A novel robust automated FFT-based segmentation and features selection algorithm for acoustic emission condition based monitoring systems

This paper aims at developing a robust, fast-response and automated FFT-based features selection algorithm for the development of acoustic emission practical condition based monitoring applications of mechanical systems. Further scope of this work is to investigate the suitability of acoustic emission for the fault diagnostic of high speed centrifugal equipment using a single AE sensor. Experiments were conducted using an industrial air blower system with a rotational speed of 15,650 RPM. Five experiments for five different machine conditions were carried out. Ten data sets were collected for each machine condition with a total number of 50 data sets. Fifty percent of the data sets were used for training and the remaining data sets were used for verification. Tailor made programs for spectral features selection and for classification of faults were developed using Maltab to implement the proposed algorithm to an industrial air blower system. The results showed the suitability of the acoustic emission spectral features technique for the fault diagnostic of centrifugal equipment and proved the effectiveness and competitiveness of the proposed automated features selection algorithm. The sets of features selected by the algorithm yielded a detection accuracy of 100%.