DEVELOPMENT OF A MONTE-CARLO IN VERSION TECHNIQUE FOR MASW DATA

Abstract

Multi-channel Analysis of Surface waves (MASW) method is widely used in near surface geophysical and geotechnical applications. inversion of l-D dispersion curve has been attempted using various algorithms in literature. The present effort is devoted to develop a Monte Carlo inversion technique for the MASW data. But due to the known inefficiency of pure Monte Carlo methods, the method was appended to its advanced form, called Simulated Annealing, with the power of simplex algorithm. The hybrid search method was used to invert the dispersion curve of MASW. The code is developed in MATLAB v.201 lb.The theoretical dispersion curve was attempted by WAVe method (Leong, 2012). 1-lere, both the numerical and experimental testing is reported. The developed method is compared with its competitors like neighborhood-search (a widely used free surface wave inversion program). The hybrid method is solely developed in, MATLAB v. 201 1b . Most of the global optimization techniques suffer from a common drawback of late convergence. Moreover, in the process the user needs to supply the upper and lower bounds of the earth parameters. In our code, an automated apriori model is selected based on the given dispersion curve. This aids in the next step of inversion. The developed code is built around the Nelder-Mcad Simplex algorithm, which is a local search method. The SA finds the solution in the neighborhood of the global optimum and the final convergence is accelerated using the simplex algorithm. The hybrid method was tested for inverting for 3 to 9 layered earth models for the synthetic data, with an RMS error as less as 0.2-0.01. In the case of inverse dispersive profiles the largest error was reported was less than 5 in the RMS scale. With such accuracy it becomes a reliable inversion tool for the user. The method is always ready for incorporating new techniques and changes.