Observer-based real-time frequency analysis for combustion engine-based power trains with applications to identification and control

The typical motion of a combustion engine, such as the four-stroke Otto engine, is characterized by a periodic variation of its torque over the rotation angle of the driven shaft. This torque variation leads to the effect of time-varying accelerations and angular velocities. Modeling approaches for combustion engines that can be used for the implementation of control strategies for corresponding engine test rigs can follow two different approaches. Firstly, it is possible to describe the torque variations on the basis of the thermodynamic processes in the interior of the cylinder. Secondly, black box models can be derived on the basis of a frequency analysis of the measured torque and/ or the measured angular velocities and accelerations. Especially, black box models belonging to the second class are sufficient if a synchronization of a combustion engine with an electric motor/ generator unit acting on the same driven shaft is desired. However, classical frequency analysis techniques such as the Fast Fourier Transform are usually only applicable offline. Therefore, an online-applicable observer approach is developed in this paper which allows for the real-time frequency analysis of measured angular velocities. The estimation results, for both stationary and non-stationary operating points, are then applied on the one hand for a system identification and on the other hand for the implementation of control strategies in the framework of combustion engine test rigs. A numerical stability analysis and experimental results conclude this paper.