Shen YZ, Du CW, Zhou JM, Ma F. Release profile predictions of controlled release fertilisers: Least Squares Support Vector Machines. Biosystems Engineering, 2018, 172: 67-74
Abstract
Modelling nutrient release profiles from a population of controlled release fertiliser (CRF) is essential for synchronising their release with plant nutrient requirements. A mechanistic model succeeds in describing the release profiles from a single granule. Yet, a large deviation will appear when it is used for a population of CRF due to the large variation in a CRF population. Least Squares Support Vector Machines (LS-SVM) that incorporated the variation were used to predict nutrient release from a CRF population using the photoacoustic spectra of the coating, coating percentage, and frequency distribution of granule radii as inputs. The predicted release profiles were in good agreement with observations both for ‘S’ and inverse ‘L’ release patterns. LS-SVM model was effective and accurate in modelling nutrient release from a CRF population.
