Prediction of Production Performance in Smart Manufacturing Using Multivariate Adaptive Regression Spline

With the dynamic arrival of production orders and ever-changing shop-floor conditions within a production system, production scheduling presents a challenge for manufacturing firms to ensure production demands that are met with high productivity and low operating cost. Before a production schedule is generated to process the incoming production orders, the production planning stage must take place. Given the large number of input parameters involved in production planning, it is important to understand the interactions of input parameters between production planning and scheduling. This is to ensure that production planning and scheduling could be determined effectively and efficiently in achieving the best or optimal production performance with minimizing cost. In this study, by utilizing the capabilities of data pervasiveness in smart manufacturing setting, we propose an approach to develop a surrogate model to predict the production performance using the input parameters from a production plan. Based on three categories of input parameters, namely current production system load, machine-based and product-based parameters, the prediction is performed by developing a surrogate model using multivariate adaptive regression spline (MARS). The effectiveness of the proposed MARS model is demonstrated using an industrial case study of a wafer fabrication production through the random sampling of varying numbers of training data set.