12 Automatic Analysers in Oenology 669
mathematical treatment. The methods used to extract this information form part
of a relatively recent scientific technique, chemometrics. This has been defined by
GELADI and DABAKK, (1995) as “the science of the use of mathematical, sta-
tistical and computing methods which aim to extract useful information from data
obtained by chemical measurements”.
Analysts usually have two principal objectives, the first of these being an
exploratory method which enables spectra to be classified into pre-defined families
of compounds. Such a tool enables the sample to be identified as a must, a must
in fermentation, a dry wine, a liqueur wine or a naturally sweet wine. The second
objective involves a quantitative approach which enables the attribution of analyt-
ical values or indices particular to the wine or must on the basis of the previously
acquired reference data (calibration). It is above all this second approach which
is used by analytical laboratories where ispossible to replace classical analytical
techniques by FTIR.
There are several chemometric methods applicable which provide numerical ana-
lytical data and their implementation remains a specialist affair. They vary from sim-
ple statistical tools such as principal component analysis (PCA) to highly complex
methods such as neural networks.
All manufacturers offer optical measurement tools, the performance of which are
in general satisfactory, though methods of standardisation of the measurement, tools
for the treatment of spectral data and the quality of the chemometric tools available
are usually more variable. The quality of the chemometrics used to obtain spectral
data of a wine or a must depend on the reliability and the precision of the analytical
data.
12.5.5 Analysis of Near Infra-Red Spectra
It is useful to take into consideration the great complexity of the information con-
tained in a mid-IR spectrum of a wine or a must. This complexity arises from several
factors:
- The organic and mineral composition of these products which is one of the rich-
est found in the agri-food industry
- The absorption bands of organic molecules are multiple, particularly those aris-
ing from significant resonance effects
- There are several types interactions between the different organic compounds
resulting largely from matrix effects which are impossible to model
The procedure applied in the experimental calibration approach consists of
acquiring spectral data for the maximum of sample varieties of precisely known
values for the compound or indices to be measured, before applying mathematical
statistical methods for quantification It is essential to understand that this step is
fundamental for the construction of a high-performance analytical tool. Not only
must the operator have at his or her disposal the greatest number of representative