Abstract
Key words: neural modelling, automotive brake, brake performance
The automotive brake´s performance results from the complex interrelated phenomena occurring in the contact of the friction pair. These complex braking phenomena are mostly affected by the physicochemical properties of friction materials´ ingredients and brake´s operation regimes. Analytical models of brakes performance are difficult, even impossible to obtain due to complex and highly nonlinear phenomena involved during braking. That is why, in this paper all relevant influences on the brake performance have been integrated by means of artificial neural networks. The influences of 26 input parameters, defined by the friction material composition (18 ingredients), manufacturing conditions (5 parameters), and brake´s operation conditions (3 parameters) have been modelled versus changes of the brake factor C. The neural models of the cold brake, hot brake, and brake recovery performance have been developed in this paper. These models have been developed by testing of 90 different neural models obtained by training of 18 different architectures of neural networks with the five learning algorithms.