| Abstract |
In conventional combustion engines, the combustion process and oscillating masses produce torque
pulsation which affects the life cycle of the power train and also reduces the driver’s comfort. It has
been suggested to absorb this pulsation using an AC-machine as actuator which is mounted on the
crankshaft to generate an inverse torque sequence [2].
In this paper, an approach for online identification of this combustional torque pulsation will be
presented. A Harmonic Activation Neural Network (HANN) is used to estimate this periodic
sequence at variable operating points of a combustion engine. For compensation, the actuator is fed
with the identified and analytically phase corrected signal. The underlying algorithm copes with
restrictions such as reduced sensor quality, limited torque range of the AC-machine and time delayed
actuator operation. Simulation results are added for illustration. |
