Toward data-driven production simulation modeling: dispatching rule identification by machine learning techniques

Abstract

Production simulation is useful to predict and optimize future production. However, it requires effort and expertise to create accurate simulation models. For instance, operational control rules, such as job sequencing rules, are modeled based on interviews with shop-floor managers and some assumptions since those rules are tacit in general. In this paper, we consider a data-driven approach to model operational control rules. We develop job sequencing rule identification methods that model rules from production data using machine learning techniques. These methods are evaluated based on accuracy and robustness against uncertainty in human decision making using virtual and real production data.