Frequency based Audio Classification for Preventive Maintenance in Automobile Engines

Preventive maintenance is carried out in order to prevent sudden breakdowns and to enhance the life of the engine. This paper presents a method of preventive maintenance where the audio data from the parts of the defective automobile engine is taken and compared with the normal automobile engine. The data of the ball bearing from a Royal Enfield Classic 350 engine has been used. Using the methods of feature extraction, the data is processed and visual representations of each data are obtained and analysed with CNN classifiers. The performance of the proposed method is evaluated and compared with state-of-the-art classifiers such as Alexnet and VGG16. The experimental results show that the proposed system classifies the audio data with 75% of accuracy. Upon considering the significance of the proposed method, it can be employed to identify the anomalies in the audio data for preventive maintenance.