Clay
Preliminary Investigation of Near Infrared Spectroscopy for Green Sand Component Identification Dr. Scott R. Giese University of Northern Iowa
Research Objective • Green sand molding characterization for process control – Numerous tests • Development of control charts by Heine • Fairly slow for results for some tests • Too generic in some instances
• Develop on-line method for determining compositional components
Compositional Components • Major components – Clay • Serves as bonding agent and determines strength
– Water • Actives clay and varies strength
• Minor components – Additives such as seacoal
– Recycled core materials
Near-Infrared (NIR) Spectroscopy Wavelength = 780 – 2500 nm (12500 – 4000 cm-1) Combination and overtones of –CH, -OH, and –NH vibration Normal Mode Vibration (MIR)
High degree of Excitation
Combination and Overtones (NIR)
Low degree of Excitation
Source - http://www.docstoc.com/docs/79530219/Near-Infrared-Spectroscopy
Characteristics of NIR • Reflection intensity stronger • Silica can be used as transmittance materials
• Sample preparation not required • NIR spectra characteristic to material
Example of Bentonite Stretching and Bending -OH Stretch
Absorbed Water -OH Stretch Structural Hydroxyl Groups
Dispersive Spectrometer Fiber-Optic Wavelength 1000 – 2500 nm with resolution 10-15 nm InGaAs Linear Array Detector
Experimental Technique
Initial Experiments • Equipment designed for reflectance and absorbance – Silica sand chemistry and distribution influence reflectance behavior
Silica Sand GFN 110 Silica Sand GFN 70
Lake Sand GFN 55
Clay Characterization Western vs. Southern
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Clay Characterization Baseline NIR Spectra Western Bentonite shows slightly higher absorbance
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Influence of Clay on Reflectance Southern Bentonite Green Sand Decreasing Reflectance Trend in Spectra As Clay Content Increases
Influence of Water 9% Western & 2.6% Water
Experimental Plan First Analysis • Prepare Green sand mixture at 3, 6, 9, and 12% bentonite levels • Vary water content of 1, 2, 3, and 4%
• Develop calibration curves from linear partial least squares analysis (PLS1)
Typical Raw NIR Spectra 12% Western with Varying Water
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Calibration Database Clay Content Example
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Calibration Database Water Content Example
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
PLS1 Models R2
# of factors
1100-2300 (full)
0.9979
18
B %Clay
1400-2160
0.9960
18
C %Clay
1200-1700
0.9956
30
A %H2O
1100-2300 (full)
0.9930
26
B %H2O
1650-1875
0.9917
16
Model
Spectral Region
A %Clay
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
PLS1 Model Verification Sample
%Clay A Actual %Clay
B %Clay
C %Clay
%Moisture Actual
A %Moisture
B %Moisture
1
9.00
7.60
6.89
8.25
2.35
3.00
2.70
2
6.00
5.46
5.86
5.86
2.50
3.07
2.86
3
12.00
8.25
8.23
9.43
2.25
3.23
2.52
4
3.00
7.29
6.67
7.47
3.00
2.74
2.52
5
12.00
9.32
8.77
8.31
4.14
3.51
3.57
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Observations of Analysis • High number of factors is detrimental to prediction capabilities – Improved reliability when factors are centered on predicting variable • PLS1 model for clay was fair – Outer limits were poor (3 and 12%)
• PLS1 model for water was satisfactory – Possible clay influence
Experimental Plan Second Analysis • Repeat design of experiment on first attempt • Baseline all curves to reduce potential noise
Base Lined Spectra Clay and Water Considered • Curves show separation at 1425 and 1920 nm peaks
Baseline Spectra Example Varying Western Bentonite at 2% Water Change in 1420 and 1920 nm Peak Heights
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Baseline Spectra Example Varying Water at 9% Western Bentonite Change in 1420 and 1920 nm Peak Height and Width
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
1920 nm Spectral Peak 2% H2O
3% H2O
4% H2O
H
L
Large Peak Ratio
1950 1750 1550
1350 1150
5
7
9
11
Western Bentonite (wt%)
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Findings on Composition • Chemistry and distribution of sand grains influence NIR reflectance intensity
• Southern and western bentonite can be distinguished. • Water is associated with plateau between 1725 and 1860 nm
Findings on Predictive Model • Simple PLS1 models can be tuned to a specific clay and water content, loses accuracy when limits are large – Water appears to most influential in affecting limits – Difficulty in simultaneously differentiating influence of water and clay
Findings on Predictive Model • Linear relationship found when determining the width to height ratio at 1920 nm spectral peak
Recommendation on NIR • Duality effect of clay and water – Further investigation to W/H Ratio – Differentiation of curves might yield more suited data
• Inclusion of other green sand components – Might amplify the water/clay duality effect
For additional information, please contact: • Dr. Scott Giese • University of Northern Iowa Cedar Falls, IA 50614-0178 • Phone: 319-273-7083 • Fax: 319-273-5818 • Email: [email protected]
Research Objective • Green sand molding characterization for process control – Numerous tests • Development of control charts by Heine • Fairly slow for results for some tests • Too generic in some instances
• Develop on-line method for determining compositional components
Compositional Components • Major components – Clay • Serves as bonding agent and determines strength
– Water • Actives clay and varies strength
• Minor components – Additives such as seacoal
– Recycled core materials
Near-Infrared (NIR) Spectroscopy Wavelength = 780 – 2500 nm (12500 – 4000 cm-1) Combination and overtones of –CH, -OH, and –NH vibration Normal Mode Vibration (MIR)
High degree of Excitation
Combination and Overtones (NIR)
Low degree of Excitation
Source - http://www.docstoc.com/docs/79530219/Near-Infrared-Spectroscopy
Characteristics of NIR • Reflection intensity stronger • Silica can be used as transmittance materials
• Sample preparation not required • NIR spectra characteristic to material
Example of Bentonite Stretching and Bending -OH Stretch
Absorbed Water -OH Stretch Structural Hydroxyl Groups
Dispersive Spectrometer Fiber-Optic Wavelength 1000 – 2500 nm with resolution 10-15 nm InGaAs Linear Array Detector
Experimental Technique
Initial Experiments • Equipment designed for reflectance and absorbance – Silica sand chemistry and distribution influence reflectance behavior
Silica Sand GFN 110 Silica Sand GFN 70
Lake Sand GFN 55
Clay Characterization Western vs. Southern
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Clay Characterization Baseline NIR Spectra Western Bentonite shows slightly higher absorbance
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Influence of Clay on Reflectance Southern Bentonite Green Sand Decreasing Reflectance Trend in Spectra As Clay Content Increases
Influence of Water 9% Western & 2.6% Water
Experimental Plan First Analysis • Prepare Green sand mixture at 3, 6, 9, and 12% bentonite levels • Vary water content of 1, 2, 3, and 4%
• Develop calibration curves from linear partial least squares analysis (PLS1)
Typical Raw NIR Spectra 12% Western with Varying Water
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Calibration Database Clay Content Example
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Calibration Database Water Content Example
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
PLS1 Models R2
# of factors
1100-2300 (full)
0.9979
18
B %Clay
1400-2160
0.9960
18
C %Clay
1200-1700
0.9956
30
A %H2O
1100-2300 (full)
0.9930
26
B %H2O
1650-1875
0.9917
16
Model
Spectral Region
A %Clay
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
PLS1 Model Verification Sample
%Clay A Actual %Clay
B %Clay
C %Clay
%Moisture Actual
A %Moisture
B %Moisture
1
9.00
7.60
6.89
8.25
2.35
3.00
2.70
2
6.00
5.46
5.86
5.86
2.50
3.07
2.86
3
12.00
8.25
8.23
9.43
2.25
3.23
2.52
4
3.00
7.29
6.67
7.47
3.00
2.74
2.52
5
12.00
9.32
8.77
8.31
4.14
3.51
3.57
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Observations of Analysis • High number of factors is detrimental to prediction capabilities – Improved reliability when factors are centered on predicting variable • PLS1 model for clay was fair – Outer limits were poor (3 and 12%)
• PLS1 model for water was satisfactory – Possible clay influence
Experimental Plan Second Analysis • Repeat design of experiment on first attempt • Baseline all curves to reduce potential noise
Base Lined Spectra Clay and Water Considered • Curves show separation at 1425 and 1920 nm peaks
Baseline Spectra Example Varying Western Bentonite at 2% Water Change in 1420 and 1920 nm Peak Heights
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Baseline Spectra Example Varying Water at 9% Western Bentonite Change in 1420 and 1920 nm Peak Height and Width
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
1920 nm Spectral Peak 2% H2O
3% H2O
4% H2O
H
L
Large Peak Ratio
1950 1750 1550
1350 1150
5
7
9
11
Western Bentonite (wt%)
118th Metalcasting Congress April 8-11, 2014 – Schaumburg, IL USA
Findings on Composition • Chemistry and distribution of sand grains influence NIR reflectance intensity
• Southern and western bentonite can be distinguished. • Water is associated with plateau between 1725 and 1860 nm
Findings on Predictive Model • Simple PLS1 models can be tuned to a specific clay and water content, loses accuracy when limits are large – Water appears to most influential in affecting limits – Difficulty in simultaneously differentiating influence of water and clay
Findings on Predictive Model • Linear relationship found when determining the width to height ratio at 1920 nm spectral peak
Recommendation on NIR • Duality effect of clay and water – Further investigation to W/H Ratio – Differentiation of curves might yield more suited data
• Inclusion of other green sand components – Might amplify the water/clay duality effect
For additional information, please contact: • Dr. Scott Giese • University of Northern Iowa Cedar Falls, IA 50614-0178 • Phone: 319-273-7083 • Fax: 319-273-5818 • Email: [email protected]
