Nondestructive Detection of Moisture Content in Palm Oil by Using Portable Vibrational Spectroscopy and Optimal...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Accurate moisture content determination is crucial for crude palm oil (CPO) quality and shelf-life.
• Traditional methods for moisture analysis can be time-consuming and destructive.
• Developing rapid, non-destructive techniques is essential for efficient CPO quality control.

Purpose of the Study:

• To develop and optimize a quantitative model for rapid, non-destructive moisture content measurement in CPO.
• To compare various spectral preprocessing techniques and Partial Least Squares (PLS) regression algorithms.
• To identify the best performing model for predicting moisture levels in CPO.

Main Methods:

• Utilized a micro Near-Infrared (NIR) spectrometer to acquire spectral data from CPO samples.
• Applied several spectral preprocessing methods: Standard Normal Variant (SNV), Multiplicative Scatter Correction (MSC), Savitzky-Golay first derivative (SGD1), and second derivative (SGD2).
• Compared different PLS regression techniques including full PLS, interval PLS (iPLS), synergy interval PLS (SiPLS), genetic algorithm PLS (GAPLS), and successive projection algorithm PLS (SPA-PLS).

Main Results:

• The optimal model, combining SGD1 preprocessing with SiPLS regression (SGD1+SiPLS), achieved high performance.
• Calibration set performance: coefficient of determination (Rc) = 0.968, root mean square error of calibration (RMSEC) = 0.468.
• Prediction set performance: coefficient of determination (Rp) = 0.956, root mean square error of prediction (RMSEP) = 0.361.

Conclusions:

• Rapid and non-destructive determination of moisture content in CPO is feasible using micro NIR spectroscopy.
• The SGD1+SiPLS model provides a robust and accurate method for CPO moisture analysis.
• This technique can significantly facilitate quality control processes in the CPO industry.