Remaining useful life (RUL) prediction of internal combustion engine timing belt based on vibration signals and artificial neural network

Timing belt rupture, which can develop quickly and cause severe harm to various engine components, usually occurs unexpectedly and without prior warning signs. Due to the rapid occurrence of timing belt rupture, fault diagnosis strategies are not indeed efficient. In this paper, a vibration-based intelligent method has been proposed to predict the remaining useful life (RUL) of the timing belt using data mining techniques and multi-layer perceptron neural network (MLP-NN). To achieve this goal, two categories of experimental tests were designed and carried out on the timing belt, namely, fault thresholding (FT) and accelerated life (AL) tests. FT test was performed by comparing the defect-free belt vibration signals with those of a faulty belt to determine the failure threshold. This is while, in the AL test, the engine was continuously run with a defect-free timing belt until initiation of rupture or damage was detected. Four feature functions, regarded as timing belt health indicators, were applied to the collected vibration signals, namely, energy, impulse factor, root mean square, and FM4. The extracted features were then fed to the MLP-NN to predict the timing belt RUL by continuously comparing the feature values and the failure threshold. The results showed that the proposed method is able to predict the timing belt RUL with an accuracy of 90%. In addition, data predicted by the MLP-NN showed high correlation with the actual measured data, which emphasizes the robustness and precision of the proposed method for timing belt RUL prediction.