Particle Swarm Optimization of Electromagnetic Data with Parallel Computing in the 2D Case

This work presents the application to electromagnetic data processing of the particle swarm optimization (PSO) algorithm, with the advantage of its high speed of convergence to the problem solution compared with other evolutionary methods. We focus on the inversion of MT synthetic data in the two-dimensional case. The PSO problem is solved complying with upper and lower bounds and a priori information, which is initially given only to a small amount of particles of the swarm. The fitness function is properly defined including smoothing parameter (Occam's razor). Since it is a computationally demanding problem, a practical tool of parallel computing is tested and validated, thus allowing large computation time savings. Results show encouraging outcomes in terms of minimization of fitness function and data fitting. This approach represents the starting point for the 2D PSO application to MT data, Audio Magneto Telluric data and other near surface applications implying two-dimensional interpretation