A Gaussian Adaptive Particle Swarm Optimisation (G-APSO) Algorithm to Invert Rayleigh Wave Dispersion

Particle Swarm Optimisation (PSO) techniques are based on the principles of stochastic inversion, employing random sampling alongside the collective intelligence of a swarm. We have effectively applied these techniques, which use evolutionary objective functions, to various geophysical inversion challenges. In this research, we present a Gaussian-Adaptive Particle Swarm Optimisation (G-APSO) method aimed at optimising dispersion curve inversion. This method incorporates a Gaussian distribution within the PSO algorithm to identify the normal distribution of inertia weight and coefficients. We then applied the G-APSO algorithm to invert for Rayleigh-wave dispersion, thereby deriving a one-dimensional shear-wave velocity (v_S). Convergence analysis was conducted to assess optimal parameter coefficients and validate the algorithm using both synthetic tests and a real passive seismic data set. Results indicate that G-APSO achieves efficient computational times and produces robust results for both synthetic data and real noise seismic data, yielding lower root mean square (RMS) errors and faster processing times compared to related stochastic optimiser algorithms. For practical application, we utilised phase-velocity dispersion-curve samples from seismic ambient-noise tomography collected in the Jakarta area of Indonesia. Furthermore, the results demonstrate that the G-APSO algorithm can effectively manage complex and occasionally biased data sets, inverting them into a one-dimensional v_S model.