An on-the-go monitoring algorithm for separation processes in combine harvesters

Modelling the internal process of a combine harvester is a difficult task. A large number of parameters influence the process, which is characterised by highly non-linear and stochastic phenomena. Therefore, it is useful to install a monitoring system to measure those process parameters that are directly related to instantaneous process behaviour. A New Holland CX820 conventional combine harvester was equipped with extra sensors to measure the actual cutting width, crop throughput and separation. Measurements were performed on a field with varying crop conditions and a feasibility study was carried out to estimate the instantaneous separation performance. A two-step strategy is proposed to design recursive algorithms that estimate the grain separation performance on-line based on an off-line non-linear correlation analysis and a subsequent on-line quasi-Newton optimisation method. Both off-line and on-line algorithms were applied on a measured data set and illustrate the relation between the separation behaviour and the local field conditions. The presented off-line analysis resulted in a more accurate time delay estimation of the dynamic crop throughput-loss relation, while the on-line method revealed a site-specific impact of local field topography on instantaneous machine efficiency. (C) 2004 Elsevier B.V. All rights reserved.