Multivariate calibration with Raman spectroscopic data: a case study

An industrial process separating p-xylene from mainly other C-8 aromatic co mpounds is monitored with an online remote Raman analyser. The concentratio ns of six constituents are currently evaluated with a classical calibration method. The aim of the study being to improve the precision of the monitor ing of the process, inverse calibration linear methods were applied on a sy nthetic dataset, in order to evaluate the improvement in prediction such me thods could yield. Several methods were tested including principal componen t regression with variable selection, partial least square regression or mu ltiple linear regression with variable selection (stepwise or based on gene tic algorithm). Methods based on selected wavelengths are of great interest because the obtained models can be expected to be very robust toward exper imental conditions. However, because of the important noise in the spectra due to short accumulation time, variable selection methods selected a lot o f irrelevant variables by chance correlation. Strategies were investigated to solve this problem and build reliable robust models. These strategies in clude the use of signal pre-processing (smoothing and filtering in the Four ier or wavelets domain), and the use of an improved variable selection algo rithm based on the selection of spectral windows instead of single waveleng ths when this leads to a better model. The best results were achieved with multiple linear regression and stepwise variable selection applied to spect ra denoised in the Fourier domain. (C) 2000 Elsevier Science B.V. All right s reserved.