关键词: 虚拟单元, 重调度, 重调度驱动决策, 混合粒子群算法

Abstract: In the actual production environment, new orders arrive in succession. Usually, a new optimal scheduling plan is built every time a new order arrives, to make the production system runs more efficiently. As the scheduling plan changes too frequently, the "nervousness" of scheduling production has appeared. The control of the workshop is affected, and the stability of the production system is weakened. At present, studies on rescheduling problems mainly focus on the optimization of rescheduling method and the evaluation of rescheduling performance. The rescheduling driven decision-making has not gained enough attention.The rescheduling driven is a mechanism which determines when to reschedule.
The rescheduling driven in the virtual cellular manufacturing system is proposed, in order to reduce the times of rescheduling with new orders arriving continually, while keeping the high efficiency of the production system.
A new rescheduling driven decision-making method is propsed through the loss and gain of scheduling in virtual cellular manufacturing system. In this paper, loss and gain of scheduling is assessed from the perspective of time. The starting time deviations of tasks which exist in both the new scheduling plan and the old one are calculated, to measure the loss. Difference between the completion times before and after the rescheduling is considered as the profit of scheduling.
To ensure the optimization of rescheduling, a nonlinear 0-1 integer programming mathematical model with multi-objectives is established according to the characteristics of virtual cellular manufacturing system. Then a discrete particle swarm optimization algorithm with genetic operation is designed to solve the rescheduling model. Based on the rolling time horizon theory, the whole plan is divided into several periods by rescheduling points. Alternative rescheduling points are generated by a mixture of periodic and event strategy. At each point, the loss is compared with the gain of rescheduling based on the mathematical programming model, to find out whether it is appropriate to implement the rescheduling. If the gain can make up for the loss, the rescheduling will be carried out immediately; otherwise, this alternative time point will be canceled. With the proposed decision-making method, the reasonable rescheduling points are selected and the unnecessary ones are filtered out. As a result, the frequency of rescheduling is reduced.
The case of shipbuilding verifies that the proposed rescheduling driven decision-making method is feasible and efficient. The analysis of the case indicates that, compared to other driven strategies, this method can obviously decrease the rate of changing production plan without worsening scheduling performance index.

Key words: virtual cellular, reschedule, rescheduling driven decision, hybrid particle swarm algorithm