A Hybrid of Artificial Neural Networks and Particle Swarm Optimization Algorithm for Inverse Modeling of Leakage in Earth Dams

Abstract

A new intelligent hybrid method for inverse modeling (Parameter Identification) of leakage from the body and foundation of earth dams considering transient flow model has been presented in this paper. The main objective is to determine the permeability in different parts of the dams using observation data. An objective function which concurrently employs time series of hydraulic heads and flow rates observations has been defined to overcome the ill-posedness issue (nonuniqueness and instability of the identified parameters). A finite element model which considers all construction phases of an earth dam has been generated and then orthogonal design, back propagation artificial neural network and Particle Swarm Optimization algorithm has been used simultaneously to perform inverse modeling. The suggested method has been used for inverse modeling of seepage in Baft dam in Kerman, Iran as a case study. Permeability coefficients of different parts of the dam have been inspected for three distinct predefined cases and in all three cases excellent results have been attained. The highly fitting results confirm the applicability of the recommended procedure in the inverse modeling of real large-scale problems to find the origin of leakage channels which not only reduces the calculation cost but also raises the consistency and efficacy in such problems.