Development of NIR spectroscopy models for starch content prediction and ethanol production from mutant grain sorghum

Abstract

The growing demands for renewable energy sources have led researchers to investigate other biomass sources, aside from maize. Grain sorghum is comparable to maize in its starch content and can be grown in regions with drier climates, where maize is a less suitable crop for these areas. In attempts to increase yield prior to harvest and for ethanol production, this study focuses on mutant grain sorghum. One hundred and nine mutant grain sorghum samples were analyzed for their chemical and physical properties and fermented into ethanol. The current method for starch analysis is time-consuming and tedious. Near infrared spectroscopy (NIR) models were developed as fast, cost-effective, and non-destructive methods for grain sorghum starch content analysis. Each mutated grain sorghum sample was scanned in a wavelength range from 4,000 to 10,000 cmˉ¹ as a whole grain and in flour form. Partial Least Squares (PLS) regression method was used for NIR model development. The coefficients of determination (R²) of 0.77 and 0.90 were achieved for starch content calibration and prediction models, respectively. This model demonstrates the possibility of a positive correlation between the actual and calculated values for starch content. Another PLS first derivative model with R² = 0.95 for calibration and a reduced wavelength range (4,000-5,176 cmˉ¹), using 39 of the original 109 samples (27 for calibration and 8 for validation), was created to predict the fermentation efficiency.