An Artificial Neural Network Approach – Performance Measure of a Re-entrant Line in a Reflow Screening Operation

This paper presents an artificial neural network (ANN) method applied to a multistage re-entrant line system. Generally, queuing networks adopt analytical methods or use simulation packages to determine their performance measure. The contribution of this paper is the development of an alternate solution method using ANN approach to determine performance measure namely the total cycle time for a Reflow Screening (RS) operation in a semiconductor assembly plant. Performance measure of an operation is an important aspect in management decision making. In order to validate the proposed method, comparison results were made using the analytical method based on mean value analysis (MVA) technique for the re-entrant line and with some historical data collected from the operation. In this paper, Back Propagation Network (BPN) learning algorithm is proposed for the computation of the total cycle time with respect to the number of lots circulating in the system. Extensive training and testing of the proposed ANN method is performed which enables the BPN model to be used to determine the required total cycle time.