Stage Selection for Machine Combination Optimization in Multi-Process Production System

Abstract

In a multi-process production system, the yield rate of the final products depends not only on the performance of each machine but also on the combination of machines at different stages. In the previous study, it was demonstrated that field-aware factorization machines (FFM) can estimate the yield rates achieved by unused machine combinations and identify important machine pairs with high accuracy. However, when the number of stages is large and the data of used machine combinations is not enough, the prediction accuracy will be low. Hence, we proposed important stage selection methods. Through two numerical examples, we confirmed that RMSE between the actual and predicted values of yield rates decreased by up to 66 % in comparison with a model using all stages.