Predicting sorghum moisture content with radiofrequency dielectric functions

Accurate grain and food moisture content (mc) estimation is an issue of great economic importance in agro-food industry operations such as drying, storage, and processing. Nonetheless, there is a high level of unmet need for on-line mc sensing that could lead to more reliable grain processing automation. Dielectric models that relate product mc to its relative permittivity ε′ and loss factor ε" are considered promising solutions for continuous measurement of moisture in grain. The object of this paper was to investigate the applicability of three dielectric models derived from the function [(ε' – 1)/ε"] for mc measurement of sorghum from 7 to 23% w.b. Of the three dielectric models tested, the most satisfactory was the function [(ε′-1)/ε"]/[(2√ε′)/(1+√ε′)] at 0.1 and 1.0 MHz. However, the moisture range had to be divided into two intervals. The corresponding equations could estimate sorghum mc with standard errors of calibration of 0.6 and 0.7, and standard errors of prediction of 0.6 and 0.8 percentage point moisture, for calibration and validation sets respectively. Comparison of the results obtained here with those reported in literature revealed that the function [(ε′-1)/ε"]/[(2√ε′)/(1+√ε′)] produced mc values which are well within the performance of commercial moisture meters.