A deeper understanding of bearing axial cracking enables reliability assessments for individual turbines
The gearbox of a wind turbine is responsible for converting the relatively slow rotations of a turbine’s blades into the high speeds needed to generate electricity. These hard-working components often do not reach their expected 20-year lifetime, despite meeting industry standards, because of a failure mode called axial or “white-etch” cracking in the rolling-element bearings inside the gearbox.
Given that gearbox bearing failures lead to costly unplanned maintenance and turbine downtime, there is an opportunity to reduce operation and maintenance costs at wind power plants by accurately predicting the probability of failure and implementing solutions to reduce the frequency of these failures.
“This issue is not unique to the wind industry,” explained Jonathan Keller, a senior engineer at the National Renewable Energy Laboratory (NREL) and principal investigator of the Drivetrain Reliability Collaborative. “This failure happens in other industries, including automotive, rail, even washing machine bearings. However, when it occurs in a gearbox weighing 15 tons or more and suspended 250 feet up in the air, the cost implications are greater than, say, your car, which you can drive to a shop.”
As wind turbines increase in size and capacity, gearbox failures are expected to continue being a problem for wind power plant operators unless bearing axial cracking can be reproduced in the laboratory, computationally modeled, and compared with actual power plant results. Funding from the Wind Energy Technologies Office (WETO) has enabled Keller’s team at NREL to join forces with Argonne National Laboratory (Argonne) and industry partners such as SKF, Winergy, General Electric, and Afton Chemical to undertake these challenges.
Argonne was able to reproduce the issue in the lab using a benchtop testing rig, which literally sits on top of a bench. About the size of an oven, Argonne’s accelerated benchtop testing setup uses material specimens and oils to help determine the parameters that cause steel bearings to develop axial cracks. The Argonne team, led by Aaron Greco, examined the sliding that takes place between the bearing roller and the surface on the bearing ring that the roller rolls upon, called the raceway.
“We determined that, instead of a roller rolling, it sometimes slides across the two metal surfaces,” said Greco. “These slipping or skidding conditions resulted in axial cracking, allowing us to confirm a causal link. Using the benchtop testing approach, we have also determined that certain chemical additives in the lubricant, and separately, the presence of electrical current across the bearing surface, can also lead to this axial cracking failure.”
Next, NREL researchers outfitted a 1.5-megawatt wind turbine at NREL's Flatirons Campus with tailored instrumentation to gather experimental data at scale. Their findings, published in the journal Tribology International, confirmed that bearing slip occurs during wind turbine operations as a result of factors including bearing design, load, speed, lubrication, and temperature. Furthermore, NREL researchers used the data to develop a probability of failure model that fills an industry gap in evaluating component reliability, and a roller sliding model that is scalable to different turbine and gearbox platforms.
“These models leverage the benefits from both physics and data domains, with the former trying to capture the underlying mechanism for this investigated failure mode, and the latter trying to fine-tune based on individual turbine operational conditions and their impacts on component reliability progression,” said Shawn Sheng, a senior engineer at NREL.
Because these experimental setups only partially represent conditions at commercial wind facilities, the researchers are currently working to compare their models with actual wind power plant failure data from more wind power plants. The researchers are working to validate their models against failure statistics and operational data from a wind power plant operating about 100 turbines over 10 years.
“It’s a great start, but we need to work with many more wind power plants to access their data,” said Keller. “With this hindcasting approach, we can see if our model is a good predictor of the actual failure numbers that have occurred in the past. That comparison will then tell us whether we have not just a contributing factor, but the primary factor for causing axial cracking.”
This combined research effort has yielded a reliability assessment and prognosis methodology for bearing axial cracking in individual wind turbines, comprising both the roller sliding model and the probability of failure model with data domain inputs. These findings, which connect the reliability forecast with bearing design, controller settings, and turbine operations, were recently published in the journal Renewable & Sustainable Energy Reviews. However, it remains difficult to implement effective mitigation techniques until the definitive cause—or causes—of bearing axial cracks in the field can be determined.
“We would love for more industry representatives to reach out to us so we can pinpoint exactly what’s causing gearbox failures at wind facilities,” said Keller. “Our goal is to help the entire wind industry reduce maintenance downtime and increase energy production.”