Wind Turbine Bearing Failure: A Personal View

Abstract

The writer describes his experience of a lifetime of casting metals, and how the casting technique controls the quality of the metal and offers answers to engineering failures. In view of the wide denial of this aspect of process metallurgy, the author takes the opportunity to present a personal view, backed up by additional evidence in a bibliography. It is a concern that the failure of wind turbine bearings continues, on occasions, to defy substantial metallurgical efforts. It is proposed here that there is good reason to identify the casting process as the generator of pervasive defects, which the writer calls bifilms. These defects originate from the casting process during the pouring of the liquid steel. They are simply doubled-over oxide films originating from the surface of the melt. They are inherited by the solidified steel and are resistant to bonding by mechanical working. They, therefore, exist in finished steel components as a substantial population of cracks. These pre-existing cracks are usually the initiators of fatigue failure, as well as other failure modes. Techniques to eliminate bifilm cracks during the casting of steel are now known and require to be implemented to produce steels that will naturally eliminate failure. We shall have, for the first time, steels we can trust.

1. Introduction

Wind turbines are typically designed for a minimum 20-year life. However, failure of the main bearing of the turbine after only a few years, perhaps 5 years, can involve the immense expense of dismantling, lowering, transporting away for repair or replacement, raising, and re-installing the new bearing. These costs are enhanced for off-shore turbines and threaten the economic case for wind energy.

It is important therefore that bearings are reliable. The engineering involved in the modern bearing designs, optimised by computer simulation, and manufacture involving precision machining ensure an extremely high standard of ‘designed in’ reliability. The steel is also held to within close limits of its chemical specification, generally based on the composition 1C–1.5Cr, the typical ‘carbon chrome’ steel which is widely used for ball and roller bearings. The use of this bearing steel for over 70 years or more has generated an optimised material backed up by an immense volume of development and production experience.

It is a source of surprise and disappointment, therefore, despite all this vast accumulation of knowledge and experience, bearing failures of large turbines still occur prematurely.

A higher strength steel, with an unusual structure of lower bainite, has more recently been available; it is hoped that this improved material will provide greater reliability and longer life. Experience with its longevity should be emerging over the next few years, so it is too early to include the new bearings in this report. This chapter concentrates, therefore, on the known behaviour of the carbon-chrome steel bearings. Even so, it seems likely that the proposals in this chapter will also benefit the low bainitic steels in due course.

This chapter draws attention to the universally neglected role of the casting process in the behaviour of the steels. Unfortunately, defects are introduced by the casting process which can be sufficiently serious to dominate the failure mechanism of the steel. This widely overlooked effect is considered in detail. The mechanisms cited in this chapter are described in more detail by the author elsewhere [1, 2, 3].

2. The background to inclusion creation and the cracking of the liquid metal

In the liquid state, all metals oxidise to some degree in the air, forming a thin surface oxide film. This surface oxide is not a problem while on the surface, and is valuable, limiting the rate of further reaction with oxygen, conferring a kind of pseudo-inertness of the liquid, preventing the liquid completely oxidising away or even, some metals, bursting into flames.

The structure of the surface oxide on the liquid is significant. Its lower surface is completely ‘wetted’, being in atomic contact with the liquid, from which it has grown, atom by atom. In contrast, the top surface of the oxide is completely dry, and when viewed under the microscope has an irregular surface like sandpaper, or sometimes a microscopic downtown Manhattan.

When filling a mould, if the metal proceeds upwardly, the surface oxide trapped between the liquid and the mould wall cannot rise with the metal. Consequently, the oxide splits at the crown of the upward progressing meniscus, to allow the metal to rise. Instantly, new oxide forms at the newly revealed liquid surface. The newly forming oxide splits and moves sideways to become trapped as the skin of the casting against the mould. Importantly, the oxide does not become entrained in the matrix. In this way, the filling of the mould is achieved without the formation of defects, and the presence of the surface oxide, thickened during its travel across the meniscus and down the walls, protects the metal from the ingress of contaminants such as gases and other solutes by reaction with the mould, and mechanically supports the surface, bridging the gaps between asperities to confer a smooth ‘averaged’ cast surface. When the mould filling is completed only by this mechanism, it is known as Counter-Gravity casting (Figure 1). It is described in more detail below.

Figure 1.

Contrasting casting techniques.

However, as everyone knows, most castings are not made only by this upward counter-gravity progress of the liquid. Most castings are poured, using gravity, which accelerates the metal as it falls into the mould. The result is vigorous turbulence and mixing of the bulk liquid and its surface so that its surface oxide film becomes entrained into the bulk liquid metal.

The entrainment of the surface film happens by a number of interesting, related mechanisms. The surface can (i) fold, or (ii) form droplets and splashes, or (iii) form bubbles which during their motion through the liquid, slough off their surface oxide to form oxide tubes sometimes meters long, which collapse to form lengthy bifilms (Figure 1). Naturally, the bubble trails are generally shredded to shorter lengths in the violent turbulence, so that the final bifilms population consists of a fragmented collection of largely indistinguishable defects, although some bubble trails survive to form leak paths through 100 mm thick walls in shaped castings (if well-fed to avoid any shrinkage porosity, how otherwise could leaks through thick steel walls be explained?). In all cases, the mutual impingement of surfaces causes the oxides on the opposing surfaces to come together as a dry side to dry side. Only the high spots meet, so the films make little contact on a microscale, and little bonding can take place. This double film now takes on a life of its own. I call it a ‘bifilm’. Each bifilm finds itself now immersed in the bulk liquid, but has practically no bonding between its two opposed films. It, therefore, acts as a crack in the liquid. Very severe turbulence, as in the pouring of a steel ingot, results in a dense population of cracks in suspension, akin to a snowstorm [4].

For liquid steel, the snowstorm takes time to clear; although the oxides are less dense than the steel and therefore expected to float out, (i) they have nearly zero volume because of their extreme thinness and so can exert only minimal buoyancy force, plus (ii) their relatively large area is characterised by high drag—a parachute action to slow progress.

Thus, the liquid steel, now damaged by a semi-permanent population of cracks only slowly recovers its integrity: the larger bifilms separated by flotation within minutes, forming the observed layer of oxide slag on the surface of the steel. However, a large population of smaller bifilms remains to be trapped by solidification.

Because bifilms are not ‘clean’ cracks but faced with highly stable oxides such as alumina (Al₂O₃) and chromia (Cr₂O₃), the cracks tend to survive plastic working such as forging and rolling. After plastic working, most of our steels remain impaired by a dense population of cracks introduced by the casting process. This is particularly common during the pouring of ingots for special purposes, such as large bearing rings for wind turbines, because top pouring of the ingot is the cheapest casting technique. As will be explained below, even if the ingots are bottom gated (uphill poured) the current technology only improves the surface of the ingots a little but makes little change to the internal integrity.

For a particularly large ingot, required for the largest bearing rings, as the steel freezes in a direction away from the mould wall, the advancing dendrites ‘push’ the bifilms, tending to concentrate them in the centre of the ingot. When the ingot is pierced and opened by forging to form a ring, the bifilm defects are naturally concentrated on the inner working surface of the ring. Regrettably, the distribution of defects could not be worse.

This chapter reviews the damaging mechanisms of the casting process and proposes alternative casting techniques to reduce or avoid damage. It is noteworthy that the current carbon-chrome steel is usually capable of providing good service life of main bearings, and most probably suffers early failure from the presence of occasional material defects. It is expected that the implementation of processes to eliminate the larger bifilm cracks, the major defects in the material, should significantly assist to eliminate failure.

3. Steel casting processes

Bulk steels are now commonly cast by the continuous casting process. This is an efficient process in which a ladle of steel spends a significant length of time delivering metal from a bottom nozzle into a launder (a trough, or channel) which continuously supplies the water-cooled mould. The metal in the ladle, therefore, has a lengthy period in which its bifilm population can be reduced by upward flotation (despite huge research and development efforts, relatively little can be separated in the launders). Fortuitously, the metal is delivered to the mould from the base of the ladle, providing the cleanest metal. The result is that continuous cast steels are usually significantly cleaner and more fatigue resistant than ingot cast steels.

However, for limited batches of special steels, and especially those required in large pieces, ingot casting remains the only practical production route. The ingots are filled from a ladle which the crane positions above the mould. The metal is poured from the base of the ladle via a nozzle sealed with a stopper, or by a sliding gate.

A common casting technique for large ingots is top pouring, in which a ladle opens its bottom nozzle above the mould (Figure 1). The metal jets from the nozzle at speeds of the order of 5 m/s. The speed of the falling jet increases during the remainder of its fall into the ingot mould, reaching up to approximately 10 m/s. This high-energy liquid fragments and churns, introducing masses of bifilm cracks. It is a casting technique to be avoided for steel bearings.

The alternative technique for introducing the liquid metal at the base of the mould is widely used in the belief that it constitutes a kind of counter-gravity filling. It certainly improves the surface finish on the outside of the cast ingot because of the reduced amount of splashing, slopping and surging (Figure 2). However, it is not generally realised that the interior of the metal suffers a disaster. The reason for this is the universal use of a conical intake (the trumpet) at the entrance to the running system. The slightly ragged edges of the high-velocity falling jet of metal accelerate the closely surrounding air and take this down into the filling system. The conical entrance acts as a venturi pump, concentrating air into the system. A resulting 50/50 mix of air and steel is now known to be formed and is clearly observed in water model experiments and computer simulations. An observer looking down into the mould during the filling process observes the melt to ‘boil’ as it rises. Once again, the steel is severely degraded. Some restoration of the properties will occur by flotation before freezing, but an immense population of defects will remain and be trapped by the freezing process. A viewing of a water model of the filling process should convince even the most sceptical traditional steelmaker that the 50/50 emulsion of steel and air will be impossible to convert into a reasonably clean steel (Video 1, https://acrobat.adobe.com/link/track?uri=urn:aaid:scds:US:8ce23dc8-fbc5-4892-9934-439f732f3e57). We need to agree that this is not the way to treat liquid steel.

Figure 2.

Bottom-gated or uphill teemed ingot.

The problem is the conical intake which acts as an air pump. During the casting process, it is necessary to eliminate the massive ingestion of air at the entrance to the filling system.

Shrouds are widely used to reduce air entrainment by the trumpet. The trumpet is surrounded by an enclosure (the ‘shroud’) filled with argon. In this way, the percentage of oxygen from air entrained into the filling system is reduced from 20% down to perhaps 5%. However, when it is considered that pouring in a vacuum, in which the oxygen levels may be less than 0.01%, the amount of oxygen is clearly still massively over-sufficient to form very effective bifilm cracks throughout most steels. Thus, even though the use of shrouds reduces the content of oxide inclusion particles, and reduces oxygen in solution in the steel, the amount of turbulence is largely unaltered, so the area of bifilms is unchanged. However, of course, the bifilms are now thinner, so they are now much more difficult to see even though they continue to act as effective cracks.

It is necessary to conclude that the pouring of steel is the most damaging feature of the steelmaking process. It has to be improved if ultra-low defect steels are to be achieved.

4. Contact pour

The procedure which is practically 100% effective to reduce air entrainment during pouring is ‘contact pour’ in which the ladle nozzle is directly put into contact with the entrance to the filling system pipework. After the first few seconds of eliminating the air already in the filling system, the system runs clear of bubbles: entrained air is excluded (Video 2, https://acrobat.adobe.com/link/track?uri=urn:aaid:scds:US:4a0df8eb-5ec4-480a-b908-34a98165cef4). This simple solution has been demonstrated with success for those ladles with stopper/nozzle systems for steel castings over the size range of 1–50 metric tons. If contact pour can be implemented, no massive investment in new equipment would be required to revolutionise steel quality. There are additional helpful developments in which the few seconds of damaged metal and bubbles originating from the priming process can also be diverted from entering the ingot [1, 2]. Thus, it is possible for metal totally free from contamination with air to be cast into the ingot. The metal will be free from its normal population of cracks.

Even so, there are several issues that many steel casting shops would need to solve to implement the system successfully, and success may not be easy. Although the x-y positioning of the ladle is a challenge, this is not expected to be insuperable in view of modern laser triangulation and computer control of the crane. The z-control of the crane is more significant to avoid the hundred or more tons of ladle crushing the mould filling system. Most difficult is the need for some casting systems to use oxygen lancing to initiate pour. The latter may be a non-trivial requirement.

If contact pour cannot be implemented, other solutions to avoid the damage of pouring are needed. The known outstanding solution is described below.

5. Counter-gravity

Counter-gravity casting is the optimum solution. In Figure 1, the metal is contained in a vessel below the mould and is preferably treated to be of high quality. Zero entrainments of the surface oxide film occur if the process is carried out correctly. The process is widely used in the aluminium casting industry, where uphill displacement by differential pressure or pump is relatively easy and is widely used. It is not so common for the casting of steel. However, there are some notable success stories for smaller steel castings of up to 100 kg which have been routinely made in their millions with excellent freedom from inclusions. A certain amount of scaling up seems feasible, so that steel ingots of a ton or more might be possible. In particular, for large bearing rings, the ingot could be cast as a ring, saving the cost of reheating and forging.

The standard objections to the elimination of forging are the loss of so-called ‘densification’ and the possible loss of a so-called ‘favourable grain flow’, or texture. Neither of these conventional benefits of forging are to be expected to be required for an ingot without bifilms. The oxide cracks are not present to pin grain boundaries and so the alignment to develop texture which occurs in traditionally cast steels is not to be expected. Similarly, the oxide cracks are the usual initiators of porosity of various kinds, which once again will be absent, so the natural soundness of counter-gravity cast steel should be near-perfect and not improvable by plastic working. The counter-gravity cast structure is expected to have excellent, and substantially unimprovable, homogeneous properties which cannot be enhanced by forging.

6. Fatigue failure

Turning now to a practical example. Figure 3 shows an image of the microstructure of a main outer bearing ring of carbon-chrome steel which had failed by rolling contact fatigue [5]. It seems to have been generally assumed that the large array of cracks had been formed by fatigue. This is not true.

Figure 3.

(a) Array of bifilm cracks under the bearing surface; (b) inclusion with butterfly wings and adjacent white-etching cracks (courtesy ref. [5]).

The large array of cracks is a typical tangle of bifilms, introduced into the steel as a result of the turbulence during pouring of the originating ingot. This was probably top poured for the economy, but even if uphill teemed (bottom gated) the turbulence and air entrainment issues are immensely damaging and certainly capable of creating such extensive defects. The bifilm population in the solidifying ingot will tend to be segregated into the ingot centre because of the ‘pushing’ action of the advancing solidification front (advancing dendrites cannot grow through the ‘air layer’ in the bifilm). Probably, the solidified ingot is now forged, opening it into the shape of a ring. The inner working surface of the bearing will naturally contain the highest density of bifilm defects from the centre of the original ingot, typical of those seen in Figure 3.

The enlarge detail provided in Figure 3b shows a fractured inclusion together with light etching cracks and ‘wings’ on either side, as in a classical fatigue structure. The ‘fractured’ inclusion appears fractured because of its growth either side of a bifilm (it is worth emphasising that the ‘fracture’ of inclusions is not normally the result of stress, but of growth on bifilms). However, this diminutive region constitutes the real fatigue failure. One can imagine that among the massive bifilm array, of the order of millimetres in size, large blocks of metal will be stressed by the passing of the rollers, and the stress will be concentrated in those small remaining regions which connect the block to the main mass of the bearing. The gradual failure of these ligaments by fatigue will eventually release the block into the rollers, causing catastrophic failure. The size of the ‘butterfly wings’ is of the order of 10 μm—only 1% of the size of the pre-existing bifilm cracks, but, of course, necessary for releasing the final failure.

In summary, extensive pre-cracks (bifilms) provide major weakening of wind turbine bearings, but final failure is by the fatigue of microscopic ligaments in which the rolling stresses are concentrated. The ligaments may or may not contain inclusions. It seems that extensive bifilm pre-cracks and microscopic fatigue cracks may be expected to be common conditions for failure. Work on the newer bainitic steels [6] is expected to reduce the fatigue failures of wind turbine bearings which is, of course, welcome. However, the complete elimination of failures is only to be expected if casting techniques can be improved [3].

7. Conclusions

Any improvement to the casting technique will be of immense value to a wide range of designers and fabricators worldwide. Without the initiation sites for failure which bifilms provide, steels (and other metals and alloys) should become free from the normal modes of failure such as tensile fracture, creep, fatigue, stress corrosion cracking and hydrogen embrittlement [3]. For the first time, the world would have metals that it could trust.