Random?Objective Waveform Inversion of 3D?9C Shallow?Seismic Field Data

Robustness and uncertainty estimation are two challenging topics in full-waveform inversion (FWI). To overcome these challenges, we present the methodology of random-objective waveform (ROWI), which adopts a multi-objective framework preconditioned stochastic gradient descent optimization algorithm. The use one shot per iteration avoids using redundant data reduces computational cost. Pareto solutions represent group most likely their differences quantifies model associated with trade-off between conflicting objective functions. Due to high dimensionality spaces, it is prohibitively expensive check optimality all explicitly. Thus, decompose original function into shot-related subproblems for analysis. We apply ROWI field multi-component shallow-seismic set acquired Rheinstetten, Germany. 3D near-surface successfully reconstructed by main target, refilled trench, delineated. compare results estimated conventional least squares FWI prove efficiency ROWI. run six tests on different solution paths robustness against random path. validity verified multiple 2D ground-penetrating radar profiles. estimate 246 Another four starting from poor initial models performed, whose relatively model.