Position and velocity predictions of the piston in a wet clutch system during engagement by using a neural network modeling

Abstract

In a wet clutch system, a piston is used to compress the friction disks to close the clutch. The position and the velocity of the piston are the key effectors for achieving a good engagement performance. In a real setup, it is impossible to measure these variables. In this paper, we use transmission torque and slip to approximate the piston velocity and position information. By using this information, a process neural network is trained. This neural predictor shows good forecasting results on the piston position and velocity. It is helpful in designing a pressure profile which can result in a smooth and fast engagement in the future. This neural predictor can also be used in other model-based control techniques. © 2012 IFIP International Federation for Information Processing.