Suhandy, Diding (2018) The feasibility of using explicit method for linear correction of the particle size variation using NIR Spectroscopy combined with PLS2 regression method. In: The 3rd International Conference on Chemical Engineering Sciences and Applications 2017 (3rd ICChESA 2017), 20-21 September 2017, Aceh.

[img] Text
Yulia_2018_IOP_Conf._Ser.%3A_Mater._Sci._Eng._334_012062.pdf

Download (654kB)
Official URL: http://iopscience.iop.org/article/10.1088/1757-899...

Abstract

NIR spectra obtained from spectral data acquisition system contains both chemical information of samples as well as physical information of the samples, such as particle size and bulk density. Several methods have been established for developing calibration models that can compensate for sample physical information variations. One common approach is to include physical information variation in the calibration model both explicitly and implicitly. The objective of this study was to evaluate the feasibility of using explicit method to compensate the influence of different particle size of coffee powder in NIR calibration model performance. A number of 220 coffee powder samples with two different types of coffee (civet and noncivet) and two different particle sizes (212 and 500 μm) were prepared. Spectral data was acquired using NIR spectrometer equipped with an integrating sphere for diffuse reflectance measurement. A discrimination method based on PLS-DA was conducted and the influence of different particle size on the performance of PLS-DA was investigated. In explicit method, we add directly the particle size as predicted variable results in an X block containing only the NIR spectra and a Y block containing the particle size and type of coffee. The explicit inclusion of the particle size into the calibration model is expected to improve the accuracy of type of coffee determination. The result shows that using explicit method the quality of the developed calibration model for type of coffee determination is a little bit superior with coefficient of determination (R2) = 0.99 and root mean square error of cross-validation (RMSECV) = 0.041. The performance of the PLS2 calibration model for type of coffee determination with particle size compensation was quite good and able to predict the type of coffee in two different particle sizes with relatively high R2 pred values. The prediction also resulted in low bias and RMSEP values.

Item Type: Conference or Workshop Item (Paper)
Subjects: S Agriculture > S Agriculture (General)
Divisions: Fakultas Pertanian (FP) > Prodi Teknik Pertanian
Depositing User: Dr. Diding Suhandy
Date Deposited: 15 Nov 2018 03:06
Last Modified: 15 Nov 2018 03:06
URI: http://repository.lppm.unila.ac.id/id/eprint/9922

Actions (login required)

View Item View Item