimization algorithm combined with the reinforcement learning mechanism can show good performance when solving complex problems.Based on the fuzzy dominance fruit fly optimization algorithm,this paper uses Sarsa-learning mechanism to control the evolution of the algorithm to improve the search efficiency of the algorithm.Then,a fuzzy-dominated fruit fly optimization algorithm based on Sarsa-learning is proposed.Thirdly,this paper uses two representative algorithms in the existing research as comparison objects to verify the effectiveness of the proposed algorithm.Then,standard numerical cases were solved using all algorithms,and the experimental results were statistically analyzed.The results show that,compared with the other algorithms,the proposed algorithms has good performance in terms of HyperVolume,Mean Ideal Distance and CPU running time etc metrics.Finally,in the above theory and research methods,this paper illustrates the practicability of the proposed methods of this paper through an example of product development project of AR-based locomotive product thread assembly management platform.更多还原