Virtual Standard Slope and Bias Calibration Transfer of Partial Least Squares Jet Fuel Property Models to Multiple near Infrared Spectroscopy Instruments

Abstract

A novel calibration transfer (standardisation) method was used to transfer partial least squares (PLS) models correlating jet fuel properties with their near infrared spectra. The modelled jet fuel properties included: API gravity; %aromatics; cetane index; density; distillation temperatures for 10%, 20%, 50% and 90% recovered volume; flashpoint; %hydrogen content; %saturates; and viscosity. The transfer method involved the use of 12 neat solvents to generate virtual standards which resembled the training set spectra used to generate the PLS models. Virtual standards created from solvents acquired on the master calibration instrument and secondary slave instruments were predicted using the master-generated PLS models. The master and slave virtual predictions were then regressed to provide a simple slope and bias correction for the slave instruments. The virtual standard slope and bias correction performed as well as or better than piece-wise direct standardisation for nine of the 13 fuel parameters while eliminating the need to maintain either the master instrument or complex fuel standards.