Application of a Neural Network to Dynamic Draft Model

A dynamic draft model is necessary to analyze mechanics of tillage and to design optimal tillage tools. In order to deal with draft dynamics, a neural network paradigm was applied to develop dynamic draft models. For the development of the models, three kinds of tillage tools were used to measure drafts in the soil bin and a time lagged recurrent neural network was developed. The neural network had a structure to predict dynamic draft, having a function of one-step-ahead prediction. A procedure for network prediction model identification was established. The results show promising modeling of the dynamic drafts with developed neural network.