Prediction of moisture content of potash fertilizer using NIR spectroscopy

摘要

Near infrared spectroscopy (NIRS)was employed to explore the structure of reflectance spectra of red standard potash granules. Reflectance spectra were collected after samples had been conditioned to moisture contents in the range of 0-1%. Reflectanceat selected wavelengths was incorporated with multiple linear regression (MLR) to predict sample moisture contents. Models were evaluated based on their ability to predict moisture content in the validation data set, using the adjusted coefficient of determination (r_(adj)~2) and standard error of prediction (SEP). A three-regressor model, using reflectance at 1198, 1427, and 2016 nm was selected as a model with the best ability to estimate moisture content of red standard potash. For this model r_(adj)~2 and SEP were 0.92 and 0.005, respectively.