Abstract
One of the problems encountered in the design and implementation of a serial production line (SPL) is the buffer size between the machine tools. The buffer size of the SPL has an important impact on the productivity of the whole production system. The machine tools' characteristics including their uptimes and downtimes and the process parameters are the main factors that affect the decision regarding the buffer size, and thus the productivity of the SPL. Due to the dynamic nature of this problem, it is complex to find the optimal buffer size in SPL. Thus, in this paper, an Efficient Prediction Model (EPM) is developed using Artificial Neural Network (ANN). The purpose of the developed EPM is to find the buffer size between each succeeding pair of machine tools in SPL at any given uptimes and downtimes of machine tools. An optimization model based on genetic algorithms (GA) is used to generate the learning data for the prediction model to find the optimal or near optimal buffer size of the bay of each machine tool in SPL. The proposed approach integrates the optimization and prediction methodologies to evaluate, and predict the optimal buffer sizes for maximum productivity. Including uptime and downtime parameters enable the proposed method to be used to improve the design of running SPL as well as to design a new SPL. Numerical examples for five and fifteen machine tools were conducted independently in this research and the results show the ability of the proposed method to determine the optimal buffer sizes in a reasonable amount of time. In particular, the results of case studies show that the developed model accurately predict the optimal buffer size, especially for the case of five machines and even for a higher number of machine tools yet with acceptable but less accuracy. Finally, the performance of the proposed approach was compared with some results of the state of the art methods reported in the literature. The comparison shows the superiority of the present approach to identify buffer sizes for higher throughput under the same uptimes and downtimes.