Using Torsional Vibration Spectra to Monitor Machinery Rotor Integrity

Abstract

Machine degradation has become a key issue with respect to the operational and maintenance costs associated with industrial and power generation facilities. Current online techniques for monitoring rotor integrity are largely based on lateral overall vibration levels that may provide only a very short notice of impending failure. As an alternative, shifts in rotor torsional natural frequencies could be used as early indicators. Torsional vibration spectra have been gathered on numerous horizontal hydro turbine generator shafts at two Southern Company owned hydro plants. The data was trended for approximately 2 years and changes were compared against findings from visual and nondestructive testing. It was determined that in the very early stages of failure the torsional frequency shifts are minute and may be masked by or be indistinguishable from other phenomena but are detectable. As the degradation progresses, the frequencies shifts may increase greatly with the crack size and are easily discerned. While the degree of early warning capability based on this technique will more than likely vary with each failure occurrence, it should generally outperform existing lateral vibration based techniques.