Smartphone application for fault recognition

Abstract

Smart-phones have become an integral part of our daily life. Even though smart-phone app development has boomed, only few general purpose applications exist for feature extraction, feature selection and classification of audio data from a smart-phone. The paper presents detailed theory behind the data mining model used, which has given good results on MATLAB. The application was made to learn different fault states of an industrial air compressor. The application was tested to recognize the fault state in real time as the air compressor was running. It has performed very well with classification accuracies above 93.73%. It is believed that similar application and model with some minor changes in specifications can be used for acoustic pattern recognition in wide range of fields; even in industry.