带有不相容工件簇的单机随机调度问题研究

发布时间：2018-04-23 01:38

本文选题：不相容工件簇 + 随机调度　；参考：《合肥工业大学》2017年硕士论文

【摘要】：在生产制造业中,存在一类带有不相容工件簇的批处理机调度问题;其中,不相容工件簇是指属于不同簇的工件不能被安排在同一批中加工。本文研究了不相容工件簇在单台批处理机上的随机调度问题,其中工件具有随机的到达时间和加工时间。本文首先根据不相容工件簇在单台批处理机上的分批加工特点,以缓冲库的库存量为实时状态,以最小化系统长期平均代价为优化目标,建立了半马尔可夫决策过程模型。然后,在此模型的基础上,通过策略迭代算法求解其最优控制策略,并分析了不同缓冲库容量对系统工件流失率的影响。最后,针对工件簇总数增大时引起的“维数灾”问题,给出了基于模拟退火的Q学习算法;并分析比较了工件簇总数为二和三时,两种算法的优化效果以及工件到达率和处理率对系统性能的影响。在现代化先进生产环境中,为了减少库存成本和生产浪费,单纯以追求生产效率为目标的传统生产模式已逐渐被及时生产(just in time,JIT)方式所替代。JIT生产方式是指在需要的时候,按需要的量生产所需的产品,因此本文研究了需求驱动下不相容工件簇的单机随机调度问题。首先,以缓冲库和成品库的库存量为联合状态,以最小化系统长期平均代价为优化目标,建立了半马尔可夫决策过程模型。由于需求驱动系统模型具有较大的状态空间,为了避免理论求解的“维数灾”问题,通过基于模拟退火的Q学习算法分析了不同需求类型对系统性能的影响。最后,通过仿真结果分析说明了所建模型的合理性以及优化算法的有效性。
[Abstract]:In the manufacturing industry, there exists a class of batch scheduling problems with incompatible job clusters, in which incompatible job clusters refer to jobs belonging to different clusters that cannot be processed in the same batch. In this paper, the random scheduling problem of incompatible workpiece clusters on a single batch processor is studied, in which the workpiece has random arrival time and processing time. In this paper, according to the batch processing characteristics of incompatible workpiece clusters on a single batch processor, a semi-Markov decision process model is established based on the real-time state of the buffer stock and the objective of minimizing the long-term average cost of the system. Then, based on the model, the optimal control strategy is solved by the strategy iterative algorithm, and the influence of different buffer capacity on the workpiece loss rate is analyzed. Finally, a Q-learning algorithm based on simulated annealing is proposed to solve the problem of "dimensionality disaster" caused by increasing the total number of workpiece clusters, and the total number of workpiece clusters is analyzed and compared between two and 03:00. The optimization effect of the two algorithms and the influence of the workpiece arrival rate and the processing rate on the system performance are discussed. In modern advanced production environment, in order to reduce the cost of inventory and waste of production, the traditional production mode with the goal of pursuing production efficiency has been gradually replaced by the mode of timely production just in time JIT.JIT production means when needed. In order to produce the required products according to the demand, this paper studies the stochastic scheduling problem of a single machine for a set of incompatible jobs driven by demand. Firstly, a semi-Markov decision process model is established with the inventory of buffer and finished product as the joint state and the objective of minimizing the long-term average cost of the system as the optimization objective. Because the demand-driven system model has a large state space, in order to avoid the "dimension disaster" problem solved by theory, the influence of different requirement types on the system performance is analyzed by means of the Q learning algorithm based on simulated annealing. Finally, the simulation results show the rationality of the model and the effectiveness of the optimization algorithm.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】