Cyber-Physical Manufacturing Systems: Improving Productivity through Advanced Automation

Abstract

Advances in computing and networking technologies have connected manufacturing systems from the lowest levels of sensors and actuators, across the factory, through the supply chain, and beyond.  Large amounts of data have always been available to these systems, with currents and velocities sampled at regular intervals and used to make control decisions, and throughputs tracked hourly or daily.  The ability to collect and save this detailed low-level data, send it to a central repository, and store it for days, months, and years, enables better insight into the behavior – and misbehavior – of complex manufacturing systems.  The output from high-fidelity models and/or reams of historical data can be compared with streams of data coming off the plant floor to identify anomalies.  Early identification of anomalies, before they lead to poor quality products or machine failure, can result in significant productivity improvements.  We will discuss multiple approaches for harnessing this data, leveraging both physics-based and data-driven models, and how automation can enable timely responses. Both simulation and experimental results will be presented.