WAVELETS AND FAILURE PREDICTION

NEW TECHNIQUES HAVE BEEN DEVELOPED OVER THE PAST 36 MONTHS THAT OFFER SIGINIFICANT POTENTIAL FOR ON-BOARD MONITORING OF MECHANICAL SYSTEMS TO DETECT AND CLASSIFY INCIPENT FAILURES. MANY ARE BASED ON THE WAVELET TRANSFORM, AND ITS EXTENSION TO THE WAVE PACKETS TRANSFORM. ALL PROVIDE A MEANS TO ISOLATE CHANGES IN SIGNAL STRUCTURE IN TERMS OF SCALE AS WELL AS IN TERMS OF TIME. THIS EFFORT PROPOSES TWO COMPREHENSIVE APPROACHES TO INCIPIENT FAILURE DETECTION AND CLASSIFICATION. THE FRONT-END OF OUR PROPOSED CLASSIFIER EXTRACTS NOVEL FEATURE SETS OFFERED BY THE WAVELET TRANSFORM AND THE MULTIRESOLUTION TIME SERIES SPECTRAL ESTIMATORS. ONE APPROACH USES THE STREAMS OF FEATURES DETECTED IN EACH TRANSFORM CELL AS AN INPUT TO RECENTLY-DEVELOPED AND SUCCESSFULLY APPLIED TECHNIQUES FOR THE DETECTION OF CHANGES IN MODAL AND SPECTRAL CHARACTERISTICS INDICATIVE OF FAULTS IN VIBRATING SYSTEMS. THE OTHER USES THESE OUTPUTS TO DEFINE A FEATURE SPACE ON WHICH A CLASSICAL NEURAL NET CLASSIFIER MAY BE TRAINED AND EVALUATED. ANTICIPATED BENEFITS/POTENTIAL APPLICATIONS - RESULTS OF THIS EFFORT WILL DETERMINE THE EXTENT TO WHICH WAVELET TRANSFORMS AND MULTIRESOLUTION TIME SERIES ANALYSIS CAN CONTRIBUTE TO THE EARLY DETECTION AND CLASSIFICATION OF INCIPIENT FAILURES IN A VARIETY OF MECHANICAL SYSTEMS.