These have all emerged from organisations outside the mainstream renewable communities, in most cases from independent, third-party investigators. This is of particular interest because, with the exception of highly conspicuous nacelle fires, there appears to be little public knowledge of such incidents.
This may represent a considerable feat of publicity control, but if wind turbine OEMs keep any lessons learned in-house, they limit wider input that could inform the techniques and processes needed to develop long-term integrity.
Such a response would be less than optimal given the importance of establishing and sustaining integrity strategies and the feedback loops needed to keep designs up to date.
Failures are invariably related to fatigue. This is understandable in that wind turbines are rotating machines that are subject to cyclic stress.
Optimisation of both component and weld design is an important feature of fatigue mitigation, but it would appear that the understanding of materials and the stresses around interfaces such as bolting and welding has slipped under the integrity radar.
In the cases that DNV GL Loughborough has been involved in, there are two distinct groupings of problems.
One set involves component fabrication with issues that may be ascribed to quality assurance, particularly around welding, giving rise to initiation sites for fatigue cracking.
The second is related to an underestimation of the cyclic stresses present in notionally static structures.
The first group of failures studied arose from defects that were induced in thick-walled blade components during welding by subcontractors.
These defects had eliminated the need for fatigue crack initiation and had, as a result, led to blade separations at an early stage in the operational lives of the machines involved.
Those from other sources involved fractures of pitch gearing and blade separation due to assembly practices that led to early fatigue crack initiation through degradation of fastener holes.
In two cases, the fabrication of the components had resulted in the formation of cracks in the subassembly as a direct consequence of welding.
In one instance, rogue material — a medium-carbon alloy steel — had been used for one side of the weld. The use of a welding procedure designed for low-carbon structural steel resulted in significant cracking in the heat-affect zone of the rogue material.
The stresses in the welds were high because the use of automated submerged-arc welding meant that only a single internal and external weld run were needed to complete the procedure.
The stresses induced during the cooling of the external weld run were therefore high enough in the medium-carbon chromium to induce cracking of the section (figure 1). This formed the initiator from which a fatigue crack propagated.
Some areas of the initial weld crack were covered by a weld run that had no apparent relation to the main butt weld (figure 2); this was taken as a clear indication that the fabricator had noted the crack prior to shipping and attempted to cover it with a weld.
The second crack was of greater overall concern as it occurred in a bracket weld in which the weld procedure used and the two components being welded were all as specified. In this case lamella tearing — cracking beneath welds — occurred during welding (figure 3).
This mechanism arises in high-constraint welds (thick material results in high constraint) where the base material plane lies transverse to the weld and the material contains significant amounts of elongated inclusions.
Most structural steels produced in rolling mills contains some elongated manganese sulphide inclusions, but above a certain concentration these can result in lamella tearing.
This is a known problem, and in this case there was clear evidence that the fabricator had welded over the crack, having prioritised rapid output over correct manufacture.
In each case there was a clear paper trail that showed all the required non-destructive testing was complete and correct. But the additional welding revealed cracking had been detected and attempts had been made to cover it up.
In both cases, the vendor had to suspend operation of turbines in the immediate vicinity of the damaged machines to ensure that they were not suffering from similar defects, with all the reputational damage and lost revenue that ensued.
Both failures were resolved in the local context, but they gave rise to global uncertainty because the issues were present in components of relatively common machines and therefore had potential implications beyond the wind farms directly involved.
In two additional cases investigated by the DNV GL materials teams, fatigue cracking of static frames within turbine structures were found.
These cracks were all associated with interrupted partial penetration structural welds (figure 4).
This type of weld is known to be unsuitable for applications where cyclic loading is expected. In these cases, the fabrication of the welds appeared to be as specified. This indicated that the problem was a lack of understanding of the actual loads experienced in the structures.
These experiences give rise to two overarching conclusions: relatively simple errors in fabrication are occurring but being hidden to meet throughput requirements, resulting in failures; and crack propagation in static structures is occurring because live loading is not being correctly predicted.
In both instances the problems can be ascribed to teething problems in an industry that is in a rapid state of expansion. But in these early phases, it is important that accurate failure investigations are carried out to ensure that problems can be identified and addressed in design upgrades.
Jeffrey Jones is senior consultant, materials at DNV GL and a committee member of the IOM3 Materials Failure Investigation Group
This article was first published in WindMax (incorporating WindStats)