Multicomponent Measurement of Respirable Quartz, Kaolinite and Coal Dust using Fourier Transform Infrared Spectroscopy (FTIR): A Comparison Between Partial Least Squares and Principal Component Regressions

Abstract

Exposure to respirable crystalline silica (RCS) is potentially hazardous to the health of thousands of workers in Great Britain. Both X-ray diffraction (XRD) and Fourier transform infrared (FTIR) spectroscopy can be used to measure RCS to assess exposures. The current method outlined in the Health and Safety Executive’s (HSE) Methods for the Determination of Hazardous Substances (MDHS) guidance series is ‘MDHS 101 Crystalline silica in respirable airborne dust - Direct-on-filter analyses by infrared spectroscopy or x-ray’. This describes a procedure for the determination of time-weighted average concentrations of RCS either as quartz or cristobalite in airborne dust. FTIR is more commonly employed because it is less expensive, potentially portable and relatively easy to use. However, the FTIR analysis of RCS is affected by spectral interference from silicates. Chemometric techniques, known as Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR), are two computational processes that have the capability to remove spectral interference from FTIR spectra and correlate spectral features with constituent concentrations. These two common chemometric processes were tested on artificial mixtures of quartz and kaolinite in coal dust using the same commercially available software package. Calibration, validation and prediction samples were prepared by collecting aerosols of these dusts onto polyvinylchloride (PVC) filters using a Safety in Mines Personal Dust Sampler (SIMPEDS) respirable cyclone. PCR and PLSR analyses were compared when processing the same spectra. Good correlations between the target values, measured using XRD, were obtained for both the PCR and PLSR models e.g. 0.98–0.99 (quartz), 0.98–0.98 (kaolinite) and 0.96–0.97 (coal). The level of agreement between PCR and PLSR was within the 95% confidence value for each analyte. Slight differences observed between predicted PCR and PLSR values were due to the number of optimal principal components applied to each chemometric process. The presence of kaolinite in these samples caused an 18% overestimation of quartz, for the FTIR, when following MDHS 101 without a chemometric method. Chemometric methods are a useful approach to obtain interference-free results for the measurement of RCS from some workplace environments and to provide a multicomponent analysis to better characterise exposures of workers.

Application of a Fourier Transform Infrared (FTIR) Principal Component Regression (PCR) Chemometric Method for the Quantification of Respirable Crystalline Silica (Quartz), Kaolinite, and Coal in Coal Mine Dusts from Australia, UK, and South Africa

“…Both methods are based on a principal component analysis. Only slight differences between predicted results were found between these chemometric methods, when a similar number of principal components were applied (Stacey et al, 2021a).…”

“…In this study, the calibration and validation sets are based on (and extended upon herein) the preparation of samples with artificial mixtures of the three major components (quartz, kaolinite, and coal). The calibration and validation samples are mostly the same as those used in a previous study (Stacey et al, 2021a). A calibration set defines the relationship between the target values and the magnitude of the vectors for principal components in their spectra.…”

“…The procedure for the generation of prediction samples is described in Stacey et al (2021a), however, no pure quartz or kaolinite powder was added to the bulk dust mixtures. Aerosols were generated in a glass jar and the respirable dust was collected onto a 25-mm diameter PVC filter using a Safety in Mines Personal Dust Sampler (SIMPEDS) operating at a flow rate of 2.2 l min −1 (Casella, Bedford, UK).…”

“…Two bulk samples from Group 1, 2 bulk samples from Group 2, and 1 bulk sample from Group 3 (containing the highest response from the feldspar mineral) were used to generate 12, 9, and 6 samples, respectively, of aerosolized respirable dust for collection on filters. A total of 16 prediction samples were obtained from the South African mine dust with added kaolinite and quartz from the previous study (Stacey et al, 2021b) were included and 15 prediction samples from 2 coal seams from a single mine in Wales, in the UK. An additional set of 13 prediction samples (separate to other low quartz content Durrans samples added to the calibration and validation sets) of Durrans coal dust were also prepared to provide an additional check of the models prediction capabilities with low proportions (1-25%) of quartz A9950 and quartz-free kaolinite (Georgia, USA).…”

“…The procedures used to determine the target values for quartz, kaolinite, and coal are described in Stacey et al (2021a) and in Supplementary Information (available at Annals of Work Exposures and Health online). In brief, the target mass values for quartz and kaolinite were obtained using a direct on-aerosol-filter XRD method described in the Health and Safety Executive's (HSE) Method for the Determination of Hazardous Substances (MDHS) 101 (HSE, 2014).…”