Fault Detection and Classification in Automobile Engine Based on Its Audio Signature Using Support Vector Machine

Numerous attempts have been made in recent years for detecting various faults in an automobile engine. The aim of developing this technique is to detecting the faulty engine more accurately and reducing the cost of risk for an automobile industry. These developed techniques have potential to detect and classify the type of fault automatically. Some of reported and developed techniques are unsupervised and some work on supervised learning. Previously reported techniques are able to produce good results with complex algorithms. In the present paper, we have developed a simple algorithm for classifying the healthy and faulty engines and detecting the type of fault if there is any. Input signal is an acoustic signal from faulty/healthy automobile engine for a duration of 11 s. Total eleven statistical features of the audio signature has been extracted and fed to the classifier. Classification and detection algorithm is made by sequential channel of support vector machine (SVM). Total four number faults have been tested on a real time collected data. Various combination of the SVM (with different types of kernel functions), artificial neural network and logistic regression have been tested and a maximum accuracy up to 99.26% is obtained, which is greater than all the previously reported algorithms on the same data. The novelty of the algorithm is its less complexity, computed features and ability to produce better results.