An Experimental Design Approach for Structural Integrity Testing of Drilled Shafts Using Full Waveform Inversion

One of the key challenges of drilled shaft non-destructive testing (NDT) is the quality of the acquired data, which is related to the survey configuration. In the specific case of crosshole tomography (CT), relatively small spacing is used between sources and receivers in the access tubes. Even with such an extensive data acquisition program, CT can fail to detect defects that are large enough to negatively affect foundation performance. Full waveform inversion (FWI) is a high-resolution tomographic technique that can potentially address this issue. Instead of relying on first arrival times in a ray-based analysis, FWI attempts to simulate wave propagation within the drilled shaft and match the entirety of the signals acquired during crosshole testing. This approach improves resolution at the expense of higher computational costs. However, FWI might not necessarily require such extensive instrumentation for satisfactory performance. Therefore, it is beneficial to evaluate possible survey configurations that can optimize data acquisition and increase the practicality of FWI for drilled shaft structural integrity testing. In this study, it is proposed that a sequential optimal experimental design (SOED) approach can provide insights into potential optimal survey configurations. The outcomes are presented as a benefit-cost curve that allows one to select an appropriate number of crosshole sources based on available resources. The principles of SOED are first described, followed by a discussion of its performance when used to evaluate optimal survey configuration. 2D FWI is then performed using different survey configurations on numerical models of drilled shafts with known defects. The results highlight an increase in the number of crosshole sources, which directly increases the computational efforts but does not continually lead to appreciable improvements in the inverted images and subsequent interpretation of defect geometry and location.