Abstract
Vehicle brake systems are among the most critical safety components in modern automobiles, yet conventional maintenance strategies remain predominantly reactive or schedule-based, often failing to detect early-stage degradation. Continuous exposure to mechanical friction, thermal stress, hydraulic pressure variations and varying driving conditions accelerates brake component wear, making fault prediction a complex nonlinear problem. This study presents an artificial intelligence (AI)-based predictive maintenance framework specifically designed for vehicle brake systems to enable early fault detection and improve operational safety. The proposed framework utilizes multivariate time-series data acquired from brake-related sensors, including hydraulic pressure, temperature, vibration and actuation signals. A hybrid deep learning architecture is employed to capture both temporal dependencies and nonlinear degradation patterns associated with brake wear, pressure leakage, thermal abnormalities and abnormal braking behavior. The collected sensor data undergo preprocessing, normalization and anomaly detection to identify incipient faults before critical failures occur. The proposed model is evaluated using representative brake operating scenarios and demonstrates superior fault prediction performance compared with conventional maintenance approaches. Experimental results achieve a prediction accuracy of 97.01%, along with high precision, recall and early fault detection capability, enabling maintenance interventions several days before potential system failure. The framework reduces false alarms, improves maintenance planning, minimizes unexpected breakdowns and enhances vehicle reliability. The findings demonstrate that AI-driven predictive maintenance provides an effective and scalable solution for intelligent brake health monitoring and contributes to the development of safer, more reliable and nextgeneration connected vehicle systems.
Keywords: Anomaly Detection, Artificial Intelligence, Brake System, Machine Learning, Predictive Maintenance, Vehicle Safety.