关键词: 机器故障, 柔性作业车间, 重调度方式决策, 数据仿真, BP神经网络

Abstract: After a machine failure,the production rescheduling mode utilized shows a direct effect on the production efficiency and stability within flexible job shops.For this reason,a decision-making model combining data simulation,genetic optimization and BP neural network was proposed,so that the optimal rescheduling mode could be estimated economically,efficiently and quickly under a given fault situation.For the flexible job shops,a myriad of re-scheduling schemes were first generated for different simulated failure situations with all the predefined re-scheduling modes.By comparing three evaluation indicators,maximum completion time difference,completion time difference and process adjustment cost,one rescheduling mode with the smallest comprehensive index was labeled as the optimal mode in a given situation,and a large-scale labeled sample set was hereafter generated.On this basis,a decision-making model combining genetic optimization and neural network was proposed to explore the internal relationship between machine failures and rescheduling modes and to estimate the optimal rescheduling mode for each given fault circumstance.In this decision-making model,the genetic algorithm was employed to first determine the structure of the BP neural network,and then to optimize the weights and thresholds in the network.Experimental results showed that the proposed decision-making model could significantly improve the decision-making efficiency and response capability of the flexible job shops in a sudden fault.

Key words: machine failure, flexible job shop, rescheduling mode decision, data simulation, BP neural network