Abstract

Sinkhole collapse may result in significant property damage and even loss of life. Early detection of pre-collapsed sinkholes (voids) are critical to limit the cost of remediation. One of the most promising ways to obtain subsurface imaging is 3D seismic full-waveform inversion. For demonstration, a recently developed 3D Gauss-Newton full-waveform inversion (3D GN-FWI) method is used to image buried voids, raveling soils, and characterize variable subsurface soil/rock layering. The method is based on a finite-difference solution of 3D elastic wave equations for the forward simulation and the Gauss-Newton optimization to extract material properties. The Gauss-Newton approach is well recognized as a robust and effective technique for numerical optimizations of non-linear problems, with faster convergence rates than the traditional gradient method. More importantly, for near-surface imaging, the inverse Hessian matrix used in the Gauss-Newton inversion acts as a weighting function to balance the gradient vector and model updating during the inversion. It reduces the dominancy of Rayleigh waves (less weights for shallow cells) and increases the contribution of body waves in the far-field data (more weights for deeper cells), and thus helps resolve deeper structures. Field experiment was conducted at a retention pond in Newberry, Florida for detection of unknown voids. The test site consists of medium dense, fine sand and silt underlain by highly variable limestone. This was a blind seismic survey (class-A-prediction), as the test area was an open and flat ground without indication of subsidence or chimneys. More importantly, no previous knowledge of voids, soil/rock layering, pinnacles or raveling existed at the test area. The seismic field experiment was conducted for the test area of 18 × 36 m, with a test configuration of 72 receivers (4.5-Hz vertical geophones) and 91 shots (source impacts) located in 2D uniform grids on the ground surface. The receiver and source grids were 6 × 12 and 7 × 13, respectively, both at 3 m spacing. A propelled energy generator with a 40 kg drop-weight was used to induce seismic waves at each of the 91 shot locations, and the wavefield generated from each shot was simultaneously recorded by the 72 geophones. The measured field data was then analyzed by the 3D GN-FWI method to extract both P-wave and S-wave velocity profiles. The analyzed domain of 36 × 18 × 18 m (length × width × depth) was discretized into 0.75 m cells for both the forward simulation and model updating. The measured data was filtered through two frequency bandwidths: 5-25 Hz and 5-35 Hz and used in two inversion runs. The first run employed a basic 1D gradient initial model using the 5-25 Hz data. The inverted profile from the first run was subsequently used as the input model for the second run using the 5-35 Hz data. The inverted results reveal that the 3D waveform analysis was able to identify unknown voids, raveling, as well as laterally variable soil/rock layering including rock pinnacles. The results were confirmed later by standard penetration tests (SPT), including depth to bedrock, two buried voids, and a raveling soil zone.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call