Michael B Simpson, Technical Sales Manager

1
FTNIR Calibration Transport For Refinery Product
Measurements

• Michael B Simpson, Technical Sales Manager
• David Reece, Applications Scientist
• Fred Baudais, Senior Applications Scientist
• ABB Bomem, HPI Chemicals Team

2
ABB Refinery Applications
3
Gasoline Blender
n-Butane
Light straight run gasoline
Isomerate gasoline
Reformate gasoline
FCC gasoline
Gasoline Blender
Coker gasoline
Hyrodcracker gasoline
Alkylate gasoline
MTBE
Other
4
Profitability of Gasoline Blending Operation

• Factors affecting profitability are
• Optimization of blending component production
• Inventory control
• Minimizing quality giveaway
• Ever tightening legislative regulations
• FTNIR can help optimize the blending operation
  efficiency

5
Process FTNIR AnalysisTypical Performance
Characteristics
6
Analysis Stability - RON
8 hour stability
7
Uses of a Lab FTNIR to Support the On-line FTNIR
Analyzer

• Feasibility and training
• Collecting calibration spectra
• before on-line installation
• transport calibration to on-line analyzer
• Back-up to on-line analyzer
• Extend on-line analyzer calibrations

8
Calibration Transport
How difficult is this?
9
Blended Gasoline Spectrum
10

• A Chemical or Physical Property

CALIBRATION MODEL
The IR Absorbance Spectrum
11
SPECIFIC GRAVITY (GAS OIL)
12
Factors from the Training Set
Q1 Q2 Q3 Q4 Q5 Q6 Q7
FACTORS
S1

S2
SCORES
13
Absorbance Differences between Analysers will
displace samples in the Score Space, and lead to
prediction errors
S1
S3
S2
14

• First Rule of Engineering
• Theres no such thing as THE SAME
• Only HOW DIFFERENT ?

15
Difference spectra (Lab minus On-line)
16
Effect of Optical Integrity on Analyser
Reproducibility
Collimated beam through interferometer
17
Effect of Optical Integrity on Analyser
Reproducibility
Must see central fringe ONLY
Image of Point Source at Detector
18
Effect of Optical Integrity on Analyser
Reproducibility
Self Apodisation distorts bandshape
19
Effect of Optical Integrity on Analyser
Reproducibility
Off-Axis Rays cause Wavelength Shift
20
Analysis Stability - Toluene, 12 Process Lab
Analysers
21
Analysis Stability - Toluene, 12 Analysers,
under Battle Conditions
22
Gasoline Spectra Comparison Between On-Line
Analysers
23
Gasoline Spectral Variance Between On-Line and
Lab Analysers
24
Gasoline Calibration Data Set All Data Overlaid
25
Gasoline Data Set Score Projection of 2 and
5mAbs Spectral Difference
26
Gasoline Data Set Prediction Error from 5mAbs
Variance
27
Test Data for Gasoil Model
28
Gasoil Test /- 0.003 Abs
29
Gasoil Test Data /- 0.009 Abs
30
Test Data for Gasoil Model - Density
31
Test Data for Gasoil Model - Cloud Point
32
Test Data Calibration Model Transport
These data suggest that Calibration
Transportability between analysers, to give
agreement well inside the SEV of the relevant
property prediction model, will require absolute
spectral measurement differences of around 0.002
Absorbance Units.
33
Gasoline Property Prediction Differences in
Practice

• What measurement agreement is achieved when PLS
  type calibrations, used to determine gasoline
  properties on one analyzer, are transported
  without modification or correction to another
  analyzer for the same sample?
• No slope or offset adjustment
• No mapping between instruments

34
Lab to On-line Calibration Transport
35
On-line to On-line Calibration Transport
36
Calibration Transport
37
Conclusions

• Transporting calibrations between FTNIR
  instrument types (lab to on-line, on-line to
  on-line) is achievable on a production basis
  without using mathematical techniques.