SPECIAL TECHNIQUES – RUNDOWN MONITORING

Cracked rotor shafts have plagued nearly all turbinemanufacturers (and of course their users) to a greater or lesser extent over the years,due to the costly and disruptive problems they cause.Vibration monitoring via pedestal vibration or eccentricity is not appropriate as it is not good when it comes to detecting cracked shafts at the early stage required.In order for a crack to be detected by vibrationmonitoring alone the size of the crack will be approaching 20 to 30% and by this point the shaft will have to be scrapped.This late diagnosis can however be avoided so that future planning can take into account predictions of the shaft’s life-span.A more sophisticated technique than simply monitoringthe overall or frequency components of shaft vibration must be adopted to detect shaft cracks at such an early stage.When a crack appears in a shaft it affects the stiffness ofthat shaft.The stiffness of the shaft is dependant upon the width of the crack.When the crack is at the top,the weight of the rotor forces the crack to close and when the crack is at the bottom,the weight of the rotor causes it to open.Therefore during one revolution of a shaftwith a crack in it,the stiffness varies according to the crack’s position.The varying stiffness as the shaft rotates means that thedeflections are not proportional to the forces causing them.This varying stiffness/deflection causes harmonics of the running speed to be generated. When one of these harmonics coincides with a rotorcritical,as it is in the case of a run-up or run-down,the vibration response at that frequency changes.A run-down spectra of a cracked shaft will therefore show a change inresponse at a frequency corresponding to one of theknown critical speeds when the rotor is actually spinning at one-half,one-third etc of that critical speed.This allows a crack to be spotted when it is a fraction ofthe size of a crack spotted using vibration monitoring alone and therefore long before the shaft has to be replaced.Trending of vibration data is an extremelyvaluable tool for machine health monitoring.Not onlydoes it indicate problems but it enables estimates to be made as to when the problem will become serious enough for the machine to be taken out of service.Trending of vibration data however requires the machinetrend data to be taken when machine conditions are identical.The diagrams below show that even when there are relatively small changes in the turbine load,the vibration spectra are markedly different.Trending also becomes difficult when the vibrationsignature of the machine is constantly changing,due not only to generator load but a number of factors including steam conditions.Conditions between a number of run- downs do however remain constant so comparison between rundown signatures provides valid trending data.
Typical run-down plot of vibrationRun-down plot from a turbine with cracked rotor spectum from steam turbine at two different loads

22