Data-Driven Modelling of Dry Clutch Lining Wear Process and Coefficient of Friction

Dry friction clutches are key components of automotive powertrain systems, where wear behaviour and friction coefficient variability directly affect durability, comfort, and transmission control performance. Although dry clutches have long been used in manual transmissions, their application in modern automatic and automated transmissions requires accurate and robust models of wear and friction behaviour that are currently insufficiently addressed in the literature. The objective of this doctoral dissertation is the experimental characterization of wear and coefficient of friction (COF) in dry clutches with organic friction linings, and the development of data-driven stochastic models for predicting their expected values and variability. A cycle-wise wear model is developed for use in transmission system simulations and for online estimation of clutch remaining useful life, while the COF model predicts the instantaneous friction level and its uncertainty under given operating conditions. Wear characterization was conducted using a custom-built disc-on-disc CNC tribometer for three friction materials. The results revealed a pronounced run-in phase with elevated and highly variable wear, followed by a stabilized wear regime. In the stabilized phase, clutch temperature was identified as the dominant factor influencing wear, followed by initial slip speed, while clutch torque and closing time had a weaker effect. Based on these findings, a stochastic cycle-wise wear model consisting of run-in, wear rate expectation, and wear variability submodels was developed and successfully validated on independent experimental data. COF characterization was performed using a large dataset collected during wear experiments, revealing significant inherent variability and several non-linear effects. A stochastic COF model, comprising expectation and variability submodels as functions of temperature, normal force, and slip speed, was developed and validated, demonstrating reliable prediction of both mean values and confidence intervals. The main scientific contributions of this work are experimentally validated stochastic models for cumulative wear and COF prediction in dry clutches, providing a robust foundation for advanced transmission control strategies and realistic powertrain simulations.