This content is not included in your SAE MOBILUS subscription, or you are not logged in.
A Machine Learning Approach in the Design of Friction Materials for Automotive Applications: Correlation among Composition, Process Parameters and Functional Characteristics
ISSN: 0148-7191, e-ISSN: 2688-3627
Published October 08, 2006 by SAE International in United States
Annotation ability available
Brake pads to be used in the automotive sector are complex mixtures composed of several raw materials of different types (metals, resins, abrasives, lubricants, rubbers…) which undergo mechanical and thermal treatment during their fabrication. Because of the large number of variables affecting the properties of these materials a customization is needed for each specific application, which increases the cost of the final product. In this study a machine learning approach based on regression trees theory has been adopted aiming to support the formulation of friction materials. A model has been constructed to correlate material composition, fabrication and application parameters to the tribological behavior of brake pads tested on dyno benches. The validation of the model has been carried out with reference to a dataset composed of 1000 patterns involving 130 input variables and using as target output the mean friction coefficient measured during standard AK-Master test. A Leave-One-Out (LOO) statistic has been used to evaluate the effectiveness of the approach in predicting the tribological properties of new materials. The resulting average error is in the range 5-10% which is of the same order of the dyno tests' intrinsic deviation. The trained model has been applied to the problem of simplifying the composition of a friction material for the after-market; the results were satisfactory and allowed a reduction in the number of dyno tests needed to reach the target.
CitationBusso, M., Portesani, A., Regis, P., and Buonfico, P., "A Machine Learning Approach in the Design of Friction Materials for Automotive Applications: Correlation among Composition, Process Parameters and Functional Characteristics," SAE Technical Paper 2006-01-3201, 2006, https://doi.org/10.4271/2006-01-3201.
- Friedman J.H. Popescu B.E. “Predictive Learning via Rule Ensembles” ( 2005 ) Stanford University, Department of Statistics , Technical Report
- Breiman L. “Bagging Predictors” ( 1996 ) Machine Learning 26 123 140
- Breiman L. “Random Forest” ( 2001 ) Machine Learning 45 5 32
- Freund Y. Shapire R.E. “Experiments with a New Boosting Algorithm” ( 1996 ) Machine Learning: Proceeding of the Thirteenth International Conference Morgan Kauffman San Francisco 148 156
- Friedman J.H. “Greedy Function Approximation: a Gradient Boosting Machine” ( 2001 ) Annals of Statistic 29 1189 1232
- Breiman L. Friedman J.H. Olshen R. Stone C. “Classification and Regression Trees” ( 1983 ) Wadsworth
- Friedman J.H. Popescu B.E. “Importance Sampled Learning Ensembles” ( 2003 ) Stanford University, Department of Statistics , Technical Report
- Friedman J.H. Popescu B.E. “Gradient Directed Regularization for Linear Regression and Classification” ( 2004 ) Stanford University, Department of Statistics , Technical Report