FATUIGE failure

important parameter in designing

STATIC STRENGTH BLADE TEST :
Test procedure
Initially, two blades were tested under static load conditions using a load profile common for AE tests (see, for example, FRP fan blades [3]). The sequence was based around two characteristic load levels: the nominal operating load (OL) and the maximum test load (MTL). Normally, the blade should not undergo critical damage when loaded up to OL and so multiple tests can be performed at this level before the final strength test. However, it is known that composite structures emit when they are first loaded to a given level, so it is necessary to repeat the initial application of the OL. During this second loading, it is expected that there will be no AE if no damage is propagated. In common with many standardised AE tests it was desired to have load-holds of up to 10 minutes duration in order to establish stability criteria. Following the two load-holds at OL the load was increased by successive increments, each with a three minute load hold, until the blade failed. For the second blade the load was removed between increments.

The load application point and the sensor locations for the tests discussed herein are shown in figure 1. The blade is 4.5 m long and is attached to a reaction frame at the root. The complete load envelope to failure at 5.9 kN is presented in figure 2.

Results
The results showing AE events v. time for each channel are presented in figure 3. The applied load is superimposed on the figure in green. The blade was loaded up to a final peak of 5.9 kN, at which point it failed catastrophically due to a buckling instability which initiated at a point 2.4 m from the blade root (between sensors 5 and 6).

At the third load peak (4.5 kN) all the signals are seen to stabilise during the load-hold period, as witnessed by the convex curvature of the events v. time graph for all channels. During the fourth load peak (5.0 kN), the trace for sensor 5 (and to a lesser extent sensor 4) already exhibits a different curvature and does not properly stabilise during the hold period. At the fifth load peak (5.5 kN), there is a steep rise in the number of events recorded at sensor 5 and sensors 4 to 6 are all still strongly emitting at the end of the load-hold period. The shape of the signal from channel 5, in particular, suggests that the blade damage has reached a critical level.
Information about the type of damage taking place can be obtained from the amplitude of the AE signals. The amplitude distribution indicates a concentration of high amplitude events, usually associated with fibre breakage, at sensor 4 during the load-hold at 5.0 kN and at sensors 4 and 6 during the load-hold at 5.5 kN. During final failure practically all the high amplitude events are concentrated close to sensor 6. A remarkably similar distribution was obtained from both blade tests with this load profile.

STATIC CERTIFICATION BLADE TEST :
Test procedure
As discussed in section 1, the MTL for the blade under consideration is estimated based on a single event, namely the 1 in 50 years gust. This event is generally modelled as a rapid load "spike" with a duration of no more than 10 seconds. Pass/fail certification criteria are based on the blade successfully surviving this load. Load-holds of up to 3 minutes at sizeable fractions of the MTL were not acceptable to the blade manufacturers since these were perceived to be unrealistic loads that would not be seen in service and which might cause premature failure during the certification test

It was necessary therefore to develop a load profile which incorporated the standard certification loading and to base the AE evaluation criteria on such tests. At the same time, it was recognised that the certification loading was designed to mimic the extreme load the blade might experience in service and it was necessary to be able to determine after the event whether such a loading had caused damage within the blade. Thus, it was specified that a sequence of so-called examination loadings (AEL) should be applied to the blade after the MTL to determine the minimum load level at which AE could be detected. The resulting load profile is shown in figure 4 and was applied to a third test blade. To avoid initiating the same buckling failure mode, the load on this third blade was applied just 2m from the blade root. Final failure was again from buckling around the saddle point.

Results
Since the load history effectively combines two different test strategies, the data was first divided into two sections, which were analysed separately. Results are presented here from the sequence of standard (increasing) MTL tests only. Analysis is still continuing on data from the intermediate AEL loading

A graph showing the distribution of linearly located events along the blade v. time from the sequence of MTL loads is presented in figure 5. This shows a progression of event clusters starting from the root and moving outwards to the loading saddle as the test progressed. Arbitrary placement location was used to identify particular clusters of activity in 2-D space and identified notable concentrations around the inner end of the shear web and the maximum chord area of the blade. However, without longer load-hold periods it is not possible to make any judgement about the criticality of damage at any stage during the test. This task is made more difficult by the nature of the final failure - buckling instability - which may well not have been preceded by significant material failure.
It is hoped that future tests on blades which have been designed to avoid the buckling instability will give clearer results in this respect.

FATIGUE TEST PROCEDURE :
Test procedure
For a long test (the fatigue test was expected to take about 2 months) the quantity of non-critical AE data becomes the main problem. To keep the data to a manageable amount, only time driven data (i.e. parametrics, number of cycles, average AE signal level, absolute energy) was continuously recorded, then for 10 slow cycles every 10000 cycles the recording of hits was initiated. The sensor positions remain the same as for the static test and the threshold is set higher at 55 dB. In all, the fatigue test consisted of four load stages (see figure 6):

I. initial static loading to OL (AE hits data recorded).
II. static test loading at start and after every 500000 cycles to 10% above the peak fatigue load (AE hits data recorded).
III. fast cycles at 2 Hz sinusoidal fatigue load (R-ratio = 0.1) to the peak fatigue load (parametrics only - no AE hits data).
IV. after every 10000 cycles, the load application rate is reduced to give ten slower cycles at 0.2 Hz and during this interval the load-strain characteristic and the AE activity is monitored

Results
Due to the limitation on overall load that could be applied due to the likelihood of causing the static buckling instability it proved quite difficult to achieve a fatigue failure of the blade. Several load increases were applied to the blade in tension-tension loading, but then, as very little appeared to be happening, it was decided to extend the loading into the compression regime. The final breakage of the blade occurred soon after the change of loading but was not recorded by AE because it occurred between two low speed cycles.
A 2-parameter filter was designed and applied to all the data files from the low speed cycles. This filter is designed to separate out events due to fibre breakage or propagating damage in the blade. It accepts only events with amplitude greater than 70 dB and counts less than 100. (A second filter that has been devised to detect signals due to impending buckling failure will also be applied to the data.)

The combined results from the data collection during all the low speed cycles are presented in figure 7. The graph shows the energy (red high) v. x-position progressively through the test. The scale maximum is indicated on the right hand side since auto-scaling within the software has resulted in a change of scale for some of the later tests. It is interesting that initially there are two concentrations of activity, one in the blade root and at the start of the shear web located 0.8m along the blade (sensors 1-3) and the other between sensors 5 and 6 where the buckling instability set in to cause final failure of the blade. In the intermediate part of the test there is a relatively quiet period (test numbers 23-32), followed by another burst of activity at the root and now slightly outboard from the end of the shear web. Following the introduction of load reversal at test number 45, a wide spread of high energy signals were recorded between sensors 4 to 6, immediately prior to blade failure, which initiated in this area.

DISCUSSION:
The work reported here has sought to combine classical AE testing and analysis methodologies with the requirements of wind turbine blade certification testing. Although the AE data and analysis was able to correctly identify critical damage prior to the onset of the buckling stability in the static strength blade tests 1 and 2, this occurred at unrealistically high load levels. Using a loading more representative of the gust loading a blade might be expected to experience in the field, significant AE activity was obtained, confirming the potential to locate damage and make a preliminary assessment of damage type. However, the loading is such that the ability to determine damage criticality appears to have been lost and work is continuing to identify the lowest loads at which this can be achieved.
The fatigue test has indicated that collecting data during periodic high load cycles can, with suitable filtering, indicate where potential damage sites are located. This finding will be verified using other inspection techniques during later tests.
The results to date are very promising, but it is recognised that new criteria and analysis methods need to be developed in order to be able to determine criticality.

CONCLUSIONS:
Acoustic emission monitoring during static and fatigue wind turbine blade certification tests can give information about the location and type of damage events. The test differs markedly from conventional applications of AE and so it is difficult at present to make judgment over damage criticality from these tests. Damage criticality can be determined from the emission level during relatively long duration load-hold periods during an ultimate strength test, but such a loading is unrepresentative of blade loads in the field. Work is continuing to establish the lowest level of AE examination load at which significant emission can be expected in a seriously damaged blade.

Fatigue testing:
Fatigue as a specific failure mechanism has been recognised since the early part of the nineteenth century but it was the development of rail travel that resulted in a major increase of interest in this type of fracture.
The premature failure of wagon axles led to Wohler in Germany investigating fatigue failure under rotating loading. This led to the design of the first standardised test - a reversing stress rotating specimen, illustrated in Fig.1.

There are many mechanisms that can lead to failure but fatigue is perhaps one of the most insidious since it can lead to a catastrophic failure with little or no warning - one well known example being the failure of the Comet aircraft in the 1950s.
Failure can occur at a fluctuating load well below the yield point of the metal and below the allowable static design stress. The number of cycles at which failure occurs may vary from a couple of hundreds to millions. There will be little or no deformation at failure and the fracture has a characteristic surface, as shown in Fig.2.
The surface is smooth and shows concentric rings, known as beach marks, that radiate from the origin; these beach marks becoming coarser as the crack propagation rate increases. Viewing the surface on a scanning electron microscope at high magnification shows each cycle of stress causes a single ripple. The component finally fails by a ductile or brittle overload.
Fatigue cracks generally start at changes in section or notches where the stress is raised locally and, as a general rule, the sharper the notch the shorter the fatigue life - one reason why cracks are so damaging.
There are two stages in the process of fatigue cracking - a period of time during which a fatigue crack is nucleated and a second stage where the crack grows incrementally leaving the ripples described above. In an unwelded component the bulk of the life is spent in initiating a fatigue crack with a shorter period spent in crack propagation.
An unwelded ferritic steel component exhibits an endurance limit - a stress below which fatigue cracking will not initiate and failure will therefore not occur. This is not the case with most non-ferrous metals or with welded joints - these have no clearly defined endurance limit.
The reason for this is that in arc welded joints there is an 'intrusion' - a small defect at the toe of the weld, perhaps only some 0.1mm deep. Provided that the applied stress is sufficiently large a crack will begin to propagate within an extremely short period of time. The endurance limit for a welded joint is therefore dependent on the intrusion size that does not result in crack propagation at the applied stress range. In the case of a welded joint, therefore, a fatigue limit - a 'safe life' is specified, often the stress to cause failure at 2x10 6 or 10 7 cycles.
During fatigue the stress may alternate about zero, may vary from zero to a maximum or may vary about some value above - or below - zero.
To quantify the effect of these varying stresses fatigue testing is carried out by applying a particular stress range and this is continued until the test piece fails. The number of cycles to failure is recorded and the test then repeated at a variety of different stress ranges.
This enables an S/N curve, a graph of the applied stress range, S, against N, the number of cycles to failure, to be plotted as illustrated in Fig.3. This graph shows the results of testing a plain specimen and a welded component. The endurance limit of the plain specimen is shown as the horizontal line - if the stress is below this line the test piece will last for an infinite number of cycles. The curve for the welded sample, however, continues to trend down to a point where the stress range is insufficient to cause a crack to propagate from the intrusion.

By testing a series of identical specimens it is possible to develop S/N curves. In service however, there will be variations in stress range and frequency. The direction of the load may vary, the environment and the shape of the component will all affect the fatigue life, as explained later in this article.
When designing a test to determine service performance it is therefore necessary to simulate as closely as possible these conditions if an accurate life is to be determined. In order to enable the fatigue life to be calculated when the stress range varies in this random manner, the Palmgren-Miners cumulative damage rule is used.
This rule states that, if the life at a given stress is N and the number of cycles that the component has experienced is a smaller number, n, then the fatigue life that has been used up is n/N.
If the number of cycles at the various stress ranges are then added together - n 1 /N 1 + n 2 /N 2 + n 3 /N 3 + n 4 /N 4 etc - the fatigue life is used up when the sum is of all these ratios is 1. Although this does not give a precise estimate of fatigue life, Miners rule was generally regarded as being safe. This method, however, has now been superceded with the far more accurate approach detailed in the British Standard BS 7608.
The design of a welded joint has a dominant effect on fatigue life. It is therefore necessary to ensure that a structure that will experience fatigue loading in the individual joints has adequate strength. The commonest method for determining fatigue life is to refer to S/N curves that have been produced for the relevant weld designs.
The design rules for this range of joint designs were first developed by TWI and incorporated with the bridge code BS 5400 in 1980 and then into the industry design rules for offshore structures. Further refinements and improvements finally resulted in the publication of BS 7608 Code of practice for fatigue design and assessment of steel structures. This standard will be looked at in more detail in a future article.

Fatigue testing and assessment services :
Welded joints have poor fatigue properties which must be allowed for in design. TWI played a pioneering role in establishing fatigue design guidance for steels and aluminium alloys, now incorporated in British and International Standards.

Our services included:
Interpretation of fatigue design rules
Assessments using numerical modelling, design codes and experimental data
Guidance on joint design
Specification and application of fatigue life improvement processes
Fatigue testing of welded and unwelded joints and specimens
Interpretation of testing results and comparison with TWI's extensive fatigue database
Detailed investigation and interpretation of service failures
Rapid deployment of experts for site investigation and recommendations on repair.