Contents

Part I
Background, motive, visions

1 Background and industrial motivation

Combustion engines are complex, highly engineered systems, where established base techniques in the technological sciences are combined with industrial engineering practice to yield a system that fulfills performance requirements. Engine control systems is a central component and strategically important for achieving the desired performance. Legislators and customers are technology drivers through their requirement on reduced emissions and demands for improved fuel economy, driveability and comfort.

The development is in a direction where new components are added, giving the control systems more degrees of freedom for optimizing the performance, which in addition increases the complexity and makes them more difficult to handle. The message from the industry is clear that traditional design methods, based on expensive test cell calibration of look up tables, will not suffice for the new complex systems. New systematic design methods that can handle the system complexity must therefore be developed. A solution is model based methods that have the potential to handle the increased complexity and provide the engineers with the necessary tools. Modeling and analysis of how new control strategies and control variables affect the engine performance both with respect to emissions and efficiency are therefore central areas.

To successfully apply the model based approach both consolidation of available model knowledge and new model development is needed. Only then will we be able to cover a complete system with the necessary models. In addition it is important to also develop systematic methods for acquiring and adjusting the parameters of the models. Another area of high industrial importance is the automatic generation of program code from simulation and validation environments directly to the electronic control units, and we are aiming at developing models that are implemented in such environments e.g. Simulink.

A requirement for the control loops is that there are sensors or virtual sensors (observers) available which can be used to give the necessary information to the control systems. It is also clear that there will be a demand for even more information in the future so that the control system can optimize the trade-off between emissions and fuel consumption. An important problem is that many of the quantities that influence performance are not directly measurable in an real application, one example is the amount of oxygen in the combustion chamber. For control purposes it is desirable to acquire information about such important quantities with as small time delay as possible, which highlights the importance of applying observers. Furthermore, for cost reasons it is obvious that it is preferable to utilize algorithms if possible, instead of having to introduce extra sensors.

A component or module based thinking with systematic tuning of model parameters is also tractable, especially during the design phase when experiments are done by switching components without having to recalibrate all control and diagnosis function for the models in the system. The complexity of the models must be appropriate for implementation in a control system and there is a gap between the mean value engine models that can be used in the control system and the complex 1D-3D FEM models that have best predictive capabilities. There is a need for a link between these model areas. Research in this direction requires a cross-scientific approach combining control theory and engine research.

Reducing fuel consumption is a key issue and this project is focusing on concepts that fall into the category of down-sizing and supercharging. Supercharging have several advantages where some are obvious, but some are sometimes overlooked. For example they are not relying on the development of new catalysts nor new sensor technologies which is necessary for some other proposed new technologies. There is an increased industrial effort put into developing supercharging systems such as for example classical turbo, variable geometry, double turbo, compressors and turbo compound. This is an area where there is a lack of accurate models suitable for control design.

1.1 International perspective

Sweden is among the world leaders when it comes to turbocharged engines. SAAB were pioneers in turbo on production cars, and continue to lead the development as they are appointed within GM to be the center of expertise for turbocharged engines and for engine control.

At the project start the following research groups had the most recent publications with advances in control oriented models for turbocharged engines: DTU Denmark have published new models for the compressors. Ford have studied methods of parameter estimation for a set of published models. ETH Switzerland have published work on modeling and control of turbocharged diesel engines. Our group has contributed with both new models and a component based modeling methodology.

2 Visions and Plans

The vision is to build a strong research group, with engine control system competence, that will be an attractive partner for both industry and academia in future engine research projects. Our overall goals are to deliver relevant research but also research educated persons with central competences within modeling, simulation, and control of combustion engines. The aim in this project is to develop methods and tools that help the automotive industry in achieving emission goals and optimizing fuel consumption through model based methods. Furthermore our long term efforts are to integrate combustion engine control with the complete driveline and vehicle control, and employing methods for optimizing the entire system, making further performance improvements possible.

During the next three year period we aim at strengthening our already existing collaborations with SAAB, SCANIA and Hoerbiger Control Systems (previously known as Mecel), and solve some of the relevant industrial research problems outlined in the project descriptions below. Examples are: Control of advanced turbocharged engines. Control oriented models of the gas flows in variable cam timing engines.

Part II
The Project

All subprojects that fall in the area of this CENIIT project have been started and are now ongoing. The projects recent results are summarized here while the projects are described in more detail below.

3 Summary of results

While year 2009 saw the conclusion of two projects, where two PhD students finished their theses [29, 30], the year 2010 has seen the start-up of new PhD projects. These fall within the general area modeling and control of advanced turbo charged engines and are covered by the two subproject areas “Mean Value Engine Modeling” and “Control and optimization of gas flows in turbo engines”.

Concerning the academic results one book chapter [33] and one journal paper [34] have been published. Complementing these there are seven conference papers [35, 36, 37, 38, 40, 39, 41] that have been published since the previous yearly report.

4 Project description

The goal with the project is to build a strong research group centered around control of advanced combustion engines. The proposed project builds upon previous results which have solved many problems but have also spawned new ideas that provide a starting point for future projects. Five prioritized areas around engine modeling and control are listed below. The recurring theme in this project is engine system models and the topics range from development of new component models to methodologies for analysis of system properties based on models. In addition to this the project is slightly widening to also cover integrated engine and driveline but the theme of system modeling and integration is maintained.

4.1 In-cylinder modeling

These models describe the important in-cylinder processes that generate work and emissions and can be used for evaluation and design of control schemes. However they are to detailed for being implemented in an engine control system. The emphasis is on finding and compiling simple but sufficient models that assist the control design and can be used in simulation to evaluate how new control schemes affect the performance. It is therefore one of our future core directions. This area builds upon my PhD thesis and is extended by the recent research documented and reported in the papers [1, 2, 7]. One part of the thesis [19] fall into this area where a variable compression ratio engine is modeled, and where a new formulation of multi-zone models is presented in [3]. It is based on the estimation using an in-cylinder model.

A continued development of the already existing multi-zone in-cylinder model [3] is planned where methods for systematizing the parameter identification by combining cylinder pressure measurements and regularization techniques. The first steps have been taken in the study of the special case of compression ratio estimation [10, 11]. Another interesting area is that the models that are received from zero-dimensional modeling are have many parameter and the system identification has to be supported using regularization techniques. In the last part of [18] regularization methods have been proposed and investigated for this application some results also appear in [33].

This modeling approach has now been applied to a variable valve timing engine and is used to study gas exchange effects, [12, 13, 20, 29], and can be seen as a bridge to the next subproject. In addition a computationally efficient model for the cylinder pressure has also been developed and it has been tailored to describe the ion current [31].

4.2 Mean value engine modeling

Mean value engine models cover the cylinder processes on a macroscopic scale and include other gas flow components like: intake, exhaust, throttles, intercooler and turbocharger. Mean value models are used for observer, control and diagnosis design since they describe the engine dynamics with a level of detail suitable for implementation in engine control systems. This is therefore a key area in the project. In particular we are interested in modeling and control of supercharged engines. A first step has been to compile and evaluate models for components in turbo charged engines which was reported in [5, 6].

In the next step a control oriented model for the gas exchange process in an engine with variable cam timing has been developed, these results have been published in [12, 13, 20]. Furthermore we have continued to develop a new model [27, 14, 15] that is purposed for fuel optimal control of the variable compression engine. Gas turbines is also modeled with a diagnosis purpose [40].

An area has been identified, that is still open, and it is control oriented modeling of turbochargers, i.e. compressors and turbines. Our recent modeling achievements indicate that dimensionless numbers, from fluid dynamics, are well suited for building compact models for the compressor. This has been explored further for both compressors and turbines [16, 21, 38]. In addition compressor surge control has been investigated in [25] where a novel concept of time to surge is introduced and used for control.

4.3 Control and optimization of gas flows in turbo engines

Gas flows have a direct impact on the emissions and fuel consumption of an engine and are important to control. Multi-variable control schemes that handles the basic flow requirements have been developed and analyzed for turbocharged engines in [8].

The project has continued to develop a controller for simulataneous control of exhaust gas recirculation (EGR) and variable turbine geometry on a diesel engine which minimizes the fuel consumption and the successful results have been reported [17, 9, 24, 26, 34, 41]. In addition it is in the future interesting to extend the fuel optimal control analysis for gasoline engines presented in [4] to also cover diesel engines.

4.4 Observer and sensor system selection

This area concerns systems analysis and method development that will guide future choices of sensors and their placement as well as the development of observers that estimate important control variables that aren’t measured. An important aspect is to handle filtering of signals without introducing time lags, otherwise the control system will not be as responsive as necessary for achieving the emission levels during transients.

The aim of this subproject is to clarify the conditions and specifications of different sensor system configurations. The problem of configuring an engine system is so complicated that the industry doesn’t know exactly what sensors to choose and their exact placement for best result. The interplay between system performance and sensor performance is thus central. A systematic analysis of this interplay will be based on system modeling of the complete system and a following analysis of sensitivity and optimality based on the system specifications.

Results on modeling and sensor quality have been reported in [22, 23, 28, 32, 39].

4.5 Observers for stiff gas flow systems

Gas flow models are nonlinear and stiff in certain operating regions. An implementation of generic gas flow observers in the control systems is therefore an unsolved problem. Here we have a new idea which is to develop a tailored discretization scheme based on the structure of the gas flow models, which will allow the system of equations to be solved efficiently and enable real time implementations for engine control systems. Work in this area has been started and is growing together with the observer related project above.

Part III
Resources and collaboration

5 Research Environment

The Division of Vehicular Systems is situated in the Department of Electrical Engineering and has strong support, both scientific and technical personnel, in systems science, control, signal analysis, and real-time systems.

The experimental facilities are centered around an engine laboratory collocated with a laboratory for real-time systems. Since engine control is complicated and requires coordination of many advanced techniques under hard real-time constraints, this type of collocation seems ideal. The engine test cell is dimensioned for two engines. It is currently equipped with two advanced research engines; one is the supercharged SAAB variable compression (SVC) engine and the other is a turbocharged SAAB engine with dual independent variable cam timing. The engines are equipped with additional sensors and instrumentation for modeling and validation. The engine laboratory uses a real-time system based on PC:s, an HP-system for data acquisition, an engine control unit from dSpace specially designed for rapid prototyping of control algorithms. Our well equipped lab is a result from our fruitful collaboration and support from SAAB and Mecel. The combination of experimental work and analysis together with intense collaboration with industry has been the key to recent success.

Our main industrial partners are SAAB, Hoerbiger Control Systems, and SCANIA with whom we have collaborative projects. We also have good international contacts and exchange of information with related groups like Professor Guzzella, ETH, Switzerland, Professor Rizzoni, Ohio State Univ., USA, and Professor Gissinger, Univ. Mulhouse, France. Lars Eriksson spent one year as post doc visiting Prof. Lino Guzzella, ETH.

There are also good relations within Linköping University, both with Computer Science and with Systems Engineering. Combustion and engine related research is a collaboration area within LINK-SIC and MOVIII.

6 Relation to other CENIIT projects

In general this project shares the methodology interests, concerning the usage and analysis of models for dynamic systems, with the projects that are associated with MOVIII and LINK-SIC. Furthermore there are two direct connections to the previous project 04.14 “Model-based diagnosis of technical systems”, that was managed by Erik Frisk, in that we share the supervision of 2 PhD students. The first is Erik Höckerdal that is working under the project area “Observer and sensor system selection”. This is a good meeting point where we both share application and methodological interests around how to extract information from complex processes. In this sub-project we have project meetings according to a regular schedule this gives us insight into each others work and competences. The second is Emil Larsson who is working on diagnosis of turbo machinery with Siemens in Finspång. This latter project has a direct relation to the mean value engine modeling subprojects that follow a modeling framework that can be directly utilized in diagnosis.

Part IV
Publications

References

Funding

This project is partly funded by CENIIT, see http://www.isy.liu.se/ceniit/.

Informationsansvarig: Lars Eriksson
Senast uppdaterad: 2011-01-17