Pattern recognition algorithm for inversion of surface wave phase velocities

Inversion of surface wave phase velocities not only suffers from computational difficulties associated with being trapped by local minima for most local-search methods but also undergoes the high computational effort for most global optimization techniques due to its high nonlinearity and to its multimodality. In order to effectively overcome the above described difficulties, we proposed a new surface wave inversion scheme based on the Mesh Adaptive Direct Search (MADS) algorithm, a commonly used pattern recognition algorithm. The calculation efficiency and stability of the proposed procedure are tested on a four-layer synthetic model and a real-world example. Results from both synthetic and real field data demonstrate that the MADS algorithm applied to surface wave inversion should be considered good not only in terms of accuracy but also in terms of computation effort due to its global and fast convergence in the final stage of search.