PPI: Erik Ydstie
University: Carnegie Mellon University
The manufacturing industry in the US and around the world is currently engaged in a fundamental transformation as it moves towards a new level of automation based on the availability of almost limitless computational resources and huge amounts of real time data. The transformation taking place is sometimes referred to as the fourth industrial revolution. Advanced Process Control (APC) systems based on Model Predictive Control provide one foundation technology. However, MPC systems are presently time consuming and expensive to design and implement.
The proposed approach speeds up the design process and, as importantly, machine learning ensures that the MPC system learns and optimizes its performance as time progresses. This vision is based on the integration of machine learning, cloud-based data management and MPC into one platform for Advanced Process Control. Our research is based on the cloud-based data management supplied by OSIsoft in combination with a state-of-the-art APC system supplied by Emerson Delta V system and process control and systems tools supplied by MATLAB. Together, these software and hardware systems provide platform technology for the application of machine learning process control being developed in our research group at CMU. The proposed technology will be tested at Vitro Glass furnaces in Fresno CA, Meadville PA, and Carlisle PA. The real-time control and testing will be carried out by Industrial Learning Systems Inc., a CMU spin-off. ILS has entered into an agreement with CMU (exclusive license to adaptive control technology) and Vitro to develop software for glass furnace control.