【作者】 吴珊珊；

【导师】 李蓓智；

【作者基本信息】 东华大学 ， 机械制造及其自动化， 2011， 博士

【摘要】 客户的个性化、多样化需求使全球市场的竞争异常激烈。如何在有限时间和有限资源的情况下,最大限度地满足客户需求?如何实现多品种小批量及大规模定制条件下生产的优化调度?如何解决大规模、多目标、多资源约束等复杂工况下智能调度问题?本课题研究正是围绕上述关键问题展开的,研究的主要内容包括：(1)提出了分解-优化-融合的智能调度策略针对大规模、多目标、多资源约束等复杂工况下的调度问题,制定了分解-优化-融合(Decomposition+Optimization-Integration,DOI)智能调度策略,规划和设计了分解-优化-融合策略下不同阶段的优化方法,即首先根据分解规则,对制造系统调度周期内的相关信息进行分类,分解成若干个调度单元：然后对不同调度单元分别给出优化调度方案；最后基于复杂大规模制造系统的总目标和资源约束,对所有调度单元进行融合,完成了优化调度集成方案,有效地攻克了调度规模增大、解空间呈指数增长的技术难题。(2)构造和实现了分解规则及其计算模型在分析和研究调度类型、调度目标、资源约束及其相互冲突的基础上,根据分解阶段的优化项目要求,构造了分解规则及其计算模型,如交货期富裕度、工艺相似性等,可以给出相关单元信息,为不同单元的优化与融合提供支持。(3)构造了基于生物智能的单元调度优化算法在生物免疫系统和遗传进化机理的研究基础上,建立了基于生物免疫进化机理的生物智能计算方法,对智能计算方法与其他策略和技术的综合在复杂调度问题的应用和实现进行研究,克服了复杂调度问题建模困难和计算方法设计复杂的局限性,拓展了调度问题的研究方法并提高了运算效率。(4)提出和实现了权重自适应智能算法实际制造过程中,由于不同企业、不同调度对象、不同调度周期等,调度目标及其组合关系非常复杂。通过建立多目标优化调度问题的模型和对基于生物智能算法的特性分析,结合了生物智能优化算法的优点,提出了多目标的权重自适应智能算法(Weighted Self-Adaptive Intelligent Algorithm,WSAIA),通过生物智能算法种群的进化及设置不同级别的繁衍系数,降低对目标的人为干预或盲目设定的影响,确保种群多样化,平衡全局搜索和局部寻优,提高了多目标调度问题的求解效率和质量。(5)提出和实现了基于混沌的改良免疫算法构建了满足工艺约束与资源约束,以总工期最小为目标的资源受限调度问题的数学模型。研究了混沌系统的特征,设计了由多个混沌函数(Logistic, Tent和Sinusoidal)构成的混沌生成算子。分析了基于生物智能优化算法的特点,引入混沌生成算子和并行变异算子,提出了基于混沌的改良免疫算法(Chaos-based Improved Immune Algorithm, CBIIA)。在种群初始化阶段,用混沌生成算子替代传统的随机数生成方式。在变异阶段,提出了基于高斯策略和柯西策略的并行变异操作替代常用的点变异,并行变异操作中用柯西策略实现大步变异,用高斯策略实现小步变异,以平衡全局搜索和局部寻优性能。(6)研制了面向复杂制造的智能调度系统并进行仿真测试开发了智能调度系统,用基于标准案例设计的大规模调度问题,文献中的多目标调度问题和选自标准案例库的多资源约束调度案例进行测试。将测试结果与文献结果比照分析,算法结果和性能证明了提出的方法和策略的有效性。更多还原

【Abstract】 Customer needs are increasingly personified and diversified and require quick response, which cause fierce competition in the global market. How are customer needs satisfied to the largest extent with limited amount of resource and time? How is production optimization scheduling with multi-variety and small batch realized? How is an intelligent scheduling problem for a complex manufacturing system with large-scale, multi-objective and multi-resource constraints delivered?This study conducts relevant research regarding the above-mentioned issues, and the main contributions of this study include as below,(1) Development of a decomposition-optimization-integration intelligent scheduling strategyOn the basis of analysis of a large number of existing scheduling methodologies, this study proposes a three-fold (Decomposition-Optimization-Integration, DOI) approach for solving large-scale job shop scheduling problems and designs optimization item in three different stages. Firstly, in terms of classification and decomposition rules, manufacturing scheduling information is classified in order that scheduling units are achieved in a reasonable manner. Secondly, according to scheduling objectives and manufacturing information of scheduling units, schedules are made by the use of intelligent algorithm. Lastly, the final schedule for complex large-scale manufacturing system is optimized through "integration" approach according to overall scheduling objectives and resource constraints. Thus, the proposed methodology can solve NP-hard problems characterized by enormous solution space in an effective way.(2) Design of decomposition rules and related calculation modelsBased on analysis of features of a complex manufacturing system, such as scheduling type, scheduling objective, resource constraints and their contradictory relations, this study designs decomposition rules and relevant calculation models according to the requirements of optimization units, namely, due dates, and process similarity, etc. in order to support the integration of different scheduling modules.(3) Research on biological intelligence based scheduling optimizationThe study introduces the mechanisms of biological immune system and genetic evolution, analyzes the mechanisms of biological immune system and genetic evolution to solve complex scheduling problems, and discusses the applications of intelligent approaches combined with other strategies on complex scheduling problems.(4) Research on weighted self-adaptive intelligent algorithm for multi-objective scheduling problemIn the process of manufacturing on the floor shop, scheduling objectives and their relations are extremely complicated due to different goals in different enterprises over scheduling horizons. The study models multi-objective scheduling problem. Afterwards, based on the features of biological intelligent algorithm, the study proposes a weighted self-adaptive intelligent algorithm (WSAIA) for a multi-objective scheduling problem. By the use of evolution of intelligent algorithm and reproduction coefficient, it can overcome the limitations of conventional weighted-sum in which the importance of each objective are manually set in advance, furthermore ensure the diversity of population and balance the exploration and exploitation so that it can increase the effectiveness of search for optimal solution considering overall objectives.(5) Development of chaos-based intelligent scheduling algorithmThis study models resource constrained project scheduling problem which features multi resource types. The objective is to minimize makespan with satisfying precedence and resource constraints. The study devises a chaotic generator by using Logistic function, Tent function and Sinusoidal functions. Analyzing the features of artificial immune system, the study introduces chaotic operator and parallel mutation operator. Therefore, the study proposes Chaos-based Improved Immune Algorithm (CBIIA). In the initialization phase, chaotic generator is utilized instead of conventional random number generator. In the mutation phase, parallel mutation is deployed rather than point mutation. Parallel mutation comprises of two mutation strategies viz. Gaussian and Cauchy. Gaussian strategy is applied for small step mutation and Cauchy strategy is applied for large step mutation. The objective of parallel mutation mechanism is deployed to balance exploitation and exploration in search space.(6) Development of an intelligent scheduling software systemThis study develops an intelligent scheduling software system. Large-scale simulated instances based on test bed of job shop scheduling problem, a multi-objective job shop scheduling problem in the literature, and benchmark problems for resource-constrained project scheduling problems are tested. Test results are analyzed and compared with existing methodologies in literature, and it is proven that the proposed methodologies are effective in converging towards the optimal solution.更多还原