SEC =
2 Vi- Y02
n-l
Authors Permanyer and Perez are with Analytical Research Lab-
oratory, Nutrexpa S. A., c/Lepanto, 410-414, 08025 Barcelona,
J
Spain. where, Yi is the reference chemical value, yi the NIR values, and n
Table 1 -Development and performance of N/R spectroscopy calibration equations for the analysis of powdered cocoa products
Calibration Prediction
Correlation Standard error Correlation Standard error
Range Calibration Wavelengths Mathematical coefficient of calibration Range coefficient of prediction
Parameter Ph.) constants (nm) treatment (r) (SEC) w.) (d (SW
Moisture 1.5 - 4.0 KO= 3.592 1 st derivative 0.978 0.157 1.6- 2.1 0.955 0.034
Kl = 1.306 1939
Fat 1.5 - 3.5 KO= 3.840 2nd derivative 0.987 0.100 2.1- 2.7 0.842 0.051
Kl= - 10.407 2296
K2= - 2.109 1959
Sucrose 51.5 - 81.5 KO= 69.726 Ist derivative 0.998 0.526 68.0-74.5 0.962 0.680
Kl= -6.505 2080
K2= 36.671 2785
t =&xii-NIR)
WAVELENGTH (nm)
ub
IWTI where n is the number of samples and ub =
l2.K
d/1/2 (olab* + ani?) and comparing with the Students t-
distribution values with n-l degrees of freedom and a two-
sided alternative hypothesis at 01= 0.05.
The precision of the NIR method was measuredby recording
ten spectral scanson the same sample and calculating the stan-
dard deviations of the prediction data for each constituent. The
coefficients of variation were 0.86% for moisture, 1.49% for
fat and 0.61% for sucrose.
Accuracy, comparedwith referencemethods, was measured
by comparing the root mean squareprediction error or standard
error of prediction (SEP) with the standarderror of calibration
(SEC) (Moen, 1976; Miner et al., 1978; Osborne and Fearn,
2816 2875 2116 2236 1986) and by the analysis of variance (ANOVA). The ANOVA
test showed no significant difference between the results ob-
WAVELENGTH (nm)
tained by NIR and by chemical analysis (p<O.O5), meaning
62 &Jl that the NIR method was as accurateas wet chemistry methods
16.46 for moisture, fat and sucroseof the powdered cocoa products
I examined.
lB.64 In summary, a NIR spectroscopymethod was developed to
provide a routine, fast and efficient method for determining
moisture, fat and sucrose in powdered cocoa products. Cali-
bration equationswere checkedfor precision and accuracyand
good prediction results were achieved. Therefore, the NIR
spectroscopy technique could be used in quality control of
powdered cocoa products, for moisture, fat, and sucrose.
-6.G
REFERENCES
Baer, R. J., Frank, J. F., and Loewenstein, M. 1983. Compositional analy-
sis of nonfat dry milk by vsing near infrared reflectance spectroscopy. J
Assoc. Off. Anal. Chem. 66: 868.
Boehringer Mannheim. 1974. Sucrose/glucose food analysis method. In
WAVELENGTH inm) Methods of Enzymatic Analysis, Bergmeyer, H.U. (Ed.) Vol. 3, p. 1176.
Verlag Chemie Weinheim Academic Press, Inc. New York and London.
Fig. 1 -Near infrared spectra of powdered cocoa product: (a) Giangiacomo, R., Magee, J. B., Birth, G. S., and Dul!, G. G. 1981. Pre-
original spectrum; (b) 1st derivative-spectrum; (c) 2nd deriva- dicting concentrations of individual sugars in dry mixtures by near in-
tive-spectrum. frared reflectance s ectroscopy. J. Food Sci.46: 531.
IOCCC. 1952a. Met R ods of Analysis.3-F/1952. Verlag Max Glattli. Zu-
rich. Switzerland.
IOCCC. 1952b. Methods of Analysis. 8aF/1952. Verlag Max Glattli. Zu-
is the number of samples. Accuracy and precision of calibration were rich. Switzerland.
evaluated by the standard error of prediction (SEP): Kaffka, K. J., Norris, K. H., Kulcsar, F., and Draskovits, I. 1982. Attempts
to determine fat, protein and carbohydrate content in cocoa powder by
the NIR technique. Acta Aliment. 11: 271.
Miner, D. C!., Ziomek, J. V., and Landa, I. J. 1976. Near infrared instru-