摘要: Multichannel seismic data collected by NRL-SSC near the Blake Outer Ridge showregions of anomalous SVO (spectrum vs. offset) behavior which is modeled by thin sediment layers of high shear velocity. We apply a nonlinear inversion ... 展开 Multichannel seismic data collected by NRL-SSC near the Blake Outer Ridge showregions of anomalous SVO (spectrum vs. offset) behavior which is modeled by thin sediment layers of high shear velocity. We apply a nonlinear inversion algorithm called simulated annealing (SA) to this problem with the aim of developing a tool for routine analysis of similar data with minimal user interaction. The use of a computationally intensive nonlinear scheme is necessary as the energy function is highly complicated, with minimal user interaction. The use of a computationally intensive nonlinear scheme is necessary as the energy function is highly complicated, with many local minima and with large regions of very low gradient. This makes it impossible for iterative gradient-based inversion methods to work. To test the algorithm rapidly we generated synthetic (T,p) gathers and used a classical Metropolis SA and a variant of it called Very Fast Simulated Annealing (VFSA). Both of the algorithms and four different types of error functions work extremely well. We are also able to invert the data in frequency-offset (w,x) domain. Although forward modeling in the (w,x) domain, transformation of most data into the (T,p) domain without aliasing is very difficult due to limited spatial sampling in most observations. An inversion of thin layers in the (T,p) domain requires high frequency data and many closely spaced receivers (or sources); inversion in the (w,x) domain can resolve thin layer properties with a few more widely spaced receivers. Most field data have limited offset range and spatial sampling thus justifying the slower (w,x) technique. (AN). 收起