Many industrial processes/systems have some of the following characteristics that impose high demands on controller capabilities:
- System constraints
- Multiple and/or complex objectives: economic objectives (e.g. energy, TCO) instead of “tracking error”
- MIMO and/or nonlinear dynamics
At the same time, two trends offer increasing preview information in processes/systems that could be exploited for improved controller performance:
- Predictions of scenarios/events/upcoming disturbances through big data
- (Many) cheaper (possibly noisy) sensors are available / heterogeneous embedded systems / adaptive software
Model Predictive Control (MPC) is the only methodology that can effectively address this combination of demands imposed on current industrial controllers! It is furthermore ideally suited to exploit preview information.
- Its industrial value has not yet been proven in mechatronics industry (proven in chemical processing industry!)
- Development and tuning requires experts in several domains
- Current MPC approaches are too “heavy” (memory&computations) for industrial computing hardware and/or it cannot be guaranteed that the control signal is available on time (within time range of control cycle)
-> Need deterministic economic MPC and approximate solutions
Bring deterministic economic MPC to application in Flanders’ mechatronic industry, making it part of the industrial state-of-practice control toolkit.
- Enable MPC and approximate MPC for industrial processes
- Focus on guaranteed solution times on less powerful industrial computing hardware
- Enable user prioritization of complex objectives and constraints to achieve desired balance between implementation cost and performance
- Reduce development and implementation costs and risks
- Focus on easy/user friendly, automated and accelerated MPC development and analysis of implementation issues (cost) vs. performance (profit)
- Demonstrate and validate
- Convince Flanders industry of high potential through industrial success stories
DIRAC is a Strategic Basic Research (SBO) project. We are looking for companies to join the User Group and work with us on the valorisation of the project.