Enhanced discrimination and calibration of biomass NIR spectral data using non-linear kernel methods

Area of Science:

• Biomass characterization
• Spectroscopy
• Chemometrics

Background:

• Accurate characterization of biomass is crucial for efficient energy utilization.
• Traditional methods for determining ash and char content are time-consuming.
• Rapid analytical techniques are needed to support biomass-to-energy applications.

Purpose of the Study:

• To develop rapid methods for characterizing biomass for energy purposes.
• To evaluate near-infrared spectroscopy (NIRS) for predicting ash and char content in diverse biomass types.
• To compare various multivariate statistical approaches for optimal model performance.

Main Methods:

• Near-infrared spectroscopy (NIRS) was employed for biomass analysis.
• Multivariate statistical techniques including Principal Component Analysis (PCA), Orthogonal Signal Correction (OSC), Partial Least Squares (PLS), and Kernel PCA were investigated.
• The OSC-treated Kernel PLS method was specifically tested for its efficacy.

Main Results:

• Highly accurate predictive models for ash and char content were developed using NIRS, independent of biomass type.
• Several chemometric methods were evaluated for their ability to classify and predict biomass characteristics.
• The OSC-treated Kernel PLS method demonstrated superior performance, achieving the highest correlation coefficient and lowest Root Mean Square Error of Prediction (RMSEP).

Conclusions:

• NIRS combined with multivariate analysis provides a robust and rapid method for biomass characterization.
• The OSC-treated Kernel PLS model is highly effective for predicting ash and char content in various biomass samples.
• This approach facilitates efficient assessment of biomass for energy utilization.