论文部分内容阅读
在生产制造过程中,设备状态的衰变会影响产品质量,导致成品率水平的逐渐下降.本文研究此类具有多成品率水平的衰变设备预防维修问题.建立隐马氏决策过程模型,在成品率水平不可直接获知的情况下,用产品质检信息作为系统观测状态进行维修决策.模型考虑两类质检误差及收益和成本参数,通过强化学习算法,学习各观测状态下的最优维修行动.针对不同的设备衰变模式和质检误差水平,进行算例分析,结果显示基于强化学习的预防维修策略与传统的固定周期的维修策略相比,能够很大程度上提高系统的平均收益.
In the manufacturing process, the decay of equipment status will affect the product quality, resulting in a gradual decline in the level of finished product.This paper studies the problem of the maintenance of decay equipment with multi-product rate.This paper establishes a hidden Markov decision process model, Level can not be directly informed of the situation, the product quality inspection information as the system observation state for maintenance decision-making model to consider two types of quality control error and gain and cost parameters, through reinforcement learning algorithm, learn the best maintenance action under each observation state. The results show that the preventive maintenance strategy based on reinforcement learning can greatly improve the average profit of the system compared with the traditional fixed cycle maintenance strategy.