Particle swarm optimization: A new tool to invert geophysical data

Abstract

Particle swarm optimization (PSO) is a global optimization strategy that simulates the social behavior observed in a flock (swarm) of birds searching for food. A simple search strategy in PSO guides the algorithm toward the best solution through constant updating of the cognitive knowledge and social behavior of the particles in the swarm. To evaluate the applicability of PSO to inversion of geophysical data, we inverted three noise-corrupted synthetic sounding data sets over a multilayered 1D earth model by using DC, induced polarization (IP), and magnetotelluric (MT) methods. The results show that acceptable solutions can be obtained with a swarm of about 300 particles and that convergence occurs in less than 100 iterations. The time required to execute a PSO algorithm is comparable to that of a genetic algorithm (GA). Similarly, the models estimated from PSO and GA are close to the true solutions. Whereas a ridge regression (RR) algorithm converges in four to eight iterations, it yields satisfactory results only when the initial model is very close to the true model. Models estimated from PSO explain observed, vertical electric sounding (VES) and MT data, from Bhiwani district, Haryana, India, and the Chottanagpur gneissic complex, Dhanbad, India. The results are consistent with RR and GA inversions.

2D dipping dike magnetic data interpretation using a robust particle swarm optimization

“…It depends on the reproduction of the obvious social conduct 60 of birds, fishes and insects in nourishment searching. PSO-calculation effectively connected and applied in many disciplines, like model development (Cedeno and Agrafiotis, 2003), biomedical pictures (Wachowiak et al, 2004), electromagnetic optimizations (Boeringer and Werner, 2004), hydrological issues (Chau, 2008), and different geophysical application (Alvarez et al, 2006;Shaw and Srivastava, 2007). In this calculation the birds representing the particles or models, every particle has an area vector which represent the parameters value and a speed vector.…”

Particle Swarm Optimization (PSO) for Magnetotelluric (MT) 1D Inversion Modeling

“…[8,9,10]. The PSO algorithm has been applied to non-linear inversion of magnetotelluric (MT) and Vertical Electrical Sounding (VES) data using 1-D model with satisfactory results [11,12]. In this paper, we follow the same approach with the emphasis for MT 1-D inversion of synthetic data.…”

Seismic inversion and uncertainty analysis using a transdimensional Markov chain Monte Carlo method

“…In the PSO, the models, called particles, are navigated in the model space by following the current optimal model as well as their individual best location in the moving history. This method is relatively newer than the above methods, and it is now used to invert geophysical data and characterize reservoir (Shaw and Srivastava, 2007;Fernández-Martínez et al, 2008Zhe and Gu, 2013).…”