Analysis and design of split and merge unpaced assembly systems by metamodelling and stochastic search

The study presents a new approach in optimal interstage buffer allocation problem of split-and-merge unpaced open assembly systems, which are increasingly being used in modern manufacturing systems, particularly in automotive industries. Allocations of interstage buffers to accommodate the work-in-process inventories are optimized in an attempt to maximize the overall system production rate. A simulation model developed is used in conjunction with genetic algorithms (GA) to find optimal interstage buffer configurations yielding a maximum production rate. However, simulation is extremely time-consuming due to lengthy computational requirements, especially when used in a stochastic search algorithm. In an attempt to overcome this problem, an alternative approach in simulation metamodelling based on artificial neural networks (ANN) is developed. The optimization problem previously conducted through simulation and GA is reconsidered by integrating the metamodelling approach into the GA, replacing the simulation model. The new ANN–GA approach not only gives solutions with no statistically significant difference in comparison with the original simulation GA approach, but also demands significantly less computational time. The proposed methodology intends to help practising system design engineers to take quicker decisions regarding the assembly system design parameters. The potential of metamodelling to solve manufacturing systems problems is also discussed.