Using Genetic Algorithm to Identify Completely Unknown System in FTIR Spectra Analysis

Abstract

The article mainly focuses on how to identify completely unknown systems in FTIR spectra analysis by Genetic Algorithm (GA). The technique of multiplespot crossover combined with quantitative and adaptive mutation is used and the whole process is controlled by the fitness value and probability of selection in GA. A standard pure FTIR spectra database containing 101 air toxic organic compounds is set up. Six unknown samples are identified and quantitatively analyzed. The results indicate that GA can successfully identify even a perplexing sample containing 18 components. The concentration of each component in the samples can be determined at the same time. The maximal relative error is no more than 10%. It means that GA has the capability of simultaneous identification and prediction, especially for those completely unknown mixed spectra in FTIR.