Combustion ProcessesCopyright: CC-BY-SA-3.0-2.5-2.0-1.0 (http://creativecommons.org/licenses/by-sa/3.0)
The Combustion Processes Group investigates model-based control concepts to enable and develop innovative combustion processes. Combustion processes are present in different industrial sectors, particularly within the application of combustion engines and gas turbines.
The system dynamics of combustion processes offer challenges for the control engineering:
- All combustion processes are nonlinear.
- Many combustion processes are inherently instable.
- Combustion systems can be modeled by the various arising physico-chemical processes, such as combustion chemistry, fluid mechanics, thermodynamics, and so on. However, the resulting models are too complex for an access to the application in control engineering. Therefore, these models need to be reduced for the purpose of control or other suitable data-driven models have to be identified.
- All combustion processes are multiscale: the relevant time scales range from very small as for chemical reactions to bigger time scales as for the cycle of a gasoline engine.
- The use of combustion engines leads to cyclic processes.
- The possibility of measurements is restricted. Reasons therefore are on the one hand the harsh environmental conditions, for example in the high temperature combustion chamber, which prevent the use of sensors or lead to a high noise level. Furthermore the absence of suitable sensors for high dynamic temperature measurement and the consideration regarding costs also prevent a use of sensors.
- The control task consists in reference tracking as for the load in a combustion engine and as well in the consideration of limit values for exhaust and noise emissions and the desire to be economic, for instance fuel efficiency.
Copyright: Reise Reise (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons
To manage these challenges, the Combustion Processes group researches on modern and innovative control algorithms that go beyond the common control algorithms used in industrial practice. The developed methods are centered around the field of model-based control.
- Use of real-time nonlinear model predictive control. Here, the systematic non-linear, unstable behavior is taken into account and at the same time reference tracking as well as the consideration of limit values can be achieved.
- Combination of model predictive and iterative learning control for cyclic processes.
- Nonlinear black-box identification for components, which can only be modeled with high complexity, using neural networks or piecewise-affine models.
- Development of reduced order white-box models.
- Nonlinear observer concepts to estimate the unmeasurable states quickly and reliably.
Selected Completed Projects
- Control of thermoacoustic instabilities
- Control of Diesel Homogenous Charge Compressiong Ignition, HCCI for short, and Gasoline Controlled Autoignition, GCAI for short, combustion engines
- Air path control for 2-stage turbocharged gasoline engines
- MILD- combustion in gas turbines