Reduction of ultrasonic multiple scattering applied to flaw detection with array probes in polycrystalline materials

Abstract

Flaw detection using ultrasonic evaluation of coarse-grain steels is perturbed by a high structural noise due to scattering. This leads to a decrease of the detection capabilities, particularly at high frequencies and large depths for which multiple scattering dominates. Recent academic studies have shown that the contribution of multiple scattering could be dramatically reduced. These results were obtained on a model random medium made of parallel steel rods immersed in water. The ability to detect a target could be significantly increased using a specific filtering method, based on the full matrix capture (F.M.C.) combined with a smart post-treatment based on random matrix theory, in supplement with the DORT method (i.e., decomposition of the time-reversal operator). Here, the same technique to separate simple and multiple scattering contributions is now applied to a real material. Experimental results were obtained on a nickel-based alloy (Inconel600®) with a thermicallyinduced coarse grain structure and manufactured flaws (side drilled holes) at different depths. The experimental set-up used a multi-element ultrasonic array. Results are presented and compared to other detection techniques, at various depths and frequencies. Despite a dominant multiple scattering noise, a significant improvement of the detection performances is observed.

The detection of flaws in austenitic welds using the decomposition of the time-reversal operator

“…The overall location and characterization of the flaw is improved in this image, however a less impressive SNR of only 16.8 dB is achieved. The higher SNR evident in the image constructed by the DORT algorithm also suggests that it has improved detection capabilities in coarse-grained media over the standard TFM and indeed, further evidence to support this conclusion is shown in [ 42 ]. The overall conclusion is that due to the limited information used by the DORT algorithm, characterization proves problematic.…”