| dc.description.abstract | Fourier Transform Infrared (FTIR) spectroscopy combined with
chemometrics has been developed for simple analysis of lard in the mixtures with
palm oil. In this research, measurements were made on pure palm oil and that
adulterated with varying concentrations of lard (0.5–80% v/v in palm oil). Two
multivariate calibrations, namely Partial Least Square (PLS) and Principal
Component Regression (PCR) were optimized for constructing the calibration
models, either for normal spectra or its first and second derivatives. The
discriminant analysis (DA) was used for classification analysis between palm oil
and that adulterated with lard. According to result of this research, PLS has an
ability to construct the calibration model better than PCR model. PLS model at
normal spectra with frequencies at fingerprint region (1500-800 cm
) revealed
the best calibration models for predicting the concentration of adulterated lard
samples, with highest coefficient of determination (R
2
) of 0,995 and lowest Root
Mean Square Error of Calibration (RMSEC) of 1,82. In addition, validation of
this model has a good result of R
2
Leave One Out of Cross Validation (RMSEC)
of 0,916 and R
2
prediction of 0,996. DA was able to classify pure palm oil and
adulterated with lard on the basis of their FTIR spectra with accuracy 100% and
no misclassified group of samples. In this model, DA was performed at frequency
regions of 1500-800 cm
-1
using 6 principal components. According to this result
of test set validation, a level of 2% of LD can be detected in adulterated with PO.
PLS and DA developed models was further used to predict the classes of unknown
(commercial) oil in instant noodles samples. Using this model, all of local
commercial instant noodles are in the class of pure palm oil without lard, but in
the imported instant noodles samples are one sample is detected as a mixture with
the lard with a concentration of 60,03 % | en_US |
