Algorithms to predict moisture content of grain using relative humidity time-series

Abstract

Post-harvest losses to grain crops are conservatively estimated at 10-20% (ranging up to 40%) in many countries. In particular, grains must be properly dried to avoid spoilage, harmful mycotoxins from mold, and financial loss. Smallholder farmers can thus greatly benefit from a means to assess Moisture Content (MC) in their grain. We describe a two-step algorithm, with very low computational cost, that calculates MC with high accuracy, using Relative Humidity (RH) and Temperature (T) time-series. The time-series do not need to reach equilibrium state, enabling fast (12-minute) time-to-result. The algorithm first curve-fits the RH time-series to estimate asymptotic RH, in order to leverage the physics of the RH-T-MC equilibrium relationship. It then uses regression to estimate MC to within ±1% on ≥95% of samples over a wide range of ambient RH-T conditions, on both Lab and Field samples of 10 different grains.