|
|
|
 |
| BACK |
|
| Gearbox Failures – A Historical Review |
|
Gearboxes failure rates for wind turbines have traditionally been well above expectations basedon other industries. In recent years some industry experts have suggested that the situation isimproving, but the evidence does not always support this. For example, consider the quarterlyfailure statistics from the German wind power industry published in “Wind Stats Newsletter”. Bythe criterion of unplanned downtime, gearboxes are still usually the worst offenders among thefault categories identified: 6513 out of 12815 hours in the Winter 2007 issue, for example.
Many possible causes have been advanced for these high failure rates, relating to the design,manufacture and operation of wind turbines. No evidence has emerged that any one causeaccounts for more than a restricted group of failures, although a rule of thumb is that thebearings rather than the teeth are where the problems initially manifest themselves. Indeed,over the years the focus has moved across different features of the gearbox. In the 1990’s,planetary bearings seemed to be the Achilles’ heel; at other times the bearings on the highspeed shaft have posed more of a problem. Concerns have been raised about the bearing typeselection, the bearing life calculations, the lubrication, bearing support rigidity, the loadspecification, the testing regime, and so on. Eventually, the industry has developed aconsensus that improvements must be sought on every front.
Possible Causes of Failure at the Design StageOne aspect of design in which there is scope for significant improvement is the simulation ofdrive train dynamics. A widely used traditional method of representing drive trains in windturbine simulations is simply as a single spring representing the combined torsional stiffness ofall the elements between the rotor and the generator. The justification advanced for this is thatthe referred inertias of the turbine and generator rotors are so much greater than those of theintervening drive train components that natural frequencies involving the latter are entirelyseparated from the rotor mode frequencies. While the substance of this argument isundoubtedly true, there are a variety of complications which it leaves out of account.
One problem is that the combined flexibility is itself difficult to estimate with any accuracy.
Measured flexibilities are typically substantially higher than theoretical estimates based onsimple hand calculations. Then there are non-linearities in the response, arising from backlashand contact mechanics. Gearbox casing deformations may exacerbate both of thesecomplications. Finally, transient loading events, such those associated with braking and faultconditions, have a much higher frequency content to excite the higher modes of the gearbox internals.
It may well be that the long term duty cycle expressed in conventional terms is not greatlyaffected by the higher torsional modes. However it is now becoming normal for gearbox manufacturers to require more than standard duty cycle information, since it appears thatbearing durability can be reduced by unusual loading patterns that do not directly affect theequivalent load calculation but may, for example, give rise to a degree of skidding. Theadditional information required is in the form of complete time histories of significant events,such as transient load events. These events cannot be fully described with only a singleflexibility representing the drive train.
|
|
|
