Application of partial least squares calibration to measurements of polycyclic aromatic hydrocarbons in coal tar pitch volatiles

Abstract

Samples of air taken from a high temperature coke oven plant were assayed for the concentration of PAHs. The analyses were focused on a group of 16 PAHs as recommended by US Environmental Protection Agency. Their electronic absorption spectra and GC–MS traces were recorded. The GC–MS was used to provide information on the concentrations. Partial least squares regression was applied to form a calibration model between the EAS spectra and the corresponding concentrations of the PAHs in a complex mixture, such as coal tar pitch volatiles. The influence of different data preprocessing techniques was investigated. Emphasis was placed on the selection of optimal calibration data sets. Variable (wavelength channel) selection plays an important role in the formation of multivariate calibration models and influences its prediction ability. Variables were selected according to correlation coefficients with the corresponding concentration profiles.