Prediction of dry matter, protein, and acidity in corn steep liquor using near infrared spectroscopy

The near infrared spectroscopy methods for determining the dry matter, protein, and acidity in corn steep liquid were investigated. The samples were divided into calibration and validation subsets randomly for multivariate modeling. Outlier was eliminated carefully and cross validated by different method. The preprocessing method was tested by different combination and the most effective method in practical (1st derivative+SNV) is selected. The partial least squares (PLS) were applied in the modeling process. The coefficient of determination of validation R², the root mean square error of prediction (RMSEP) and the ratio of standard error of prediction to standard deviation (RPD) of the obtained optimum models were 0.94, 5.02 and 4.26 for dry matter, 0.93, 2.41 and 3.92 for protein, and 0.67, 0.48 and 1.75 for acidity, respectively. The results indicated that near infrared spectroscopy is a validated approach for predicting dry matter and protein of corn steep liquid rapidly and accurately.