Title: Monitoring of highly absorbing species in viscous and corrosive melts with inline visNIR spectroscop
1Monitoring of highly absorbing species in viscous
and corrosive melts with in-line vis-NIR
spectroscopy
- Ewan Polwart
- Process Analysis Team
- Process Technology
2(No Transcript)
3The formation and History of Avecia
INTEGRATED BIOSCIENCE BUSINESS
PHARMACEUTICALS (Merger April 1999)
Merger of Novartis Zeneca Agri
Businesses December 1999
4Avecias Core products
- Pharmaceutical Intermediates and Actives
- Biotechnology - Advanced Medicines
- Ink Jet Printing Materials
- Electronic Display Materials
- Metal Extraction chemicals
- Water-borne Resins
- Non Chlorine Pool Spa products
5Chemistry
- Reaction modifies physio-chemical properties of
CuPc - Only very slight change in shade
- Reaction gives controlled mixture of compounds
- These combined have desired properties
6Analysis Options
- Off-line analysis
- ? Sample collected and take to laboratory
- ? Expert analyst
- Analysis turn around time 1-2 h
- Requires manual handling of sample
- On-line analysis
- ? Automated analysis in reactor
- Rapid and frequent results
- No manual sampling
- ? Technically very demanding
- At-line analysis
- ? Sample analysed in local laboratory by process
operators - ? More rapid results turnaround
- Limited local analytical expertise
- Requires manual sampling
7Collecting Spectra Probing the Reaction
- Attenuated total reflectance
- Evanesant wave interacts with sample
- Only very short path length in sample
- Ideal for strongly absorbing samples
- Light absorption measured as characteristic of
wavelength - Zeiss MCS522 Spectrometer coupled to probe by
fibre optics - Ceramic bodied immersion probe with sapphire ATR
crystal - Chemically resistant
8Off-line/Reference Measurements
- Parameter 1 numerical scale describing the
total level of substitution of CuPc (assesses
total number of groups P Q). - Parameter 2 numerical scale describing the
number of Q groups attached to the CuPc molecule.
- Complex sample preparation
- Time consuming analysis
9Process Development Data
10Univariate Data AnalysisParameter 1
- Correlation between Parameter 1 and maximum
absorbance in range 400 to 500 nm. - R2 0.914 SEC/critical range 0.31
11Univariate Data Analysis Parameter 2
- Correlation between Parameter 2 and absorbance _at_
781 nm. - R20.988 SEC/Critical range 0.041
12Multivariate Data Analysis Single batch
- Parameter 1
- SG 2nd deriv, 11 pt/quad
- 3 latent variables
- SEC/Critical range 0.14
- Parameter 2
- SG 2nd deriv, 11 pt/quad
- 1 latent variables
- SEC/Critical range 0.13
13Multivariate vs. Univariate
Parameter 2
Parameter 1
14Multivariate vs. Univariate
Parameter 1
Parameter 2
15Multivariate Data Analysis Multi-batch
- Partial Least Squares (PLS) regression applied to
data set - Samples from 6 batches
- 14 for calibration
- 20 for testing model
- Preprocessing
- 2nd derivative (Savitzky-Golay, 11 pt, quadratic)
- Normalisation
- Mean centred
- Merit Stats
- Calibration
- LV 3
- SEC/Critical range 0.20
- R2 0.954
- Test
- SEP/Critical range 0.14
16Reactor
17Collecting Spectra Probing the Reaction
optical fibres to/from spectrometer
- One bounce ATR probe employed
- Probe installed in sampling baffle
- Probe only sees sample when sampler is open
- Should increase lifetime of probe
- Allows washing of probe, even during reaction
sample collected or rinse to waste
air driven sample delivery or wash
ATR probe
Sample delivery system
pneumatically actuated plate allows sample to
enter system
18Collecting Spectra Probing the Reaction
- Only collect spectra when the sampler is open
- Therefore only a limited quantity of data per
batch, i.e. no longer continuous like development
batches
19Plant Data
20Multivariate Data Analysis
- PLS
- Samples
- 11 samples from 2 batches
- 5 used for calibration
- 6 used to test model
- Pre-processing
- Apply same as for development batches
- 2nd deriv (SG, 11 pt, quad)
- Normalisation
- Mean centring
- Merit Stats
- Calibration
- LV 2
- SEC/Critical range 0.0033
- R2 0.999
- Test
- SEP/Critical range 0.21
21Future Work
- Obtain more data over coming manufacturing
campaigns - Incorporate data processing in to spectrometer
software to allow reporting of Parameters 1 2 - Side-by side validation with current off-line
methods - Implement of on-line analyser in to control regime
22Conclusions
- Successfully used ATR-vis-NIR spectroscopy in the
development lab to predict the desired end point
for reaction - Have successfully transferred the lab-based
methodology to plant reactor as a permanent
installation - Have begun process of obtaining useful
information from the plant data using chemometric
methodologies developed using the lab data
23Acknowledgements
- Ian Wells, Steve Gilligan and Karen Johnston
- Stevie Gavin and Fiona Keenan
- All on 22 Workstation on P1 plant, Grangemouth
- In particular Brian Morrison