Development of a solid state IR analyser system using chemometrics
There are many applications for gas and vapour measuring systems in environmental and industrial analysis. Different analytical techniques and detectors employed include; chromatography, catalytic filaments and pellistors, metal oxide and semiconductors, electrochemical sensors, piezoelectric and surface acoustic wave detectors, and spectrometry. The technique employed in this work is infrared spectrometry and it is part of a joint project between the Analytical Science and Optoelectronic Groups at the University of Hull. The aim of the project is to develop a robust multi-component infrared gas analyser using a fixed array of quantum well intersubband photodetectors QWIPs in which the photodetectors are tailored to measure at specific wavelengths. In order to provide a low cost portable detector, only a limited number of detectors can be incorporated into the array, therefore a method of selecting suitable wavelengths to measure is needed. To achieve this, FTIR spectra were collected for mixtures of CH4, CO, CO2 and N20. A variable selection algorithm based on multivariate linear regression (VS-MLR) was applied to the spectra to reduce the number of measurements required down to five. It was found that it was possible to build prediction models using low resolution and that baseline variations could be compensated for by using an internal standard. Based on the wavenumber selection models, the concentrations of CH4, CO, CO2 and N20 in new validation data could be predicted using wavenumbers 2129,2284,3071,3148 and 3488 cm 1, with prediction errors averaging less than 14 %. This gave comparable results to PLS models built from the entire spectra. An exponential dilution system was constructed to enable different concentrations of a gas standard to be prepared and a GC was included into the system to act as a reference method for IR measurements. Good correlation was achieved between IR and GC responses for CH4, C02 and N20 with RZ values greater than 0.988. A prototype system was built (which incorporated the dilution system), utilising narrowband filters (at wavelengths 2.9,3.25,4.3 5,4.73 and 10.62 μm) combined with general IR detectors, based on the results of the wavelength selection. This was shown to give promising results and future work will be to replace the IR detectors with QWIPs.# Also explored in this work was the feasibility of developing a system to measure volatile organicc ompounds(V OCs). This was investigatedb y collecting FTIR spectra for a variety of VOCs and employing PCA and cluster analysis to see if it was possible to distinguish between different types of compounds. Partial least squares (PLS) and VS-MLR was performed on a simple four component vapour mixture to demonstrate that the same methods employed on the CH4, CO, CO2 and N20 mix could also be applied to other mixtures. In conclusion, the project demonstrated the role chemometrics can play in the development of a solid state gas analyser, from experimental design to data treatment and analysis. It was proved the VS-MLR could be successfully utilised to determine suitable wavelengths to measure in the array for gas and vapour applications and a prototype system was constructed using narrow band filters to demonstrate the operation of such a system.