GridMPC: a Service-Oriented Grid Architecture for Coupling Simulation and Control of Industrial Systems

It is now commonplace for industrial systems to be automatically controlled by computers. Furthermore, it is becoming more common for industrial systems to use Model Predictive Control (MPC) software to optimize the behaviour of a physical process through manipulation of control variables. It has been shown that MPC control can be improved through coupling with Computer Aided Engineering (CAE) simulations. We propose a grid architecture, GridMPC, to provide MPC coupled with CAE simulations for the realtime control of industrial systems. GridMPC bridges the gap between powerful simulation tools and actual physical processes. A job management service makes computing resources available to the grid for CAE simulations and a repository service stores the simulation results to avoid identical queries and provide immediately the data to the process controller. GridMPC has been built using web services, taking security in consideration. The overhead of the grid architecture is found to be acceptably small.