Advances in GC with Alignment

for Real-Time Decision Making Brian Rohrback and John Crandall

October 31, 2010

CPAC Workshop

Poll of Process Users

1. Analytical failure prediction

2. Result validation

3. More process-specific information (timely, higher quality,

more focused)

4. Simplification of procedures

5. Elimination of analytical discrepancies

6. Reduction in the lifecycle cost (cost of ownership)

Driven in part by dwindling manpower,

skill sets and capabilities !

2

Rethinking Gas Chromatography

Application Efficiency

Flexibility/Ease of Use

Multiple applications

Lab or Line

Speed of analysis

Seconds or a small number

of minutes

Automated interpretation

Lifecycle Cost

Price paid

Installation

Size

Simplicity

Maintenance

Power

Resources( supplies,

people)

Modularity

Rugged/Reliable 3

Greener Analysis

Contract analytical services company in California

Environmental, remediation, hazmat

18,000 foot2 facility, 19 GC or GC/MS instruments

AC year-round, capacity is 62 tons, accounts

for 50% of electrical use. Summer electric use

5,000 kWh/day

“Air conditioning is our biggest maintenance problem.”

Mike Brech & David Tsubota of BSK Analytical

A 6890 GC consumes between 2250 and 2950 kW

at peak draw.

Air conditioning requires 40% of that energy to

remove the waste heat.

Power delivery is 33% efficient

5

I assume the 6890 averages 1kWh and include air

conditioning at 0.4 kWh, so,

A conventional GC in the lab requires 4.2 kWh of

production at the power plant,

peak at 12.6 kW!

Delivering Information

Just having the measurements does not translate into control

Remember, there are not enough skilled technicians to handle even

the current workload.

Chemometrics solves the information processing problem with 2

technologies:

Alignment enables us to sell instruments that have vastly-lower

calibration requirements.

Interpretation algorithms automates the generation and the

qualification of the information derived from the raw data.

And if we can make all of our instruments

look as much alike as possible.

Interchangeability

common interpretive base 6

Raw process chromatograms

Full Data;2

0 10 20 30 40

Time Index (E +03)

0.0

0.1

0.2

Re

sp

on

se

7

The same chromatograms after

alignment

Full Data;2

0 10 20 30 40

Time Index (E +03)

0.0

0.1

0.2

Re

sp

on

se

8

Chemometrics for instrumentation:

the value proposition

All this results in the ability to make the most of the data you are

collecting and enables

Continuous validation of the instrument, possibly the entire

process

A vast improvement in the ability to automatically interpret the

stream of data, leading to better feed-back and feed-forward

control

A better ability to maintain the process, the instrument and the

multivariate model.

9

Software Goal: continuous validation of

the system performance and guarantee

the quality of the data.

10

The data to information

transition

Raw data in the C15 to C19 region

for ten oils run in duplicate

C15 to C19 region after retention

time alignment

C15 to C19 region after area

normalization and alignment

QC of x-ray contrast agents

14

Original HPLC data

Aligned HPLC data

15

QC of x-ray contrast agents

16

QC of x-ray contrast agents

17

Amino acid analysis

Capillary

Electrophoresis

18

Capillary Electropherograms

19

Gating Problem

20

Gating Problem Solved

21

Biodegraded crude oil

Raw data

20 40 60 Time (seconds)

Chromatographic Alignment

3 instruments

23

Auto-Aligned

20 40 60 Time (seconds)

Chromatographic Alignment

3 instruments

24

Automated alignment works

5 year period, 6 GCs

25

Automated alignment works

aligned

26

5 year period, 6 GCs

Chemometrics for instrumentation:

the value proposition

Anything you can do to improve precision of the multivariate

measurements collected by the instrument will allow you to tighten

the control – essentially for free.

One way is to construct an application-specific, objective evaluation

system:

Experimental design

Exploratory data analysis

Leading to

Multivariate modeling (qualitative and quantitative analysis)

Just as key is the signal processing aspect of chemometrics to reduce

instrument-derived variability

Within an instrument (e.g., noise reduction)

Between instruments (i.e., transfer of calibration)

27

Interpretation

Data, data and more data

Human qualities

Good at seeing small differences

Bad at quantifying small differences

Fair at recollection

Poor at seeing patterns in tables of numbers

Objectives

Quantify a property or attribute?

Characterize the sample?

28

54 samples of M. intracellulare (1

lab)

29

38 samples of M. simiae (5 labs)

30

Variation in M. asiaticum

31

PCA Scores Plot After Alignment

Samples of the same

species cluster, some

species to a greater extent

than others.

Also, species known to be

similar express similar

chromatographic profiles

and cluster near each other

in this factor space.

32

Use of decision points, hierarchical

models

Journal of Chromatographic

Science vol.32 1994 33

34

Continuous validation of a

multivariate instrument

We can correct retention times to match an application-

specific relevant sample

This eliminates the transfer of calibration problem in

chromatography

Common regression and classification algorithms can be

applied automatically to infer physical properties or

characteristics

This allows us to bring more complex analyses into on-line

use