Explainable machine learning and multi-objective optimization of rubber-based brake friction materials using Krauss test machines through a computational framework
Year: 2026
Type: article
Abstract: Rubber-based brake friction materials are evaluated using friction, wear, thermal behavior, and friction stability responses. This study uses, an L27 Taguchi design is performed with respect to five types of rubbers (NR, NBR, SBR, EPDM, and MSR) and three operating factors such as pressure, sliding speed, and interface temperature. Eighty-one experimental data points are obtained based on physical Krauss fade-recovery tests, including three replicates for each design point. Six responses are studied: coefficient of friction, friction stability, fade, recovery, specific wear rate, and maximum interface temperature. A total of 1119 artificial observations were created using the constrained response surface approach, yielding a 1200 data points dataset for the machine learning approach. Comparison of linear regression, random forest, support vector regression, gradient boosting, XGBoost, and ANN methods are done using random and grouped validation. SHAP analysis is performed on the estimated models, while NSGA-II investigates the balance between friction, stability, fade, recovery, wear, and temperature. The 81 observations constitute experimental data, while the remaining 1119 observations are computationally based.
Source: Next Materials
Institution: Vellore Institute of Technology University
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,280
OpenAlex ID: W7235068897
