Mathematical model for improved mass flow estimation in the feeder housing of a forage harvester

For the implementation of precision agriculture practices, accurate yield maps are needed. These yield maps depend on accurate mass flow measurements made during harvest, measurements that are difficult to make because of variability in yield and crop conditions within each field. One way to estimate mass flow rate in a self-propelled forage harvester is to measure the displacement of feed rollers as material enters the machine. In this work, a mathematical model that captures the interaction between the feed rollers and the incoming material was developed to improve the accuracy of the mass flow estimation. The model was developed based on physical principles, various material compaction models, and discrete element method simulations. When tested with experimental data, the model performed better than the current sensor system for some parameter sets, and as well as the current sensor system for others. Because of the mathematical complexity of the full dynamic model, and in order to reduce computation time, a reduced order quasi-static polynomial model was developed to approximate the relationship between feed roller displacement and mass flow rate. Finally, the perturbation induced learning technique was investigated as a method to real-time calibration of model parameters