MicroCT Analysis of Cracks in Wet Welds

3D Analysis of Cracks in Underwater Wet Welds

  • 3D rendering of micro-cracks in the thin section of an UWW specimen.3D rendering of micro-cracks in the thin section of an UWW specimen.
  • 3D rendering of micro-cracks in the thin section of an UWW specimen.
  • Fig. 1: Center: Variable width tensile test coupon. Left: Stress and strain values for each section. Right: MicroCT setup.
  • Fig. 2: Image processing secquence for a reconstructed microCT slice. (a) Original Image. (b) Noise reduction with anisotropic diffusion filter. (c) Background correction. (d) Segmentation with manual threshold. (e) Small object elimination.
  • Fig. 3: Quantitative results for the specimen sections: Left: All objects: Top) Number of objects. Center) Average object length. Bottom) Average object thickness. Right: After shape filtering with AR>8: Top) Average crack length. Bottom) Average crack thickness.

Underwater wet welding is a critical procedure for the repair of offshore structures. High cooling rates due to direct contact with water and the presence of hydrogen derived from water dissociation can lead to the formation of cracks in the weld metal. The detectability of this kind of crack by x-ray microCT was evaluated with an all weld test coupon that was scanned after a tensile test. A correlation between stress/strain level and crack length/thickness was determined by 3D Image Analysis.

Introduction

Underwater wet welding (UWW) is a critical procedure for the repair of offshore structures, mainly related to oil production and transportation. The harsh environmental conditions in which the weld is performed has strong consequences to the structural reliability of the welded joints. High cooling rates due to direct contact with water and the presence of hydrogen derived from water dissociation can lead to the formation of defects, such as pores and cracks in the weld metal (WM), which adversely affect mechanical properties. During cooling, weld beads contract both in transverse and longitudinal directions. It is well established that longitudinal contractions are responsible for higher residual stress after welding. Consequently, in wet welds, the embrittlement associated with high hydrogen content in the WM can lead to nucleation of transversal cracks [1,2].

The formation of cracks in the WM depends on the diffusible hydrogen content, which depends on the specific type of electrode employed. For instance, rutillic electrodes tend to produce weldments with very high diffusible hydrogen (≈90ml/100g) [3].

Previous works have employed both traditional 2D microscopy [4] and x-ray microCT [5] to reveal cracks in welds. Hydrogen micro-cracks (HMC´s) in the WM are platelet-shaped objects and show a predominant orientation transverse to the weld axis. These characteristics highlight the relevance of 3D techniques, as traditional 2D microscopy would not be able to reveal these complex shapes and relationships.

Moreover, this kind of crack can be very closed and very thin at the tips (< 1µm), probably beyond the typical resolution of microCT.

In this work microCT was employed to detect and measure HMC´s in a wet welded steel sample, evaluating the detection limits of the technique for this kind of defect.

A specially machined coupon with varying cross section was scanned before and after being submitted to a tensile test (TT). The cross section variation leads to stress/strain changes which, in turn, lead to change in HMC opening, thickness and length. These changes were quantified by microCT and Image Analysis.

Experimental

Rutillic high diffusible hydrogen content coated electrodes were used to weld an ASTM A-36 steel plate (150 x 280 x 16 mm). A 45o V-groove butt weld was performed along the longest dimension of the plate. To simulate the conditions of UWW the welds were conducted in a water tank with a mechanized gravity welding system. After welding an all weld TT sample with its length parallel to the welding direction was machined to create six regions (A-F) with different widths. The rationale was to create regions with different stress/strain values and, consequently, different crack openings after the TT. Figure 1 shows the recomposed specimen after the TT. As expected, the coupon failed at its thinnest section (F).

Figure 1 also shows the microCT setup. Displacing the specimen vertically 6 separate tomographs were obtained before and after the TT. A GE v|tome|x s microCT operating at 160 kV and 9 µm/pixel was employed.

After reconstruction the image stacks were processed with the FIJI/ImageJ software.

Results and Discussion

Measuring the specimen before and after the TT, the values of stress and strain in each section were obtained, as also shown in figure 1 alongside the specimen picture.

The images of the specimen before the TT were analyzed. Very few HMC´s were visible indicating, as expected, that they were still closed. No attempt at processing these images was made, as the contrast was too low.

After the TT the HMC´s became visible. Figure 2 shows the typical image processing sequence for one slice in a stack. Random noise was reduced with an anisotropic diffusion filter [5]. This filter is very efficient in removing high frequency noise while preserving object edges, what is critical in the case of small objects such as cracks. A typical background correction step, based on subtracting a low-pass version of the image, was applied. The cracks were segmented with a 2D manual threshold in each slice. Even though manual thresholding is dependent on user choice, it was possible to define a stable threshold level for all slices. Finally, very small non-elongated particles were deleted with a size-based object filter.

The same procedure was applied to all slices of the A-E image stacks. To allow comparing sections with different widths, all stacks were cropped to the width of Section E. Section F was discarded due to the damage caused by specimen failure.

The total number of objects, their thickness and length were measured. Object thickness corresponds to the average over the object length of the local 3D thickness obtained from the 3D distance map [6].

The left column of figure 3 shows the initial results. Length and thickness values shown are averages over all objects in each section. Even though, as expected, the number of measured objects increases from sections A to E, length and thickness do not show a clear trend.

To explain these results it is important, first, to mention that these measurements are restricted by microCT resolution, 9 µm/pixel. HMC´s shorter than this value are not detected. Thus, part of the variation in the number of objects comes from the thinner undetected HMC´s, e.g. in Section A, while many HMC´s in Section E are thick enough to be detected and counted. Moreover, crack tips are probably below the resolution, hindering the detection of the full crack length. That might explain why the average length does not vary between the sections.

Secondly, one must consider the presence of other types of defects, such as pores or inclusions, which are not sensitive to the stress/strain level. To test this hypothesis, a stricter shape filter was applied to keep only objects with aspect ratio > 8 (AR = length/thickness), i.e., long thin cracks.

The filtered results are shown in the right column of figure 3. Both length and thickness show a constant value for the lower stress values in sections A-C and a clear growth trend for the higher stress values of sections C-E. These results are consistent with the known yield strength (σy) of this material, between 400 and 440 MPa. Up to σy the sample is in the elastic regime and pre-existing cracks should remain closed. Above σy the material enters the plastic state and cracks open and grow, becoming thicker and longer.

Conclusions

MicroCT was successfully used to detect and measure HMC´s in a wet welded steel specimen. The variation in the cross-section of the tensile specimen led to a variation in stress/strain which, in turn, affected crack opening, length and thickness. This variation was consistently detected by 3D image analysis of the acquired image stacks.

Further analysis will employ thinner subsamples of each section and a x-ray microscope which combines geometric and optic magnification to achieve higher resolution.

References
[1] Gooch T.G.: Metal Construction 15, 164-205 (1983)
[2] Gooch T.G.: Metal Construction 15, 206-236 (1983)
[3] Santos V.R. et al.: Welding Journal 91, 319-328 (2012)
[4] Mauricio M.H.P. et al.: In: IMC17 - International Microscopy Congress, Rio de Janeiro (2010)
[5] Padilla E. et al.: Materials Characterization 83, 139-144 (2013)
[6] Dougherty R.P. et al.: Microscopy and Microanalysis 13, 1678-1679 (2007)

Authors
S. Paciornik, L. F. Silva, V. R. dos Santos

Pontifícia Universidade Católica do Rio de Janeiro, Brazil
G. Schneider, T. Bernthaler
Aalen University, Germany

Contact
Prof. Dr. Sidnei Paciornik
(corresponding author via e-mail request)
Pontifícia Universidade Católica do Rio de Janeiro
Department of Chemical and Materials Engineering
Rio de Janeiro, Brazil

Contact

Pontifícia Universidade Católica do Rio de Janeiro
Predio Leme - Sala 501L
CEP 22451- Rio de Janeiro
Brazil
Phone: +55 21 3527 1243

Register now!

The latest information directly via newsletter.

To prevent automated spam submissions leave this field empty.