Principal Component Analysis as an Exploration Tool for

Kinetic Modeling of Food Quality: A Case Study of a Dried

Apple Cluster Snack

Ratna Palupi, Emmy Novia, Elivi Sofi

 Materials and Methods
Samples of the new product, a dried apple snack called Cluster. The snacks
are made from a mixture of small cubes of apples and natural fruit sugars
(the product formulation is not shown because it is intellectual property).

The mixture of apple pieces and fruit sugars is dried using hot air and then
subjected to a puffing process to shape the snack.

The resulting product (bulk density of 0.18 g/cm3 0.05) is then packed in
high barrier metalized bags (HYC Packaging Inc., Chile, water permeability
<0.3 g/m2 day at 25C and <0.8 g/m2 - day at 35C).

The samples were incubated at 18 C, 25 C and 35 C

without exposure to light. The measured attributes
The humidity in the temperature controlled chambers
1. Water activity (Aw)
was maintained at 75% RH 5% (Precision Scientific,
Jouan Inc., Winchester, USA) with air circulation. 2. Degradation Color DE
The sampling consisted of randomly removing three
3. Moisture
samples for analytical determination (in triplicate)
and six samples for use in a sensory panel. All of the 4. Sensory properties
quality attributes were measured during 17.5 months
of incubation.
(aroma, taste, texture and
color).
Results; Principal Component Analysis as an Exploration Tool for Kinetic
Modeling of Food Quality: A Case Study of a Dried Apple Cluster Snack
 ANALYSIS
To study the overall product stability, the data for all of the
quality attributes at every sampling time point were arranged
in a matrix according to each storage condition (XT matrix:
28 observations and 8 variables), where each observation is
the average value obtained from repeated measurements.

A PCA of the entire XT matrix (the data for the three storage
conditions, analyzed at the same time) and separately for
each storage temperature (XH,T matrices) was performed. All
of these computations were performed using SIMCA-
P+software (version 12, Umetrics AB, Sweden).

The XT matrices for all of the storage conditions (18C, 25C,

and 35C) were simultaneously analyzed using PCA.

The underlying global interactions between the relevant

attributes that were detected by PCA can be observed in the
loadings plot.
 Conclusion

Calculations were performed using the first

PCA related with time (68% of total
explained variability). Fig. 5 shows that the
product degradation follows first-order
kinetics for all of the storage conditions.

The univariate method considered only

moisture as the limiting attribute. This
method involved calculations based on the
worst case criteria, leading to an under-
estimated shelf-life due to higher values in
the kinetic parameters of this type of
deterioration.

The shelf-life estimated using MALST was

validated according to its compliance with
the final quality of the product which
served as the validation criteria for the
multivariate methodology.

Principal Component Analysis (PCA