Prediction of nitrate release from polymer-coated fertilizers using an artificial neural network model

Polymer-coated fertilizers (PCF) are currently the most popular controlled-release fertilizers, and offer great advantages over conventional fertilizers. To understand the release of PCF, a generalized regression neural network (GRNN) was used to predict nitrate release profiles. A total of 30 PCFs were used for the experiment, and nitrate release profiles in water were obtained to train the GRNN model. Input vectors of the model were controlled-release parameters, including membrane thickness, temperature, granule radius and saturated concentration of the nutrient, and output vectors of the model were nitrate release profiles. The results showed that the predicted values were in fairly good agreement with the observed ones, and the performance of the GRNN model was superior to a theoretical model. The GRNN model revealed that the thickness of coating membrane was the most important parameter in controlling nitrate release, followed by temperature, granule radius and saturated concentration of nitrate. The GRNN model was a useful tool in solving non-linear prediction problems in the development of PCF, and the nitrate release profile of PCF could be optimized with GRNN by adjusting controlled-release parameters, which provides an alternative method to achieve a PCF product with the desired nutrient release characteristics.