The development of a reliable on-line method to monitor process streams is important for improved process control in oil sand extraction plants. The suitability of diffuse reflectance NJR spectroscopy for this purpose was investigated in a pilot plant environment using an on-line fibre-optic probe. Spectra of a feed slurry flowing through a 25-mm pipe were acquired continuously over a nine-hour period during which ore type and slurry water content were varied.
The feasibility of monitoring feed stream conditions by online NJR spectroscopy is demonstrated by means of a principal component analysis of the measured spectra. Clustering of these spectra according to ore type and water content enables the detection of deviations from and transitions between steady-state conditions of the process. Estimates are given of characteristic times for the process to reach a steady state after a change in condition was initiated. The use of artificial neural networks for classifying spectra on the basis of ore type also is illustrated.
Primary analytical determinations for oil, solids, and water contents of slurry samples collected during the run were used in multivariate calibrations of the spectra. The results suggest that quantitative analysis of an on-line spectrum should be preceded by a classification step (according to ore type, for instance) to permit selection of the appropriate calibration model.
Variability in the composition of oil sand ores makes accurate on-line analysis essential for optimal process control in an extraction plant One technique that has aroused significant interest in various applications is near-infrared reflectance analysis (NlRA).This is an attractive method for characterizing suitable samples as it is rapid, non-destructive, and usually requires little, if any, sample preparation. Large samples can be measured since the NIR region comprises primarily overtone bands, which generally have lowabsorption coefficients. Furthermore, recent advances in fibre-optic technology allow coupling of the spectrometer to process streams.
At Syncrude's extraction plant, the bitumen content of incoming oil sand ores is measured on-line by an NIR reflectance spectrometer1. This information is used to control the rate of caustic addition to the ore on the basis of mpirical results indicating that bitumen content varies inversely with fine clay content, which determines the caustic requirement. However, as intensity data are acquired at only two wavelengths, much of the information present in an NIR spectrum of oil sand is not used.
The NIR region of a transformed diffuse reflectance spectrum of a raw oil sand (as measured on a BOMEM FTNIR spectrometer) shown in Fig. 1 is similar to spectra displayed in papers by Dougan1 and Shaw and Kratochvil2. The prominent features are CH aliphatic and aromatic first combination bands at 2300 nm, an AJO-H first combination band in clay at 2200 mo, overtone bands of OH in water and clay at 1900 and 1400 nm, and a CH overtone band in oil at 1700 nm. The decaying baseline between 1100 and 2000 nm epresents the tail of an electronic excitation band in the visible spectrum of oil. A similar feature also appears in the NIR spectrum of illite2, one of the clays present in Athabasca oil sand. Silica has no absorption bands in the NIR and its presence is manifested only as a baseline shift.