Multi-Objective Optimization of Spectra Using Genetic Algorithms

This paper applies genetic algorithms (GAs), a powerful general-purpose biologically motivated optimization technique, to the multi-objective problem of spectrum optimization. Two objectives, color and efficiency, are address using real spectra, although the addition of other objectives (e.g., color rendering, color temperature) is relatively straightforward. The direct application of the method presented is to transform the spectrum of newly developed lighting technologies to have desirable color properties while maximizing efficiency. Other applications of this methodology include the design of a filter for the input of a fiber optic system such that the color at then end of a given length of fiber has particular properties (e.g., appears “white”), while the efficiency of the system is minimally affected. The principal findings described in this paper are the implementation of an efficient multi-objective fitness function tailored to this problem and a method for speeding convergence of the GA by "smoothing the chromosomes." An algorithm, data and results from several approaches are presented. Introduction It is the goal of lighting manufacturers to produce light sources with maximum luminous efficacy, the ratio of the total luminous flux to total power input (i.e., "amount of light" per Watt). Luminous flux (F) is defined as: