NIR Spectrometer Calibration Curve for Oil Loss Monitoring

(Officer In-charge: Dr. Andrew Yap Kian Chung)

 

One of the efforts to increase the oil extraction rate is to reduce oil loss in palm oil mills. MPOB has developed hexane soxtec or soxhlet extraction method for oil content analysis which involves time consuming sample pre-treatment procedures. The advance of technology in near infrared (NIR) spectroscopy is a non-destructive fast analytical technique using waves in the near-infrared region of the electromagnetic spectrum wavelength from about 800nm to 2500nm without sample preparation requirement has produced proven reliable and acceptable results for samples contain such bond as CH, NH or OH with concentrations exceeding about 0.1% of total composition.

 

Description

The FOSS InfraXact 7500 NIR spectrometer equipped with ISIScan 4.2.0 and WinISI III Version 1.60 software was installed in Sime Darby R & D Palm Laboratory, Carey Island, Klang, Selangor. Samples were taken hourly from the respective sampling points of a nearby palm oil mill during the mill operation according to the Protocol for Measuring Oil Losses (Zulkifli et. al., 2010) and analysed with NIR spectrometer. The respective sample is then analysed in laboratory according to the MPOB Test Method. Graph of Laboratory reading versus NIR spectrometer is plotted for each sampling point. Statistical analysis is performed using WinISI III software to determine the accuracy and the level of confidence for each calibration curve.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FOSS Infraxact 7500 NIR spectra was acquired as an average of 32 scans between 570 and 1850 nm, with 32 scans using air as background.  Wavelength increments of 2 nm were used.  The sample data were recorded as log (R-1) where R is the reflectance energy. A sample was replicated ten times to determine the mean and standard deviation (SD) at each wavelength and then summed to determine the replicate instrumental error in NIR predictions.  Diagnostic test for wavelength (0.5±0.002 nm) were performed daily before samples were scanned.

 

 

 

 

 

 

 

 

 

 

 

 

 

Principal component analysis (PCA) was used to evaluate the respective samples spectra then subject to various mathematical processing treatments such as standard normal variant (SNV) transformation and de-trend scatter correction (D) which removes or reduces linear and quadratic curvature of each spectrum. Calibration performance based on the statistics such as standard error of calibration (SEC), standard error of cross-validation (SECV) and coefficient of determination (R2).