APPLICATION OF INDEPENDENT COMPONENT ANALYSIS TO CHEMICAL REACTIONS

摘要: ABSTRACT The analysis of kinetic data monitored using spectroscopic techniques and its resolution into unknown components is described. Independent Component Analysis (ICA) can be considered a calibration free technique with the outcome analyses being spectral profiles species. This enables realisation qualitative information concerning identification number type present within reaction mixture over time. ICA approaches FastICA JADE multivariate curve resolution-alternating least squares were applied to spectra first order synthetic reaction. For all signal was successfully separated from constituent components. 1. INTRODUCTION Reaction monitoring major challenge across process industries. form typically involves measurement prediction concentration in chemical determination mixture. Furthermore extracted importance terms understanding. Traditionally models are built predict property interest. However success approach depends on factors. Calibration modelling time consuming final model sensitive changes conditions. In addition it only provides quantitative about interest no side reactions intermediates [1]. Moreover industrial processes may because operational disturbances or lack detailed understanding mechanisms. Thus by-products produced that their existence cannot determined existing interpretation methods. There consequently need for more advanced methods mixtures reactions. paper investigates mixtures. A simulated set generated different rates. Techniques included Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) [2] [3]. Specifically two algorithms investigated, [4] Joint Approximate Diagonalization Eigenmatrices (JADE) [5] results used as an initial estimate MCR-ALS algorithm. Applications have previously been reported areas voice sound separation, biomedical processing, financial series, wireless communications image feature extraction.