The failures were traced to fine white cracks on the surfaces of the bearings, Kadiric explained in January in a presentation to the 21st International Colloquium Tribology. Such cracks were first cited as a source of failure in steel bearings and high-speed tools in 1996, and the problem since has spread-but why? Recent literature proposes various causes, yet no consensus has been reached yet regarding the main root cause, he remarked.

Theres no lack of theories, Kadiric went on. One leading contender is contact fatigue due to inclusions in the bearing material. Another (which he dismissed as unlikely) may be lubricant composition that leads to brittle fracture. Direction and magnitude of the sliding contact may have some effect, and stray electrical currents are also under scrutiny. The effect of hydrogen has been probed, but is notoriously difficult, Kadiric admitted. Anyway, the point is that theres no definite agreement. And absent that, the link to actual bearing life remains elusive.

Could lubricants and their additives be propagating WEC? Maybe, say the researchers chipping away at the problem. But lubrication is not the sole source of the mayhem. Christoph Mayer of Kluber Lubrication Munchen cited the quality of the bearing material, mounting errors, stress and load dynamics. Other evidence points to environmental conditions like water ingress, hydrogen and contamination, he said during the conference. And the trail is strewn with false leads, cautioned Mihails Birjukovs of the University of Latvia, including a flawed theory on how thermal stress from stray electrical currents may cause WEC.

Kadiric and his colleagues Francesco Manieri and Pawel Rycerz set out to uncover the mechanism behind WEC. We wanted to establish if white etching cracking is a cause or a consequence of bearing failure, he told the gathering, at Germanys Technische Akademie Esslingen. We asked what is the relationship to premature rolling contact fatigue failures, and finally, we wanted to know if we can generate cracks away from the contact, so we can observe it.

In its investigations, Imperial College used a micro-pitting rig from PCS Instruments, which allows users to apply various contact geometries; this is very important to the failure mode being studied, Kadiric said. We used a flat disk on chamfered roller.

To weigh the possible influence of the lubricant on WEC, the researchers used AISI 52100 bearing steel specimens and monitored crack-propagation rates. The study compared four oils: one formulated SAE 75W-80 reference oil (transmission fluid) known to perform poorly; one bad reference engine oil; a blend of polyalphaolefin and zinc dialkyl dithiophosphate antiwear additive; and an API Group I base oil fortified with ZDDP. The test conditions are so severe that you have to have some ZDDP in the formulation or you cant run the test; you need the wear protection, Kadiric observed. As it turned out, he added, we obtained white etching cracks very easily-it was a struggle not to form them.

The results underscored the significance of both sliding direction and magnitude of sliding; and while they suggested that negative sliding plays a role in crack propagation, that didnt seem like a primary factor in how the crack begins. Instead, the authors theorize, WEC is formed by the stress profile. High stresses at the start of the test-with the highest stresses imposed early on-seemed to initiate cracks early.

Other factors are also important, such as the presence of hydrogen or stray electrical currents, which can weaken the metal, but they are not necessary to the forming of white etching cracks, Kadiric said. Rather, he saw powerful evidence that the sequence of stresses over the time of the test, the load history, could lead to failure.

Mayer, a research tribologist at Kluber, characterized WEC as a nano-crystalline area shortly below the running track surface, transected by cracks. It occurs, he said, with both oil and grease lubricated bearings. Failure occurs at just 5 to 10 percent of the calculated bearing life. In wind turbines, this can happen in the first year of operation, and in 15 percent of cases it crops up within the first five years, his presentation noted.

Kluber wanted to see what effect base oils and additives may have on WEC, using a variety of oils: five model products formulated in the lab to meet ISO viscosity grade 100, plus two ISO VG 68 commercial gear oils.

All the samples underwent an in-house axial cylindrical roller bearing test, designed around the standard FE-8 bearing test rig. At any sign of pitting formation, the test was terminated-a fail. Products earned a passing grade only if the planned number of load cycles was reached without damage (see table above). Then we did an analysis of all results, with and without pitting, Mayer continued. We also analyzed the tribo­layers formed at the bearing washer.

Tests on the first two model formulations, a mineral oil and a full-synthetic sharing the same additive package, showed severe and early pitting damage and were aborted. From this, Mayer said, we learned that if you have a WEC-sensitive additive package, it will happen regardless of base oil type.

The next two model products used the same mineral and synthetic base oils but swapped in a second additive package; and both got passing results. So the additive packaging can influence WEC formation, Mayer asserted.

The fifth model product, using the same mineral oil and a third additive package, passed the desired number of load cycles but flunked on closer scrutiny. There was no failure, but metallographic investigation did show small, early-stage WEC. So even if no surface pitting is observed, this shows that cracks can be present.

As for the two commercial ISO VG 68 products, one (a semi-synthetic gear oil) had its test aborted due to pitting formation. The second (a full synthetic from Kluber) passed all the load stages and gave no sign of WEC, Mayer stated.

Following each test, the bearing washers were checked by metallographic analysis for WEC wear fatigue, and the depth of the boundary tribolayers formed by the additives on the surfaces was checked by time-of-flight secondary ion mass spectrometry. (ToF-SIMS measures the signal intensivity of select secondary ions, indicating the thickness of the tribolayer.)

From the assembled data, Mayer and his colleagues realized that the specimens showing the thickest tribolayers (from additive package A) always failed through WEC. When the tribolayer was of medium thickness, as with additive packages B and C, WEC formation occurred more rarely. And washers exhibiting thin tribolayers didnt exhibit WEC at all. Echoing earlier research, this suggests that a highly reactive, surface-active boundary layer, constantly regenerating itself, is actually a poor defense against WEC.

Mayer summarized: First, thick is not the same as good when it comes to tribolayers that build up on the metal surface. Second, what the base oil does is not unimportant. And finally, you do have to avoid exposing fresh metal surfaces, because they allow bad things to happen.

Mihails Birjukovs brought a different issue to the conferences attention: a faulty theory. A graduate student in physics at the University of Latvia, he is digging into the idea that electrical currents, whether from lightning, stray currents from generators or static discharges from lubricants, can lead to subsurface damage in metals.

Birjukovs curiosity was sparked by a theory published in 2015 in Tribology Letters by M. Scepanskis et al, that posited a numerical model for how electric discharges could build up in and affect metal. The authors suggested that thermal stress could accumulate around inclusions in the steel structure, intensifying and wreaking havoc. I created an analytical model to test that theory using Scepanskis parameters, Birjukovs said, and while he had predicted mega-Pascals of overheating and stress levels, I could only get Pascals-no order of magnitude.

In fact, he continued, the stored discharge associated with electromagnetic force induced stresses were found to be negligible. Perplexed, Birjukovs asked for and reviewed the data and model used in the earlier study-and found that the mathematical model had flaws, including exaggerated current densities.

Closer examination and his own modeling led Birjukovs to conclude that the results of the 2015 paper represent a complete misapplication of material thermodynamic properties. The defect sites that its authors spotted are likely products of numerical artifacts introduced by using a very coarse mesh in their model.

Bearing manufacturer Schaeffler also has found that discharges in the lubrication zone vanish very quickly, so Id say its all based on an unfounded assumption about material thermodynamic properties, he stated. (View Birjukovs analysis at www.ResearchGate.net/publication/
319587898.)

While Birjukovs felt apprehensive about bringing his findings forward, he was urged to do so by his academic advisor, Andris Jakovics, and Walter Holweger of Schaeffler Technologies. Eliminating the factors that are irrelevant to WEC, he said, can free researchers to concentrate on more fruitful lines of inquiry. The audience at the tribology colloquium agreed, and one listener enthused that this is how science advances, through negative as well as positive learnings.

Also at the colloquium, Ksenija Topolovec Miklozic of PowerTrib Ltd. in Oxford, U.K., noted that another roadblock to solving WEC is the cost and time needed to run FE-8 roller bearing tests. She suggested using screening tests to evaluate wind turbine gear oils-but are they valid?

PowerTrib has found that running an FAG screener test on the PCS micro-pitting rig correlates well with the longer, costlier FE-8. To generate wear and fatigue results, the full FE-8 test runs almost 900 hours at two temperatures, and then another 600 hours with water. By contrast, she said, the three stages of the FAG screener can be done in under 90 hours.

These are relatively short micro-pitting and wear tests, Miklozic continued, but in the real world, wind turbines operate under diverse, transient, non-steady conditions. So were able to combine different modules, in stages, to see those effects, too. For example, a test might run for one hour at low-speed, mid-load conditions, then 10 hours under high speed and mid load, then undergo a rapid ramp-up in sliding for an hour, and finally run at high speed and high load for 20 hours. Thats only about 30 hours in all-yet youre able to measure friction, vibration, acceleration signal, pitting, weight loss and temperature. You can really see how fluids differentiate themselves under those changing conditions.

Eventually the right tools, research and lubricants may put an end to WEC. As Kenred Stadler, a tribologist with bearing manufacturer SKF Group, insisted at a 2014 symposium, Bearings dont commit suicide!