Improved signal preprocessing techniques for machine fault diagnosis

Machine Fault Diagnosis and condition monitoring using Acoustic Emission and Vibration Signature is an active research area of much industrial importance. Pre-Processing is an important stage after data acquisition. In this paper we have presented a preprocessing scheme which includes a filter, a smoothing algorithm, a novel segmentation technique and a normalization algorithm which is less affected by the presence of outliers. Proposed segmentation approach chooses segments suited for classifications algorithms. Experiments with real time data from an air compressor have shown promising results.