Inverse Modeling in Geoenvironmental Engineering Using a Novel Particle Swarm Optimization Algorithm

Algorithms derived by mimicking the nature are extremely useful for solving many real world problems in different engineering disciplines. Particle swarm optimization (PSO) especially has been greatly acknowledged for its simplicity and efficiency in obtaining good solutions for complex problems. However, premature convergence of the standard PSO and many of its variants is a downside particularly for its application to the inverse problems. This aspect encourages further research in developing efficient algorithms for such problems. In this work, a novel PSO algorithm is proposed by introducing fitness of a new location in the search space into the standard PSO which enables to enhance the success rate of the algorithm. The proposed algorithm uses center of mass of the population to compare the fitness of global best particle in each iteration. The proposed algorithm is applied to solve contaminant transport inverse problem. The performance of different PSO algorithms is compared on synthetic test data and it is shown that the proposed algorithm outperforms its counterparts. Further, accurate design parameters are estimated using the proposed inverse model from the experimental data.