A Multiscale Master-Slave Parallel Genetic Algorithm with Application to Groundwater Remediation Design

Numerical models are an integral part of many optimal engineering design problems. The accuracy of fitness function evaluations for these applications depends on the spatial grid sizes used in the numerical models. Fine grids usually improve the accuracy of the solutions, but they can also pose major bottlenecks in the computational efficiency of the algorithms. The need to carefully select a grid size that can maintain the numerical accuracy of the solutions, along with being computationally less exhaustive, is very crucial for applications that use genetic algorithms. In this work, we present a multiscale parallel genetic algorithm that can be used to improve the performance of engineering design problems that use spatial grids. The algorithm’s efficacy is tested using a groundwater remediation design case study.