In the spring of 2005, engineers from Lawrence Livermore National Laboratories contacted SEP to ask about the potential to use geophysical algorithms for nondestructive investigation of manufactured/machined parts. The conversation was sparked by emphasis from the LLNL management to search for existing solutions to their suite of current problems. Dr. Lehman presented SEP with a problem of investigating the interior of a layered manufactured product with a potential delamination or rugosity of an interior layer. As opposed to standard geophysical problems, the velocity/density structure of the target is completely determined.
The velocity model of the product is three two layers submerged in water. This makes the total 1D velocity model: 11mm of 1500m/s, 3mm of 4100m/s, and 30mm of 2670m/s. Data given to SEP was modeled with the elastic ED3D finite difference code from Livermore. Center frequency 2.25 MHz. First presented to our group was a data volume consisting of a single shot and 100 receiver locations over 80mm of the surface giving a receiver spacing of 0.8081mm. The target to identify was 1mm negative step in the center of the model. The zero-offset time to the anomaly is approximately 0.04 ms. However, the only shot modeled was at the extreme left of the model space.
The most obvious problem identifiable from a geophysical perspective was a strong multiple train generated from energy ringing within the high velocity middle layer. Secondly, from the standpoints of either multiple removal or imaging, the lack of redundant information from multiple shot locations were immediately identified as problematic. In all other respects, the laboratory conditions available to collect data with no velocity uncertainty promised highly successful application of conventional geophysical processing technology.
A full fold, ns=nr, data volume was modeled and delivered to SEP during the summer. Also modeled by LLNL was a similar data volume with a up/down double spoon/scallop anomaly at the base of the third layer. With standard migration algorithms, we were able to image both targets with resolution of 0.1mm vertically and about 0.40405mm horizontally. Intrabed multiples from the second layer were not time-coincident in the middle half of the offset range, so no multiple attenuation efforts were required after the far offset traces were removed before imaging. Source-receiver, shot-profile, and zero ray parameter planewave migrations were implemented. Given the simplicity and cleanliness of the data, zero-offset images from all approaches were practically identical.