DEVELOPMENT AND TESTING OF A TIMING BELT FAILURE DETECTION SYSTEM BASED ON VIBRATION SIGNALS IN INTERNAL COMBUSTION ENGINE

Abstract

This study presents the development of a vibration-based health monitoring system for internal combustion engine timing belts using advanced signal processing, statistical feature extraction, and artificial neural network (ANN) classification. Vibration signals were captured under six belt conditions one healthy and five faulty (Tooth Crack, Back Crack, Wear, Separated Tooth, and Oil Pollution)—and analyzed in the time, frequency, and time–frequency domains. Six statistical features (standard deviation, kurtosis, FM4, impulse factor, crest factor, and skewness) were extracted from each signal to characterize fault behaviour. The classification task was modelled as a multi-class problem, and results were fused using Dempster–Shafer Theory (DST) to improve diagnostic reliability. Analysis revealed that Separated Tooth and Oil Pollution were the most severe faults, exhibiting the highest deviations across all domains, while Wear showed signs of progressive degradation. Tooth Crack and Back Crack displayed moderate but detectable anomalies, and the healthy belt provided a stable baseline across all features. The DST-based classifier achieved high accuracy ranging from 96.90% to 99.34%, confirming the effectiveness of the proposed method. This research demonstrates that vibration signal analysis combined with soft computing techniques offers a robust solution for early fault detection and predictive maintenance of timing belts, potentially preventing catastrophic engine failure.