Optimization of multiple edge barriers with genetic algorithms coupled with a Nelder–Mead local search

This research work aims at developing a new multi-criteria optimization method dedicated to complex road noise barriers. Numerical simulations of the acoustical propagation have been achieved using MICADO, a 2D boundary element method (BEM) code developed at CSTB. The optimization part is carried out with the help of a Nelder–Mead algorithm (direct local search method) coupled with an evolutionary strategy in order to globalize the approach. A first application of this combination between an outdoor sound propagation numerical code and an optimization algorithm concerns the optimization of noise barrier caps with the following varying parameters: the cap size, its shape and its surface impedance. The cost function to be minimized is defined through a mean value of the insertion loss due to the added crowning compared to the straight, rigid barrier solution of same overall height, averaged on several receiver points within the barrier shadow zone. Final results show a significant improvement of the efficiency of a multiple edge noise barrier by optimizing values of both size and impedance.