Oct 8, 2014 - SPECTROSCOPY OF OIL SANDS FOR. BITUMEN GRADE ... "Bitumen Content Estimation of Athabasca Oil Sand from Broad. Band Infrared ...
PROCESSING AND INFORMATION EXTRACTION FROM NEAR INFRARED (NIR) DIFFUSE REFLECTANCE SPECTROSCOPY OF OIL SANDS FOR BITUMEN GRADE MEASUREMENT Theodore Garver, Ph.D. Matthew Morrison, E.I.T. Nancy Luo Canadian Mineral Processor’s Conference (CMP) October 8, 2014 Fort McMurray, AB
Overview Introduction Overview of the challenge of NIR
measurement of oil sands
Sources of NIR signal Diffuse reflectance Vibrational and electronic transitions Overtones and combinations Components: bitumen, clay, sand, water
Effects of Particle Size, Packing Density Sample preparation methodology is key
Characterization of spectra for ore quality descriptors. Calibration of field instruments to lab measurements
AITF Instruments NIR Analyzer™ % bitumen & water
Tank Gauge Level Measurement
K40 Analyzer™ % clay
Bitumen Quality Solvent in Bitumen
OILSAND ORE SURFACE MINE
FROTH TREATMENT
water CONDITIONING SLURRY PREP
chemical TAILING S
K40 Analyzer™ % clay
Tailings Analyzer™ % bitumen in tails
FROTH SLURRY
PRIMARY SEPARATOR SEPARATION
MIDDLINGS
Solvent in Tailings
Tailings Analyzer™ % bitumen in middlings Level Detector™ Interface Measurements
Wall Thickness Wireless Pipeline Monitor
TAILINGS
Tailings Analyzer™ % bitumen in tails
BITUMEN
PRODUCT
K40 Analyzer™ % clay
Instrumentation Output Accurate and Reproducible Spectroscopy
Data Acquisition
Process Relevant Information Extracted as Descriptors
Data Enhanced to Improve Information Extraction
Data Pre-processing
Information Extraction
Bitumen Water Fines Embedded in Instrument
Control Methodology
Information Use Client
Instrument Design
Base-line Correction
Peak Height
Feed Forward Process Control
Instrument Reliability
Filtering
Integration
Feed Backward Process Control
Smoothing
Derivatives
Mining Control
Derivatives
Defined wavelength
Inventory and Mixing Control
Multiplicative Scatter Correction
Principle Component Analysis Least Squares Fitting to Reference Spectra Chemometics
AITF is involved in improving the information quality that is available to the client for use with process control. Activities to improve instrument quality may range from issues related to spectroscopic configuration, reliability, data pre-processing and chemometric approaches to information extraction from the acquired data.
NIR Ore Analyzer NIR Ore Analyzer is placed above a belt or apron feeder. Measures diffuse reflectance near infrared light from oil sands sample. Signal is derived primarily from vibrational energies of functional group (-CH2, -CH3, R-OH, HOH) vibrational modes in the near-infrared region. Instrument software upgrade provides:
Corrected reflectance Analysis of bitumen water and clay components Diagnosis for spectrum integrity, lamp age and interference from humidity.
System Configuration
The Challenge of Diffuse Reflectance NIR Measurement of Oil Sands The diffuse reflectance signal varies with: Sample texture, roughness or granularity Particle size of mineral components
Sample heterogeneity.
The NIR surface measurement may not be related in a simple way to bulk property measurements. Clay may be covered with a layer of bitumen and sand;
other times the bitumen may be adsorbed on the clay. Differences in the absorptivity and scattering efficiency of different components tends to favor signal from absorbing substances (hydrocarbon).
Sources of Signal (I)
Infrared vibrational spectra of oil sand peaks relate to bitumen, clay, sand and water. Rivard, B., D. Lyder, et al. (2010). "Bitumen Content Estimation of Athabasca Oil Sand from Broad Band Infrared Reflectance Spectra." The Canadian Journal of Chemical Engineering 88(5): 830-838
Diffuse Reflectance NIR The near infrared instrument measures light that has scattered off of the illuminated materials. This light has interacted with the surface layers of the material and reflected, refracted and diffracted as well as absorbed. • Spectral reflectance tends to specular reflectance preserve spectrum integrity and provides little information on the surface absorbance. • Diffuse reflectance provides spectroscopic information on the absorbance of the surface. • Diffuse reflection depends on texture, particle size, packing, moisture. surface 20.00
diffuse reflectance
20.00
surface
Overtones and Combinations Second harmonic overtones (on diagonal, highlighted) and combinations. Combinations are shown in green. wavenumbers (1/cm)
wavelength (nm)
1/cm
1375
1450
2850
2920
2940
nm
7273 6897 3509 3425 3401
1375
2750
2825
4225
4295
4315
7273
3636 3540 2367 2328 2317
1450
2825
2900
4300
4370
4390
6897
3540 3448 2326 2288 2278
2850
4225
4300
5700
5770
5790
3509
2367 2326 1754 1733 1727
2920
4295
4370
5770
5840
5860
3425
2328 2288 1733 1712 1706
2940
4315
4390
5790
5860
5880
3401
2317 2278 1727 1706 1701
NIR of Oil Sand 0.2 nm Resolution Spectra
Sources of Signal (III)
Bitumen, Clay, Water Dougan described breaking the signal from oil sands into component signals from bitumen, water and mineral.(Dougan, 1989). DOUGAN, P. D. 1989. Near-Infrared Reflectance Analysis: Its Potential Application in Oil Sand Processing. AOSTRA Journal of Research, 5, 203-210.
Data Extraction Methods With offset or baseline subtraction. Peak height measurement Peak area measurement
Reference spectra fitting. Derivatives (with smoothing) Savitzky-Golay methods are most popular
Empirical Fitting Methods Partial least squares
Principle component analysis
Peak Integration
Peak Derivatives
Fitting to Reference Spectra
Light interaction with Oil Sands Sample heterogeneity and granularity 1 0.9
A.
absorbance (baseline corrected)
0.8 0.7 0.6
sy001-1 sy001-2
0.5
sy001-3 sy001-4
0.4
sy001-5 sy001-6
0.3
0.2 0.1 0 850
B.
1050
1250
1450
1650 1850 wavelength (nm)
2050
2250
2450
2650
Packing and sample measurement with loose packing in A. and dense packing in B. Bitumen is sampled more times in the loose pack structure. The variation of NIR diffuse reflection signal from oil sands as a function of sample packing was initially described by Shaw et. al. SHAW, R. C. & KRATOCHVIL, B. 1990. Near-Infrared Diffuse Reflectance Analysis of Athabasca Oil Sand. Analytical Chemistry, 62, 167-174.
Oil Signal and Particle Sizes
Particle size can influence the calibration between bitumen concentration and observed NIR signal.
Calibration to Bitumen, Water, Clay Outputs from the instrument have been calibrated to bitumen, water, and clay. The best correlation is with bitumen (generally between 0.75
and 0.90 R2. Correlation with clay or fines values is between 0.50 and 0.75 R2.
Bitumen, clay and water descriptors can be used with non-linear and multiple regression models. We are currently verifying calibration models with online data.
Data Processing
Conclusions Sampling and sample preparation methods must account for signal variation due to particle texture (granularity). Variation between sample presentation in the lab
and in the field must be accounted for when using a lab calibration in the field. Buried clay may not be detected by the analyzer and should be accounted for.
Different data extraction methods yield correlated (similar) results. The best method depends on spectral quality and resolution.
Final Comment Our integrated instrumentation development program spans from core science to instrument engineering to field implementation and service. We are working with partner companies to accelerate instrumentation development for the oil sands. The improvements in the NIR ore grade analyzer discussed here substantially improve instrument accuracy and reliability while reducing maintenance needs.