“uåX˜äŠÊ˜·³Ç

Wind turbine gearbox failures are not generally a product of poor quality components, but of weaknesses in the design process caused by a dearth of data on just what goes on in the inner workings of a wind turbine. This was the broad message from experts on gearbox and gearbox bearing failure who shared their views with delegates at the American Wind Energy Association's Los Angeles conference. If the panel of experts got no closer to establishing the causes of gearbox failure, it did agree that trouble nearly always starts with the bearings, though bearings are not necessarily the root cause. To find out what is, requires an industry-wide collaborative effort focussed on data building and analysis, the panel urged.

Reporting from the early stages of a $2.5 million project at the US National Energy Research Laboratory (NREL) in Colorado aimed at doing just that, Brian McNiff of McNiff Light Industry said gearboxes in wind turbines are not living up to their design lives. Indeed, 19.4% of all plant downtime is due to gearbox problems said Simon Roberts from Romar Technology, a company specialising in helicopters. "We think these problems are non-manufacturer specific," said McNiff. "With what we know now we can introduce some improvements, but we cannot make the problem go away," he added. "We can get rid of the nuisance problems."

Among reasons McNiff gave for gearbox failure were irregular or unexpected bearing responses, excessive flexibility of the gearbox mountings, and/or inadequate margins in some of the gearbox components. Standard fatigue analysis could not explain the failures. Roberts criticised the use of "simplified approximations" in existing fatigue load models, while from bearing supplier Timken, Michael Kotzalas explained: "There are several methods of calculating fatigue loads on bearings -- all giving different results."

The three-stage NREL program will start with drive-train analysis and modelling, followed by laboratory testing of two different optimised gearboxes using a testbed able to feed up to 2.5 MW of power into the drive-train, and conclude with field tests of a sub megawatt wind turbine at the Ponnequin wind farm, also in Colorado.

The test bed can run at up to 146 rpm, enabling a few months of testing to simulate 30 years of operation. The effects of turbulence and high transient loads can also be included. Researchers will gather data during simulation of loading events to replicate what happens in wind turbine operation. In the field, load events will be correlated with the gear/bearing response, said McNiff. "What is it doing when it slows down? Look at what the inertia is doing."

Roberts supports the concept of a full scale test bed and collaborative work between all component suppliers. In current modelling methods there is too much subjective input, he warned. In the wind industry, forces and deflections should be calculated, not just assumed, Roberts said. "It's all a matter of understanding the loads. There is not enough empirical data in the wind industry," concluded Roberts. McNiff said it is necessary to "get down and describe the full complexity of the loads" and to find out the load distribution and "make what is going on inside the gearbox clearer."

Christopher Walford from GEC, working on a Department of Energy (DOE) project under the DOE WindPACT program, said that detecting what happens is complicated by the low speed rotation of a wind turbine. WindPACT is testing a low speed gearbox in a 1.5 MW drive train made in the laboratory with a new CARB planet bearing from SKF which allows for axial movement and misalignment. The research group is studying "what is actually going on inside there with the bearings," Walford said.