Pattern Recognition based on Probabilistic Neural Network for Motorcycle Oil Fault Diagnosis

In the presented era, diagnosis of different types of faults that arise in vehicle engines especially in automotive of motorcycle is very critical in subject of repair and maintenance. Therefore, timely and accurate diagnosis of motorcycle lubricant oil fault is the key to solve the problem. In this paper, motorcycle oil fault diagnosis is proposed through Empirical Mode Decomposition (EMD) as feature extraction and Probabilistic Neural Network (PNN) as pattern recognition. EMD such a Fourier transform and wavelet decomposition is used to break the signal than the selected feature is used as input into the PNN. PNN's pattern recognition is a classification method of a self-supervised and feedforward networks. Proposed scheme is evaluated on Honda Beat motorcycle which use AHM MPX2 lubricant oil based on LabView. Result shows that EMD decomposed vibration signal into its Intrinsic Mode Function (IMF) successfully, then become the input into PNN. Result has proven that PNN is quick learner, robust and has 100% classifying accuracy.