Artificial intelligence - algorithm development for laser spectroscopy studies

Abstract

Principal component regression and partial least squares regression are some of the widely used machine learning algorithms in recent years. They are artificial intelligence models used for dimensional reduction and learning key features which are applicable for regressions. These two models were applied into the field of advanced chemometrics for quantitative analysis. The objective of this quantitative analysis was to predict the identity of gaseous compounds used within the mixture samples which had varying level of concentration. In this quantitative analysis, gaseous mixture of compounds C2H4 and NH3 were used to simulate gaseous mixture emitted from vehicle exhaust. Data collection was made possible through a home-made quantum cascade laser which captured complex spectroscopic gaseous information. These data were computed to generate dataset containing key absorption spectrums information used for the identification of the gaseous compounds. The generated data set were processed into machine language for the regression analysis to take place. The regression outputs were benchmarked through quantitative evaluation metrics which assured that the results were accurate and reliable.