Stochastic Modelling of Dry-Clutch Coefficient of Friction for a Wide Range of Operating Conditions

This paper presents a stochastic regression model for predicting the coefficient of friction (COF) in automotive dry clutches with organic linings. The influence of temperature, normal load, and slip speed on COF behaviour is investigated based on a large set of clutch wear-characterization data, collected using a custom-designed disc-on-disc tribometer that replicates realistic clutch-engagement cycles. The proposed model calculates both the expected value and standard deviation of the COF. The COF expectation model takes temperature, normal load, and slip speed as inputs, and it has a cubic polynomial form selected through a feature-selection method. The COF standard deviation model is fed by the same three inputs or alternatively the COF expectation input, and it is parameterized using the maximum likelihood method. The overall model is validated on an independent characterization dataset and an additional dataset gained through separate experiments designed to mimic real driving conditions.