Genetic Algorithm That Considers Scattering for THz Quantitative Analysis

Quantitative analysis is an important application of terahertz (THz) technology. However, scattering limits the accuracy of the commonly used least-squares and partial least-squares regression methods in quantitative analysis. This paper presents a genetic algorithm (GA) for quantitative analysis of multicomponent samples using THz absorption spectra, which exploits artificial intelligence and offers the advantage of global optimization. Because the scattering effect is considered when the fitness function is established, almost all of the quantitative errors are below 2%. Furthermore, optimal values of the crossover and mutation probabilities were selected considering their effects on the quantitative error and time used. The GA was found to be superior to the least-squares method for quantitative analysis of mixtures using their terahertz absorption spectra because of its excellent nonlinear discrimination ability and higher accuracy.