【作者】 张亮；

【导师】 王凌；

【作者基本信息】 清华大学 ， 控制科学与工程， 2004， 硕士

【摘要】 仿真优化问题具有很强的工程背景，它研究基于仿真评价的优化。仿真优化问题具有大规模性、多极小性、不确定性、建模困难、评价费时和NP完备等难点，因此它一直是学术界和工程界共同关注的热点研究课题。生产调度问题是一类典型的仿真优化问题，有效的调度与优化技术是实现先进制造和提高生产效率的基础和关键。近年来，计算智能的研究受到了广泛重视，并大量应用于生产调度领域。本学位论文针对仿真优化的诸多难点，重点研究基于序的智能算法，并用于求解典型调度问题。论文工作主要包括以下方面：1. 对仿真优化、智能优化、序优化进行了综述。2. 针对确定性优化问题，将序优化思想融入遗传算法，提出了一类基于序的遗传算法，并有效解决了Flow Shop调度问题。3. 针对随机优化问题，一方面，通过结合最优计算量分配(OCBA)技术和所提出的基于序的遗传算法，提出了遗传序优化方法；另一方面，通过将假设检验的统计思想融入智能优化算法，分别提出了基于假设检验的模拟退火和遗传算法。所提各种方法有效解决了加工时间随机的Flow Shop调度问题。4. 针对模拟退火和遗传算法的参数和操作选取问题，通过将其描述为随机优化问题，提出了基于OCBA的解决方法。同时，基于奖惩机制提出了自适应多操作遗传算法，并有效解决了流水线调度问题。5. 针对有限缓冲区的Flow Shop调度问题，提出了结合基于有向图的局部搜索和遗传算法的有效混合算法。最后，对本文工作给予了总结和展望。更多还原

【Abstract】 Simulation optimization studies the optimization problems with simulation based evaluation, which has strong engineering background. Simulation optimization problems have hardnesses, such as large scale, multi-modal, randomness, hardness to model, time-consuming evaluation and so on, so it is always a focus both in academic and engineering fields. Production scheduling problems are a class of typical simulation optimization problems, and effective scheduling and optimization technologies are critical to realize advanced manufacturing and improve production efficiency. In recent years, computional intelligence has obtained intensive attention and wide applications in production scheduling field. Aiming at the difficulties of simulation optimization problems, this dissertation stresses the study on order-based intelligent algorithms and the applications for typical production scheduling problems.The main contents of this dissertation are as follows.Surveyes on simulation optimization, intelligent optimization and ordinal optimization are provided.For deterministic optimization problems, a class ordr-based genetic algorithm is presented by incorporating the idea of ordinal optimization into genetic algorithm (GA), which is applied to effectively solve the flow shop scheduling problems.For stochastic optimization problems, genetic ordinal optimization approach is proposed, which combines the optimal computing budget allocation (OCBA) technique with the proposed order-based GA. In addition, hypothesis-test based simulated annealing (SA) and genetic algorithms are proposed respectively by incorporating the statistical idea of hypothesis test into intelligent algorithms. All the proposed <WP=5>algorithms can effectively solve the flow shop scheduling with random processing time.Aimed at the problem of selecting suitable parameters and operators for SA and GA, an OCBA based approach is proposed by formulating the considered problem as a stochastic optimization problem. Furthermore, an effective adaptive GA with multiple operators is proposed with the idea of award-penalty for flow shop scheduling problems.By combining the directed graph based local search with GA, an effective hybrid algorithm is proposed for flow shop scheduling with limited buffer size.Finally, summarization and prospect are provided for the work of this dissertation.更多还原