Predictive Maintenance of Brakepad Degradation in Automotives

Abstract

Efficient health of brake assessment is critical for maintaining its performance and ensuring functional safety in automotive systems. The condition monitoring of a brake pad in an automotive plays a significant role in ensuring vehicle and chauffeur safety with effective maintenance. Conventional brake monitoring approaches rely more upon reactive approaches to brake pad wear, which increases system complexity, introduces latency and overall safety, and may not always be practical in real time scenarios. This paper proposes an automotive framework that is modular and includes predictive maintenance and data driven approach to envisage the degradation of the brake pad and state the condition of the brake either as normal or abnormal. In this framework, multiple classification models that include Random Forest, Support Vector Machine, and eXtreme Gradient Boost have been evaluated and compared to derive the degradation state of the brake using the commonly available vehicle operational features such as odometer reading, brake pedal position, accelerator position, gear position, engine speed, and fuel consumption.