Curvature Effects on Responses from Transverse Flaws in Piping Welds

In this paper, the practical limitations surrounding the probability of detection of transverse weld flaws are discussed when performing manual, skewed ultrasonic inspection of piping welds.


Introduction
Inspection of piping welds requires processes for detection of flaws oriented parallel and transverse to the weld axis. In ASME BPVC Sec. V Art. 4, this is stated in T-472.1.2 and T-472. 1.3 (1). In most situations the weld cap is left in place, thus inspection of transverse flaws is addressed by manual skewing techniques. To perform this task, the search unit is skewed towards and along the weld axis while it is traversed around the circumference of the pipe (Figure 1).

Sensitivity Calibration
A sensitivity calibration (e.g. time-corrected gain, or TCG) is normally performed to equalize reference reflector indications at the same screen height over the sound path used. A previous paper by the authors (2) details some of the challenges of performing sensitivity calibrations on curved surfaces and the differences between orientation of reference reflectors.
For girth welds, the primary weld inspection direction is axial. Considering only the beam centreline, the plot for this direction simulates a flat plate ( Figure 2). Inspection in the purely circumferential direction is illustrated in Figure 3. Manual, skewed inspections for transverse flaws, regardless of piping weld orientation (girth or long seam) result in a beam direction that is neither purely axial nor purely circumferential but is rather a dynamic combination of the two. Attempting to calibrate with a flat wedge on a surface curved in the circumferential direction is fraught with complications (2). The curved beam direction in Figure 3 results in a longer sound path and change in angle response (if using notches), so it is true to state the axial and circumferential targets are not geometric equivalents. However, in the circumferential direction, the wildly variable coupling efficiency, beam apodization, and counter-intuitive angle effects highlighted in the previous paper effectively negate any perceived advantages of using this reflector for calibration when using manual skewing techniques.

Beam Coverage for Flat Welds
A simple plot showing the effectiveness of a 45° beam on a 9mm V-bevel with a 6mm HAZ is shown in Figure 4. These simulations address parts with little to no curvature, like those on very large pipe or on flat components.
In the first orientation, the beam is directed entirely perpendicular to the weld axis. When the probe is manually skewed only 30° to detect a transverse flaw, the beam center point rotates and fails to enter the body of the weld ( Figure 5).  These illustrations are shown using a 45° refracted angle, however use of higher angles is subject to the same effects. The coverage of the beam is based on angle, part thickness, and weld geometry. In many real-world scenarios and common weld sizes, it is evident that the effectiveness of the beam coverage rapidly diminishes with skew angle.

Beam Plots on Curved Surfaces
Previous work has been done studying the effects on the ultrasonic beam when varying wedge and part curvature (3). This work included multiple scenarios varying part diameter, thickness, wedge contour, coupling efficiency, element size and wedge rocking. All analyses were concentrated on the effects in the axial beam direction utilizing AOD targets.
One conclusion showed that in the axial direction, the wedge gap effect on coupling was of much less concern than the actual rocking of the wedge itself. However, in the case of manual scanning for transverse flaws where the probe is skewed, the beam divergence in both the lateral and vertical axes are affected by the pipe curvature. This effect increases as beam skew increases, and is maximum when the beam is purely circumferential (Figure 8). The illustrations in the preceding section dealt with essentially flat components. Since a manually skewed transverse scan is a sort of hybrid between axial and circumferential, introducing part curvature creates compounding complications. Firstly, there are beam divergence and convergence effects that occur after reflection from the curved ID and OD surfaces (Figure 9). These will combine in ways which simply cannot be predicted in a practical, real-world environment. Secondly, when the beam is skewed, the reflection off a curved surface will redirect the beam left or right. This, again, cannot be predicted accurately in real time and increases the likelihood of plotting errors. Thirdly, beam spread effects and the combining of direct and reflected path indications may make characterization and quantitative determinations of flaws effectively impossible. These effects cannot be accounted for with standard calibration techniques. Figure 9: Excerpt from reference (3,Ginzel) showing beam divergence and convergence after reflection

Weld Coverage vs Skew Angle
Civa simulations were performed to graphically illustrate the beam geometry when the probe is skewed towards the weld. The simulation parameters used are listed below.
Pipe: NPS 10" pipe Sch 40 (273mm OD 9.3mm wall) Weld: 60° bevel with 2mm root land and 2mm gap Cap 3mm excess, root 2mm excess. Probe: 12.5mm diameter 5MHz 80% bandwidth Wedge: 60° refracted angle in steel Length L1: 15mm Length L2: 15mm Height L4: 12mm Width: 22mm Round bottom holes 1mm in diameter were modeled in the weld to identify coverage responses. The holes each had a 9mm ligament from the pipe opposite surface (e.g. less than 1mm deep from the drilled surface). Far side targets are 11mm from weld centreline just inside the full skip distance with the probe wedge touching the edge of the weld cap. This is shown in Figure 10. Civa does not permit a diagonal scan on a pipe surface, so a scan was configured to raster the beam to cross the weld and then a variation was configured to continue that raster with different probe skews varying from 0° to 60°. Figure 11 illustrates the probe at 0° skew (left) and 60° skew (right). The irregular line representing the stand-off (index axis) motion is a result of the probe riding up on the weld cap. This is perhaps an optimistic approach in that the weld cap drawn in Civa is only 11mm wide. At 0° skew, we see that at the probe's closest approach the beam centre cannot detect the root.
Only the volume at the edge of the root HAZ is being covered by this standard 60° probe. Beam coverage after the reflection off the inside surface is then limited to the volume above the 2mm root land (Figure 12).   At a 45° skew, with the corner of the wedge pivoting against the weld cap, the full skip distance of the beam is only halfway across the weld cap at the near surface. The far side of the weld is not being interrogated by the beam centre ( Figure 15). This may not be of concern if the weld is inspected in this fashion from both sides, however it serves to illustrate the quickly receding zone of coverage as skew angle increases.

Comparing Amplitude Responses
To compare the relative responses from a calibration notch to a transverse planar flaw, a calibration scan was first configured using ID and OD 1mm notch-targets. The response from the calibration ID notch was 0.27 pts and the OD notch peaked at 0.084 pts. Another half skip to the ID notch indicated it was not much different than the OD notch at 0.069 pts. For convenience, these point values are converted to decibels (Figure 17). A scan was then configured using the same probe on a simulated girth weld with 2 transverse planar flaws. The flaw near the outer surface was 3mm high and 10mm long with just a 1mm ligament to the surface. The flaw at the inside surface across the root was 3mm high and 6mm long and was surface-breaking (i.e. no ligament).
When the echodynamic curves were analysed using the same relative amplitudes, it became apparent that the signals being detected were nothing more than tip diffractions (Figure 18). Having demonstrated using the round-bottom holes that the beam at a skew greater than 45° does not have the central ray interrogate anything past the weld centreline, the detections are therefore due to just a small part of the outer edge of the beam interacting with the tips of the crack edges. This is evident when we look at the signal amplitudes from the transverse cracks ( Figure 19). The near side of the upper crack gives a response of 21.4 dB below reference and the far side of the upper crack gives a response 30.6 dB below reference. It would be reasonable to expect that the response from the root notch would be lost in the noise that could exist from the root excess metal geometry. Note that these levels are far below normal evaluation levels such as 20% of reference (-14 dB).  Figure 20: Civa plot of amplitude responses (in Civa "points") from transverse cracks using skewed 45° probe. The relative amplitude of the first peak is 21.4dB below the calibrated reference level.

Conclusions
Inspection of transverse weld flaws using a manually skewed scan is fraught with complications. The limited coverage area, beam dynamics, and poor interrogation angles on the predicated flaw faces simply does not reflect the same level of reliability and repeatability as perpendicular scans for axial flaws. Using lower refracted angles such as 45° helps maintain beam integrity and predictability but provides poor coverage. Higher beam angles improve coverage but may not plot as expected and incur mixed indications from direct and indirect reflections (2).
Once the physics and geometries are illustrated and understood, the practical limitations become obvious. Detection of transverse flaws with ultrasound, when the weld cap is left in place and a manual scan must be performed, provides a very low probability of detection. This conclusion has been documented in prior research by Morgan (4).
Sensitivity calibration is based on maximizing the target responses using the centre of the beam. However, the simulations presented in this paper illustrate that it is most likely not the beam centre that will be interrogating the transverse flaw but rather the beam periphery. Added to this is the fact that responses may more likely be due to stray tip diffractions than reflections from crack faces or corner traps themselves.
While the examples in the preceding sections are shown for only one thickness and weld geometry, it should be apparent the same limitations will exist for many common weld sizes and configurations.
With the understanding of limited beam coverages, angular restrictions, and high likelihood of missed calls, one might simply avoid skewed inspections for transverse flaws and consider it a geometric restriction, which is provided for in the code as a limitation. The alternative is to knowingly perform an inspection technique with a very low probability of detection, and more than likely declare the weld free from transverse indications.
It is recommended that for welds where transverse flaws are suspected, a surface technique such as magnetic particle inspection, liquid penetrant inspection, or eddy current be performed.