Résumé
In most cases, industrial mixed flow corn dryers are controlled manually or by using very simple heuristic procedures, with a relatively poor performance. A detailed analysis, carried out using both a first principles model and a black-box approach, shows important non-linearities and gives a better insight into plant dynamics which may explain difficulties in using conventional linear control. Prior knowledge is used to derive a combined logarithmic and variable sample period co-ordinate transformation which is shown to linearize the plant exactly under reasonable simplifying assumptions. The resulting linear model has a rather special structure and better approximation capabilities than common transfer functions with many more estimated parameters. The control implications of the model are analysed. A simple PI controller is shown to be a suitable choice for an industrial application and significant improvements cannot be expected from more sophisticated designs such as LQG.