Challenge
Brake pad development involves selecting and combining material compositions and processing methods, followed by prototype fabrication and dynamometer testing to evaluate friction characteristics. The process typically requires hundreds of tests and relies heavily on the experience and intuition of researchers—resulting in high costs and long development cycles. In addition, narrowing the technology gap with leading competitors requires further optimization of the R&D process.
Approach
An AI model was trained to predict the friction characteristics of brake pads using preprocessed historical test data. By forecasting the outcomes of planned dynamometer tests, the model reduces the number of physical tests needed and accelerates the development process. To ensure usability for non-AI specialists, intuitive tools and interfaces were introduced to make prediction results actionable in real-world R&D workflows.
Value Delivered
A predictive model was developed to meet the international performance standards set by the International Union of Railways (UIC), with accuracy maintained within an acceptable margin of error. Explainable AI (XAI) was applied to enhance the reliability of the results, and additional metrics were introduced to evaluate prediction uncertainty when working with new data. A user-friendly dashboard was also implemented, allowing researchers to run prediction tests, compare results against UIC standards, and make more informed development decisions.
