Prospects for Low-Resolution NDIR Sensors to Discern Ignition Properties of Fuels

Abstract

Abstract Derived cetane number (DCN) varies within jet fuels. With the expected prevalence of alternative jet fuels, additional variability is expected. DCN is usually assigned to fuels using ASTM methods like the Ignition Quality Tester (IQT). Recently, there have been developments in chemometric techniques, which use Machine Learning (ML) to correlate infrared spectra of fuels to properties like DCN, density, and C/H ratio. These techniques have advantages over ASTM methods, and previous studies have shown high accuracies in DCN prediction. However, this accuracy is from high resolution measurements, making the equipment relatively large and power-hungry. Alternatively, nondispersive infrared (NDIR) sensors, despite having low resolution, are attractive because they can be compact, inexpensive, and power efficient. This study investigates the trade-off between complexity and accuracy and assesses the feasibility of low-resolution NDIR sensors to discern DCN by using ML models trained on real FTIR data and DCNs obtained from IQT. DCN predictions are made for blends of ATJ/F-24, CN fuels, and neat Jet A1, A2, and JP5, with an error within 10%. There seems to be sufficient variability in the near infrared range to discern DCN with a feasible number of channels, if they are narrow (60nm FWHM ). For the data set in the study, the performance of linear models is better than a non-linear model. NIR region beyond 1050 nm is found to be important in DCN prediction, primarily the regions consisting of the first and second CH overtones and the CH combination band.