| NDT.net - March 2002, Vol. 7 No. 03 |
Detection and characterization of „small cracks" by non-destructive testing methods proves difficult especially when the cracks are near other inhomogeneities – pores, material inclusions or geometric features like edges or traces of machining. But this neighbourhood is just the normal case because fatigue cracks grow preferentially in the stress field of microscopic or geometric notches. In parallel it is an important task to detect and – if possible – to characterize cracks as early as possible after their formation. By that, inspection intervals for safety relevant parts can be prolongated without increasing the risk. This can produce important economic effects.
For many testing methods the problem with small cracks is not that the sensitivity is too small. In fact the indications of cracks are masked by indications of features which are not relevant from the fracture mechanical point of view. It is more a problem of differentiation between the two groups of signals. So in ultrasonic testing of surfaces by Rayleigh waves very often edges of objects and traces of machining produce echoes which are comparable or even stronger than the cracks searched for. Therefore, the challenge is to find methods discriminating between cracks and other features.
For surface breaking cracks such a method is available by the penetrant testing and the magnetic particle method. The effects used are the capillarity of the cracks and their special influence on the magnetic fields. But the eddy current method can be adapted, too. For cracks starting from surfaces which are not accessible or from an inhomogeneity in the volume the situation is much more complicated. The former case is given for overlapping plates. Testing of such overlaps is especially relevant in aviation.
An other feature of cracks - utilizable to characterize them - is the motion of crack surfaces under varying elastic stress. So variation in stress leads to friction of the crack faces and plastic deformation at the crack tip. Among others elastic waves emanate and heat is produced; effects which can be detected by acoustic methods or thermography, respectively. On the other hand the influence of the motion of crack faces on the transmission, reflection or diffraction of ultrasound can be utilized. This could be demonstrated for stresses generated by a laser spot of pulsating intensity [1] whereby this kind of excitation is limited to surfaces which are accessible. In contrast to local stress fields from laser excitation in this paper global stress fields are applied. The intention is to detect small cracks in the volume or near covered surfaces with improved sensitivity.
The basic idea is that usually a crack in an unloaded component is partially closed in the surrounding of its tip (crack closure effect – see[2]). In an externally varying stress field the size of the closed area will vary and so the interaction of ultrasound with the crack will do synchronously. This leads to corresponding variations in the amplitudes of received signals. For the signals of other features a corresponding amplitude of variation will be much smaller.
We distinguish between two methods – the load cycle B-image method and the load cycle difference method – according to the time regime in which the ultrasonic waves probe the crack.
Fig 1: scheme of the time schedule of load cycle difference method and load cycle B-image method.
|
In the load cycle B-image (LCB-image) method (Fig. 1, bottom) this probing is done with a high repetition rate for about one load cycle. The data are represented in a kind of modified B-image, where on the second axis not the position of the sensor but the external load is plotted [3].
In the load cycle difference method the crack is probed at a much less rate e.g. only in the maximum and minimum of load. If there are crack closure effects the signal contribution from the crack will change while the signal contribution from spurious echoes will be unchanged to a large extend. So, a simple subtraction of both signals should yield A-scans nearly free of spurious echoes. Of course this A-scans can be averaged as common A-scans to enhance the signal to noise ratio further.
Both methods are illustrated with a few examples. We will start with the load cycle B-image method in a transmission arrangement [4,5]. Fig. 2 summarizes the most important information for a measurement of a CCT specimen. The original RF signals have been included (as hardcopy) to see how the LCB-image(shown above the A-scan) is composed. Both time scales are identical. Taking into account the time-dependent force given in Fig. 2 the following observations can be made: decreasing from the maximum applied tensile stress the transmitted ultrasonic signal does not change first. Beginning with F=Fclose there are significant changes in the A-scans; the amplitudes increase and the signal onset shifts to earlier times, i. e. the distance moved by the signal decreases. For increasing tensile stress the behaviour is exactly symmetrical.
Fig 2: LCB-image method for a CCT specimen containing a fatigue crack, the probes are positioned such, that the crack is in the direct (undiffracted) sound path.
|
A consistent interpretation of this observation is: for all forces F ³Fclosethe crack is opened completely. Ultrasound transmission through the crack is not possible so the measured signals come from crack tip diffraction. At Fclosecrack closure sets in. The geometric path length between ultrasound transmitter and receiver decreases with decreasing tensile stress and the signals shift to smaller travel time. Consistently the amplitudes increase because the crack tip is "illuminated" with rising intensities and detected with increasing sensitivity, respectively (due to the directivity of the probes). To which extent at minimum tensile stress a part of the ultrasonic wave is transmitted already directly through the crack needs additional investigations.
Fig. 4 shows the pulse echo arrangement applied in the load cycle difference method.
The specimen made of an aluminium alloy (Al-Cu-Mg) has dimensions 250 x 48 x 4 mm³. The diameter of the drilled hole is 8 mm. An angle beam probe A5067 with and incident angle of 45° and a centre frequency of f = 5 MHz has been used. Prior to testing the probe has been arranged in such a way that the hole echo has maximum intensity. A mean stress of 11 kN and an oscillation amplitude of ±9 kN has been applied.
Fig 3: Comparison between the hole echo at maximum force on the one side and the difference signal between the echo at maximum force and minimum force; state without fatigue crack, all signals averaged over 512 cycles. |
Fig. 3 shows the situation before occurrence of a fatigue crack. The averaged echo signals at maximum force (shown at top) differs little from the averaged echoes at minimum force (not shown). The difference is shown in the bottom. In this case the suppression of structure indications was 29 dB. Such a big suppression confirms the hope that also small crack echoes can be found.
|
Fig 4: Arrangement for the load cycle difference method The results of a corresponding crack growth test are shown in the Figures 5 and 6. In Fig. 5 the hole echo is given for various states of specimen. On the right hand side the number of fatigue cycles is given in thousands. At 240 000 cycles a crack starting from the hole became visible for the first time. Significant changes in the echo signal become visible only very near the instant of fracture of the specimen at 247 300. Very small changes can be suspected earlier but they are connected to the hole echo so closely that they cannot be identified as individual echoes – at least if only one single A-scan is available. |
Fig 5: hole echo of the specimen at various fatigue states. (in thousand cycles);
| 
Fig 6: echo-difference of the hole region for the specimen at various fatigue states (in thousand cycles);
|
The situation is much more advantageous using the echo difference (Fig.6). In this case an additional echo appears prior to the first visibility of the crack at 240 000 cycles and then continuously grows with the number of cycles.
The precise formation of these signal components is not obvious. Simulations with numeric methods e.g. the Finite Integration Technique (EFIT) can help to clear the situation. But without doubt a growing fatigue crack is involved – it can be detected by these signal parts therefore.
But most probably the crack involved is not the crack at the side of the hole opposite to the probe, which was the first detectable by optical means, but a fatigue crack on the side of the probe. Usually these cracks develop at similar cycle numbers – not all penetrate to the surface and become visible early.
It has been shown that additional information on short cracks can be obtained by load dependent registration and processing of ultrasonic signals. In detecting many A-scans over a load cycle and displaying them as LCB-images the load dependence of crack properties can be recorded. This method requires a large amount of data already for one single cycle. Therefore, it is appropriate mainly for basic research to study e.g. the crack closure effect in fracture mechanics specimens and to determine the crack closure force and the driving force parameter DKeff. On the other hand, large memories as well as advanced signal and image processing techniques become increasingly available also for routine applications such as crack detection and sizing.
It is clear from LCB-scan display that crack closure affects the propagation of ultrasonic waves considerably. To explore this effect for fast online measurements, in a second variant the ultrasonic signals are recorded in the load maxima and minima. By calculating the difference signals the load independent echoes are eliminated and only the difference between the echoes from a crack opened more or less remains.
The measurements done gave the following central results:
The shape of the signals in detail is unclear yet so the results presented has to be taken as an encouragement for additional investigations. Especially valuable should be additional EFIT simulations. By this a better understanding of the ultrasound travel paths involved should be gained and therefore the origin of the particular signal contents should be clarified.
Of course the generation of cyclic loads in a testing machine is suited for basic laboratory research only and not for field use. But it is a promising idea to use the operation loads responsible for both the fatigue and the possible crack growth also for the measurements. In cases where this turns out to be difficult, e.g. in large components like airplanes, as an additional possibility the external imprinting of a cyclic load e.g. as low frequency bending waves, should be tested.
| © NDT.net - info@ndt.net | |Top| |