Novel Fibre Bragg Grating design using multiobjective evolutionary algorithms

Abstract

A multiobjective evolutionary optimisation algorithm is applied to a fibre Bragg grating (optical filter) design problem. The design specified a dual wavelength filter with four required spectral characteristics - total bandwidth, peak separation, peak width and minimum transmission. Five parameters which described the apodised grating profile were used to define the search space and the transfer matrix method was used to numerically evaluate the transmission spectrum of candidate solutions. Various constraints on the search space were included in the design algorithm. Two separate selection schemes were tested, a distance based approach as used in the nondominated sorting genetic algorithm (NSGA-II) and a conglomerative clustering approach as used in the strength Pareto evolutionary algorithm (SPEA). Nondominated solutions are found and it is evident that particular objectives can be achieved more easily than others. Preliminary results are discussed and future work is introduced.