Estimating cocoa bean parameters by FT-NIRS and chemometrics analysis.

Area of Science:

• Agricultural Science
• Analytical Chemistry
• Spectroscopy

Background:

• Cocoa bean quality is crucial for the food industry.
• Rapid and accurate analysis methods are needed for quality assurance.
• Traditional methods can be time-consuming and labor-intensive.

Purpose of the Study:

• To develop a rapid analytical method for cocoa bean quality assessment.
• To utilize Fourier transform near infrared spectroscopy (FT-NIRS) for this purpose.
• To apply chemometric techniques for data analysis and model development.

Main Methods:

• Fourier transform near infrared spectroscopy (FT-NIRS) was employed for spectral data acquisition.
• Chemometric techniques, including back propagation artificial neural network (BPANN) and synergy interval back propagation artificial neural network regression (Si-BPANNR), were used.
• Model performance was optimized using cross-validation and evaluated on a prediction set.

Main Results:

• An optimal identification model using BPANN achieved 99.73% accuracy for cocoa bean quality categories.
• The Si-BPANNR model demonstrated high efficiency for estimating pH (Rpre=0.98, RMSEP=0.06) and fermentation index (FI) (Rpre=0.98, RMSEP=0.05).
• Variable selection using Si-BPANNR improved the accuracy of pH and FI estimations.

Conclusions:

• FT-NIRS, coupled with BPANN and Si-BPANNR models, provides a robust and efficient approach for rapid cocoa bean analysis.
• This technique can be successfully implemented for quality control and assurance in the cocoa industry.
• The study highlights the potential of spectroscopic and chemometric methods in agricultural product analysis.