Analysis of port wines using the electronic tongue

1
Analysis of port wines using the electronic tongue

• Alisa Rudnitskaya1, Ivonne Delgadillo2, Andrey
  Legin1, Silvia Rocha2, Anne-Marie Da Costa2,
  Tomás Simoes3
• 1Chemistry Department, St. Petersburg
  University, Russia
• www.elecronictongue.com
• 2Chemistry Department, University of Aveiro,
  Portugal3Instituto do Vinho do Porto, Porto,
  Portugal

2
Port wine making procedure
3
Port wine producing region
4
Port wine producing region
5
Port wine styles
Ruby Bottle aged
Tawny Cask aged
Ruby, Ruby reserve (2-3 years in the cask)
Tawny, Tawny reserve (min 6 years in the cask)
Tawny with an Indication of Age (10, 20, 30 or 40
years in the cask)
LBV (4-6 years in the cask)
Colheita (min 7 years in the cask)
Vintage (2-3 years in the cask)
6
Purpose of the study

• Development of the rapid analytical methodology
  for the assessment of the port wine age
• Evaluation of the electronic tongue multisensor
  system (ET) for the determination of the port
  wine age
• Comparison between ET and conventional chemical
  analysis data for the determination of the port
  wine age
• Evaluation of the orthogonal signal correction
  for the data filtering

7
Experimental

• Samples
• 146 samples of port wine, in particular, wines
  aged in oak casks for 10, 20, 30 and 40 years,
  Vintage, LBV and Colheita (harvest) wines of age
  varying from 2 to 70 years.
• All port wine samples together with chemical
  analysis results were obtained from Instituto do
  Vinho Do Porto

• Measurements
• Electronic tongue
• Sensor array of 28 potentiometric chemical
  sensors with both chalcogenide glass and
  polymeric membranes
• Direct measurements without sample preparation
• Chemical analysis using conventional analytical
  techniques (provided by Instituto de Vinho de
  Porto)
• 32 parameters including content of sugar (ºBé),
  ashes, reducible sugar, total SO2 and sulphates,
  tartaric and malic acids, alcohols (ethanol,
  methanol, glycerol, 1 and 2-butanol, 1-propanol,
  isopropanol, amyl and allyl alcohols), ethanal,
  ethyl acetate, volatile and total acidity, Foline
  index, density, dry extract, etc.

8
ExperimentalData processing

• PCA
• Recognition of samples and data exploration
• PLS regression
• Calibration models for prediction of port wine
  age
• ET and chemical analysis data
• Raw and OSC filtered data
• Test set validation, 1/3 of the samples were used
  as tests
• OSC
• Applied for filtering of ET and chemical analysis
  data
• Software used
• Unscrambler v. 7.8 by CAMO AS
• SIMCA-P v.11.0 by Umetrics

9
Orthogonal Signal Correction

• Wold S, Antti H, Lindgren F, Öhman J,
  Chemometrics Intell Lab. Syst. 44 (1998) 175-185
• Aim removal of variation in X that is not
  correlated with Y prior to modeling
• to Xwo, which is orthogonal to Y AND provides
  good modeling and prediction of X
• po' toX
• XOSC X Stopo

10
PCAChemical analysis data

• Good correlation between chemical analysis data
  and port wine age
• Clustering according to port wine type good
  separation between blended tawnies and LBV and
  vintage wines

11
PCAET data

• No good separation of port wines according to
  their age
• Clustering according to port wine type
• Significant temporary drift in the data

12
Prediction of the port wine agePLS regression on
the raw data
ET
Chemical analysis
PCs in the model - 2 RMSEC 5.3 RMSEP 5.4
13
OSC filtering of the data
14
OSC filtering of the dataRMSEP
ET data
Chemical analysis data
15
Effect of OSC filtering of ET data
16
Effect of OSC filtering on ET data
17
Conclusions

• Port wine age can be predicted using both
  electronic tongue and conventional chemical
  analysis data with the same precision of about 5
  years.
• Electronic tongue response has shown a temporary
  drift in port wines, especially pronounced during
  first days of measuring session
• Data pretreatment using OSC was favorable for ET
  data successfully removing time dependence and
  producing improved calibration models
• Port wine sample can be separated according to
  their types using both ET and conventional
  chemical analysis data.