Prediction of automotive friction material characteristics using artificial neural networks-cold performance

In this study, an artificial neural network technique was used to predict the cold performance of the automotive friction material. Cold performance was predicted for two cases: (i) before and (ii) after fading and recovery tests. Predictions were related to the brake factor C values versus 26 input parameters. The input parameters are defined by the friction material formulation (18 parameters), manufacturing conditions (5 parameters), and testing conditions (3 parameters). For these predictions, the five types of the friction materials were produced and tested. The quality of prediction has been evaluated by comparison of the real results obtained during testing on the single-end full-scale inertia dynamometer and predicted ones. The 15 different architectures of the artificial neural networks have been investigated. The five training algorithms have been employed for the artificial neural networks training.