Rapid Prototyping of Engine Control Functions
Project ZAMOMO - Dovetailing of Model Based Software Engineering and Model Based Controller Design
In the development of software intensive embedded systems in control applications, such as engine control units, two different disciplines meet, which in the past had developed mainly separately from each other.
- Software Engineering
- Automatic Control Engineering
The cooperation of these two disciplines does not operate smoothly in practical experience: Different vocabularies of concepts and perspectives of the subject of draft causes misunderstandings. The lack of coordination in the design process inhibits the possible mutual methodological complement.
The aim of the project ZAMOMO, which is funded by the Federal Ministry of Education and Research, is to improve this situation through a appropriate integration of the design processes of software engineering as well as controller design.
In particular the both model based approaches (model based software engineering and model based controller design) should be dovetailed and complemented. As an exemplary application domain the engine control unit is determined, which is generated with the help of Rapid Control Prototyping.
In addition to the dovetailing of the two disciplines Software Engineering and Automatic Control Engineering, the "Virtual Development" of engine controllers is another main focus of this Institute.
The traditional control function development is mainly being accomplished on the test bench. But test bench runs are very time consuming and expensive. So since a long time it is attempted to reduce the test times to a minimum. Alongside the established use of CAE-Tools for the development of mechanical systems, the virtual development of controller concepts can also contribute to the whole development process of an engine.
One benefit of the virtual controller development is, that it can be started at the same time as the mechanical engine development. For the development of the controller software there is no longer the need for a real engine. With the help of commercial one dimensional simulation tools (like Boost or GT-Power) it is now possible to obtain the necessary data with a for controller development sufficiently accuracy. However these simulations are very time consumptive, whereby they are not useable for real time controller development.
For this reason the next step in the development chain, is the automated parameterization of real time capable Hardware-in-the-Loop-Models. In this project these models are build with the object oriented modeling language Modelica. The engine models should simulate the whole engine process with the resolution of one degree crank angle. Because of the high demand of computing time of a complete engine model and the need for real time capability the accuracy of the model can not be increased arbitrarily. One possible solution to increase the model accuracy is to build sub models, which can be distributed to several real time platforms. The developed models are integrated in an engine library to improve the reusability of these models for further projects.
The parameters of the developed HIL-Models are afterwards adapted to the data which where determined with the one dimensional process simulation tool. The task of the automated determination of the model parameters should be accomplished with the optimization tool AVL Cameo. At this step the parameters are optimized such that the results of the HIL-Model fits the data of the process simulation tool best possible.
Model Based Predicitve Controller (MPC) - Library
With the help of the real time capable HIL-Models, which can be adapted to different engine types, the next step is to develop controller functions. In the focus of the project ZAMOMO is the designing of scaleable controllers for the control of the air path of a diesel engine. For this purpose the possibilities of the scalability of a Model Based Predictive Controller (MPC) under consideration of calculating speed an memory requirements are investigated. Additionally non functional requirements – which are known from Software Engineering – should be included in the process of controller development and adapted to the needs of automatic control engineering.
Examples of these non functional requirements are:
- and performance.
The created controllers can then be tested and parameterized with the help of the HIL-Models, before they are used with the real engine at a test bench. The application of the parameters of the model based predictive controller like prediction horizon or weighting matrices should be evaluated in the last step of the virtual development also by the optimization tool Cameo. The developed controllers can then be tested and validated on the real engine. For this purpose the controllers, which where build with Matlab/Simulink, can be exported to the dSpace Micro-Auto-Box using the Realtime-Workshop for code generation.
The benefit of the virtual development is, that already tested and pre-parameterized controllers can be used at the real engine. The fine tuning of the controller parameters can therefore be much faster. Furthermore the risk of damaging the new engine is minimized, because of the pretests of the controller with the HIL-Models.
Within this project there will be the possibility for writing Master-Thesis or work as an research assistant. An actual list of free positions can be seen here. For further questions concerning the project ZAMOMO please don't hesitate to contact Peter Drews or Frank Hesseler