Classification of conditions of rotating machines using higher order statistics

In this paper three approaches are outlined to classify conditions of rotating machines using higher order statistics. Horizontal and vertical accelerometer vibration data have been collected from a small rotating machine set in four different conditions at different rotational speeds. The three methods are higher order statistics based classification, artificial neural nets based classification, and higher order spectra based classification. Preliminary results from these approaches indicate that their success rates are approximately 90%. Further studies are under way for better understanding and performance