Model Based Nonlinear Observers for Torque estimation on a Combustion Engine Test Bench

Torque is an important quantity in combustion engine test bench control but unfortunately very difficult to measure. Observers can be used as an alternative to sensors to estimate the torque based on a nonlinear model and the available measured output signals, engine speed and dynamometer speed. In this paper three nonlinear observers are designed for a combustion engine test bench simulator including combustion oscillations as well as noisy measurements and disturbed inputs. The well known Extended Kalman Filter (EKF), a High Gain Observer (HGO) and a Sliding Mode Observer (SMO) are compared in terms of the quadratic estimation error, computing time and convergence rate. Finally a comparison with a linear Kalman Filter which is still frequently used is shown.