Applications of Data-driven Production Optimization for Heavy Processing Industries

Abstract

This talk focuses on real-time recommendations for the operations of manufacturing processes based on production optimization. The solution provides the predicted and optimized production in a future horizon without and with adopting our recommendation. We develop regression models based on the carefully selected control variables and real-time sensors as the current states of operation with the deep learning technique. We choose either non-linear or Mixed-integer linear programming optimization based on the process complexity and response requirement. We apply our data-driven solution to different heavy industrial applications, such as cement and paper production.