University of Cagliari Master Science in Chemical and Process Engineering Statistical control of FTIR measurements in commercial detergents production Supervisor: Ing. Massimiliano GROSSO Co-supervisor: Student: Ing. Vincenzo GUIDA Alessandra TARIS Scientific committee: in collaboration with Prof. Ing. Roberto BARATTI 2011-2012
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
University of Cagliari
Master Science in Chemical and Process Engineering
Statistical control of FTIR measurements in commercial detergents production
Supervisor:Ing. Massimiliano GROSSO
Co-supervisor: Student: Ing. Vincenzo GUIDA Alessandra TARIS
Scientific committee: in collaboration withProf. Ing. Roberto BARATTI
2011-2012
• potassium hydroxide
• surfactants (anionic, amphoteric, non ionic)
• Chelating agents
• sodium carbonate
• perfume
• ethanol
• Fatty acid
• polymers
• etc.
Aim: ensure standard quality in detergents
Focus on surfaces detergents
Complex formulations containing:
• ingredients mixing
• packaging
• quality control
Problems:
• Interpretation and manipulation of collected process variables may be difficult
• Online quality control is not always feasible
• Analytical techniques are slow (e.g. concentration measurements)
Steps in liquid detergents production:
Process deviations due to compositionvariations of detergent
FTIR spectroscopy: fast analytical technique, can be used online
Reproduction using a 142 samples set ofdetergentJoint variation of 11 experimentalconditions (compounds concentration)
Samples FTIR spectra
NPNN
P
P
PN
yyy
yyy
yyy
Y
21
22221
11211
N=142, P=1738Numero d'onda (cm-1)
Assorb
anza
Numero d'onda (cm-1)
Assorb
anza
Experimental campaign (P&G, Bruxelles)
142 spectra1738 absorbances for each spectrum
Samples FTIR spectra
Deviations reflect on spectra
Problem: How can weidentify samples differencesusing spectra analysis?
Thesis aims:
1. Development of methods for statistical control on experimentalmeasurements (spectra) using Multivariate StatisticalTechniques (to be implemented online in the future)
2. Detect compounds that significantly affect the spectra
Numero d'onda (cm-1)
Assorb
anza
PCA goals: data compression, informations extraction
Example: Bidimensional case-study (x1-x2 set)
Original variables Principal components (PC)
High dimensionsExtremely correlated
• PC1 greatest variance
• PC2 residual variance
• PC1 and PC2 indipendent (orthogonal)
fewerindipendent
x1
x2
x1
x2
x1
x2
PC1
PC2
Score1 (t1): projections on PC1 direction
Score2 (t2): projections on PC2 direction
Sscore1>>Sscore2PCA model : only one principal component (PC1)
Data coordinates in the newspace: scores (T)
Scores variance:
PC1
PC2
x1
x2
x1
x2
x1
x2
x1
x2
PC2
PC1
Hotelling T2
Measures distance from O′within PCA model
Q Statistic
Measures sample distance from PCA model
(that is from orthogonal projection on PC1 line)
Bidimensional case-study: 2 samples supposed to be out-of-control
Out-of-control samples identification using Q and T2 statistics