1. Bridging the Lab – Process Line Gap C
Bridging the Lab – Process Line Gap C. Rechsteiner Chevron ETC, Richmond, CA B. Rohrback Infometrix Inc., Bothell, WA

2. Current State
In current control applications, there is a clear preference to obtain the minimum number of process measurements that allow one to control the process.
“I don’t want too many measurements, they make my model unstable.”
For process GC applications, one either measures a few discrete components, or you need an assist from the local support laboratory.

3. Future Trends What Why Miniaturized instrumentation
Changing regulations
Changing products
Convergence of labs and process analytics
Resource limitations
manpower,
skill sets,
materials, …
Miniaturized instrumentation
Fast instrumentation
Remote access to sample points
Abundant data rich analyzers
More precise control needed

4. This will make our jobs manageable!
Future Solutions
To meet these challenges, we need:
to implement analyzers capable of measuring more critical process parameters;
to extract more information from those analyzers;
to better utilize computational advances; and
to put more smarts into our analyzers.
This will make our jobs manageable!

5. Implications
This means that:
Instruments must
be smarter
respond quicker
make data rich measurements and
smartly reduce the data to a reasonable number of “model-able” parameters.
Instruments must heal themselves (or at least act as a diagnostician)
Instruments should be the “same” in either the process or the laboratory environment

6. Common platform that spans all
The Needed Tools
Robust instruments
Fast spectral or chromatographic alignment
Fast pattern recognition with heuristics to determine how steady, steady-state is
A knowledge base allows recognition of common instrument faults and communicates the corrective actions to the appropriate party
Common platform that spans all
data-rich analyzers

7. Spanning the variety of gas chromatographic systems
Steps Along the Way
The Hardware
Spanning the variety of gas chromatographic systems

8. Different GCs Giving Similar Results

9. Must Have… a Good … System

10. Must Have… a Good Sampling System

11. Must Have… a Good Sampling-Control System

12. Must Be Self Contained

13. Must Have… a Good Control Shed

14. Steps Along the Way The Alignment Advantage
Seeing process detail that would otherwise be missed.

15. On-line SimDis (Siemens Maxum II GC)
400 Samples Un-Aligned
Same Samples Aligned

16. On-line simulated distillation
The plot overlays 400 chromatograms collected over 6 days
aligned

17. Comparison of PCA scores
Before alignment
After alignment
85% of all of the variation in the raw data is due to the misaligned peaks.
Correcting for this shows us that there are three different production regimes in these data.

18. Why Simulated Distillation (SimDis)?
The primary refinery separation process is distillation.
Physical distillation can
take significant time,
requires a largish sample
skilled manpower,
and so-so reproducibility.

19. Why Simulated Distillation (SimDis)?
SimDis can be done
at/near-line with small samples,
good reproducibility in reasonable time,
and, if there are no problems, little manpower.
SimDis retains the data-richness of chromatographic methods, which can be exploited.
SimDis performance is fairly well understood and measurable.

20. The case for alignment - 1
Unaligned Chromatograms

21. The case for alignment - 2
Boiling Point Calibrated Chromatograms

22. The case for alignment - 3
Chemometric Aligned Chromatograms

23. The case for alignment - 4
Yield curve comparison for the 12 runs
BP Calibrated
Chemometric Alignment

24. The case for alignment - 5
Data Bias

25. The case for alignment - 6
Data Bias - Closeups

26. Aligning Multiple Instruments20 40 60
Time (seconds)
Raw data 20 40 60
AutoAligned
Time (seconds)
The above chromatogram shows runs of a C8 to C19 hydrocarbon mixture on three instruments. Although the run-to-run variability is small on a single instrument, there are differences among the three instruments.

27. Steps Along the Way Building the Knowledge Component
Seeing process detail that would otherwise be missed

28. PCA for Interpretation
Reformates
Alkylates
Naphthas

29. PCA of Aligned Chromatograms
Alkylates
Naphthas
Reformates

30. Working with Data-Rich Measurements Simulating in the Laboratory
Steps Along the Way
Working with Data-Rich Measurements
Simulating in the Laboratory

31. Using DHA Reports as a Data-Rich Source
Winter gasoline
After the alignment and DHA steps have been completed, it may be useful to perform another multivariate prediction on the DHA report to confirm the original material identity.

32. Identifying the Big Problems
Outliers:
Instrument problem?
Process upset?
Stream Error?
95% confidence interval
Two additional samples were run and compared to the model. In this case, these gasolines show a pattern in the report table that is statistically-different from our expectations. The direction of the new points compared to the mass tells what peaks are responsible for this excursion.

33. Where Are We? The pieces are coming together!
We have made progress towards implementing a novel micro-GC for Simulated Distillation.
The unique trapping approach of this system is compatible with the SimDis applications.
Chemometric alignment will be essential for data-rich measurements to assure consistent data.
Chemometric alignment has value even for a low resolution chromatographic techniques.

34. Where Are We? The pieces are coming together!
Alignment can be automated for plant use.
Alignment and chemometric identification techniques can provide effective analysis of complex data at the process line.
These techniques can reduce the burden on highly skilled manpower to interpret complex data.