关键词: 模具制造, 随机动态规划, 马尔可夫链, 启发式策略, 仿真, Q学习

Abstract: To assure on-time delivery of mould and die, by analyzing uncertainties of the durations, costs and rewards as well as the characteristics of frequent repairing in the mould and die manufacturing project, a stochastic evolution model of multiple mould and die manufacturing projects was proposed based on a discrete time Markov chain. And an algorithm framework to solve this stochastic dynamic programming model was proposed. Through simulation of heuristic policies and Q-learning, the'curse of dimensionality'was overcome. Finally, the algorithm framework was explained by an example, and its feasibility and reliability were also verified.

Key words: mould and die manufacturing, stochastic dynamic programming, Markov chains, heuristic policies, simulation, Q-learning