Abstract
It is important that combine harvesters can operate efficiently under various crop and weather conditions. The Taguchi Method has been recognized as a leading tool to improve the robustness of control systems for many kind of machinery. The Taguchi Method utilizes Signal to Noise ratio to represent the factorial effects of control variables against outputs and also takes noise factors into consideration. A key tenet of the Taguchi Method is that control variables with strong cross interactions should not be included in design parameters. In implementing the Taguchi Method it is necessary to select control variables without strong cross interactions. Experimentally identifying unwanted variables is impractical due to the enormous number of control variables combination that must be examined. For this reason the Taguchi Method is not used in combine harvesters. Recently, Artificial Neural Networks (ANN) have been utilized to describe various non-linear systems. The performance of combine harvesters can be considered as a typical non-linear agricultural machinery system. In this paper, ANN were used to model non-linear combine harvester system in order to reveal hidden cross-interactions between control variables. After revealing hidden cross-interactions between control variables, the factorial effects were obtained based on an L18 matrix. A confirmation run proved the reliability of the factorial effects obtained. Combining the Taguchi Method and ANN was proved to be effective.
