GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 3.3 ______ ___ GNGTS 2023 The role of three different metaheuristic methods in geophysical data optimization F. Pace 1 , A. Godio 1 1 Department of Environment, Land and Infrastructure Engineering (DIATI), Politecnico di Torino, Torino, Italy Introduction There has been an increasing interest in global search methods for solving the geophysical inverse problem during the last decades (Sen and Stoffa, 2013). Several metaheuristic methods have been adopted in the geophysical literature, and ever novel methods are being proposed from non-geophysical studies. These methods are the so-called nature-inspired algorithms based on biological systems and social dynamics of groups of entities such as birds, fish, bats, ants, wolves and so on (Engelbrecht, 2007). These algorithms are original, sophisticated, and efficient in the optimization of geophysical data. The most popular metaheuristics are the genetic algorithm (GA) and particle swarm optimization (PSO), which have been largely adopted (Sen and Stoffa, 2013; Pace et al., 2021). An example of an emerging and promising population-based algorithm is the grey wolf optimizer (GWO) (Mirjalili et al., 2014). Then, the scientific question that arises is if all these algorithms are efficient and useful. We focused on a comparative analysis among the performances of three metaheuristic methods, that is, GA, PSO and GWO, for the optimization of time-domain electromagnetic (TDEM) data (Pace et al., 2022). The aim of the work is to highlight their advantages and weaknesses because it seems that they have been indistinctly applied to geophysical data so far. Moreover, it is rare the three algorithms are contextually compared. The three metaheuristic algorithms: GA, PSO, GWO The GA is an Evolutionary Computation paradigm and mimics the inheritance rules in nature, where the individuals with the best chromosomes survive (and the weakest individuals must die) (Sen and Stoffa, 2013). The PSO algorithm is based on the social behavior of agents sharing knowledge to achieve the best objective of the group, such as escaping from a predator or searching for food (Pace et al., 2019, 2021). The GWO algorithm is based on the social dynamics adopted by a group of wolves attacking a prey while searching for food (Mirjalili et al., 2014). The comparison among the three methods was performed by setting some common features. The same settings of the algorithms were the objective function, the forward modeling code, the number of candidate solutions sampled, the stopping criterion, the random initialization, the