Computing Collision Probability in a Series-Parallel Machines Model

抄録

In this paper, we consider collision probability in a manufacturing line model. Collision probability is the probability of jobs colliding in a manufacturing line and is a comparatively new evaluation item in systems in an unsteady state, having real-world applications in, for example, flat panel display and semiconductor manufacturing systems. Previous work in this area includes efficient methods to compute collision probability in an in-line machines model, a parallel machines model, and a generalized in-line machines model with consideration of delivery time and buffers. Two optimization problems with collision probability also exist for the generalized in-line machines model which address the minimization of tact time and total number of buffers. In an in-line machines model, when the processing time on a particular machine is longer than on others, this machine becomes a bottleneck in the production process and may lengthen the total production completion time. However, if two such machines are placed in parallel, then a reduction in the total production completion time is expected. In this paper, we generalize the in-line machines model to a series-parallel machines model and present a method which efficiently computes collision probability using computer simulation. The key idea regarding efficiency is the use of a heap data structure when searching for an idle machine among parallel machines. We also implement the method presented and show the findings obtained from computational experimentation.