Online composition measurements of refrigerant mixtures by optical fibres and infrared spectroscopy
Résumé
As refrigeration and heat pump industries use more and more mixtures as refrigerants it is useful to measure the composition of the fluid at different locations in the machine. For our own research purposes on mixture based cycles, we have implemented an online non-intrusive technique based
on infrared spectroscopy. This paper describes the apparatus used and also the data treatment that allows to extract valuable information from the recorded spectra. The calibration of the method for a given mixture is obtained by recording spectra with known concentrations, pressure and temperature. Partial least squares (PLS) or principal component regression (PCR) statistical methods are used to build a model. Then compositions can be determined for unknown mixtures. Results are presented for CO2+Propane mixtures. The sensitivity of the method to baseline treatment is tested. The best accuracy obtained so far is of the order of 2%
The method is well tested and can be applied to other mixtures.