In this paper, a new algorithm for the off-line estimation of wet clutch friction parameters is proposed for automotive transmissions, motivated by the usefulness of such an algorithm for diagnosing the condition of the clutch and transmission fluid in service. We assume that clutch pressure is measured, which is the case in dual clutch transmissions (DCT). The estimation algorithm uses measured rotational speeds and estimated accelerations at the input and output sides of a clutch, measured clutch pressures, and a simplified dynamic model of clutch friction to estimate the viscous and contact components of clutch friction torque. Coefficient of friction data is generated using the contact friction torque. A Stribeck friction model is fit to the data, and parameters in the model are then calculated by applying linear least squares estimation.
The proposed estimation algorithm is tested using the simulation of a powertrain utilizing a DCT, where the clutch friction model incorporates a realistic level of fluid film squeeze dynamics. The algorithm is evaluated under ‘noiseless’ and ‘noisy’ conditions. In the ‘noiseless’ case, clutch accelerations and pressure derivatives are obtained by differentiating stored clutch speed and pressure signals. In the ‘noisy’ case, white-Gaussian noise is introduced to the stored signals, with a noise-to-signal amplitude ratio of 10% for speed signals and 5% for pressure signals. The clutch accelerations and pressure derivatives are then estimated using a Kalman filter, and results are presented for both ‘noiseless’ and ‘noisy’ cases. In both cases, friction parameters are estimated within 10% of their respective simulated values, with the estimation accuracy being generally better for the ‘noiseless’ case.