CPAC optical moisture monitoring: Characterization of composition and physical effects on moisture determination Task 2A report

Abstract

The impact of particle size and chemical composition variations on determination of tank simulant moisture from near infrared (NIR) optical spectra are presented. This work shows particle size and chemical variations will impact moisture predictions from NIR spectra. However, the prediction errors can be minimized if calibration models are built with samples containing these variations as interferents. Prior work showed the NIR spectral region (1100 to 2500 nm) could be used to predict moisture content of BY-104 tank simulant with a standard error less of approximately 0.5 wt%. Particle size will increase moisture prediction error if calibration-models do not include the same particle size ranges as unknown samples. A combined particle size model with 0-420 {times}10{sup -6}m, 420-841 {times} 10{sup -6}m, and 841 {times} 10{sup -6} m-2 mm diameter particles predicted 0.59, 0.34 nd 0.23 wt% errors respectively for samples containing only these size ranges and 0.80 wt% error for a samples with all particle size ranges. Chemical composition would also increase moisture prediction error if calibration model samples chemically differ from unknown samples. For a BY-104 simulant, increases in NaOH, NaAlO{sub 2}, Na{sub 2} SiO{sub 3}, and Na{sub 3}PO{sub 4} produced moisture predictions that were lower than the actual moisture levels while increases in FE(NO{sub 3}){sub 3}, Ca(NO{sub 3}){sub 2}, and Mg (NO{sub 3}){sub 2} resulted in a higher than actual moisture prediction. Systematic changes in the NIR spectra could be observed for these families of materials. When all of the composition variations were included in a single model, the model had a moisture prediction error of 1.41 wt% as compared to a 2.96 wt% error without model changes. This work shows a calibration model based on a single set of tightly controlled experimental conditions will tend to have somewhat larger prediction errors when applied to samples collected with variations outside of such conditions.