A neural model of friction material behaviour

Abstract

The neural computation ability to model complex non-linear relationships directly from experimental data, without any prior assumptions about nature of the input/output relationships, has been used in this paper. An artificial neural network technique was used to develop a neural model for predicting the friction materials behavior under prescribed testing conditions. By means of neural modeling of the friction materials behavior, the relationship between 26 input parameters and one output parameter (brake factor C) has been established. The input parameters are defined by the friction material formulation (18 parameters), manufacturing conditions (5 parameters), and testing conditions (3 parameters). Prediction abilities of the neural model have been evaluated by comparison the real cold performance obtained during friction material testing on the single end full-scale inertia dynamometer and predicted ones.