On-the-go throughput prediction in a combine harvester using sensor fusion

The paper addresses design of a clean grain throughput observer for a combine harvester, i.e. delay free yield sensing. The aim is to predict grain throughput changes using the forward speed and a throughput sensor in the feederhouse. By utilising a grain flow model and sensor fusion an estimate of the current grain throughput is obtained, hence the effect from the lag in the momentary yield sensor reading due to material transport delays can be reduced. Statistical change detection is used to detect feederhouse load condition as well as sensor discrepancies using the observer innovation signal. The system is able to predict changes originating from forward speed and local crop density variations. Also temporary sensor discrepancies are detected and compensated in the grain flow estimate.