Multi-objective nonlinear identification

Project members: R. Babuška, K. Maertens (Katholieke Universiteit Leuven), T.A. Johansen (Norwegian University of Science and Technology)

Sponsored by: Fonds voor Wetenschappelijk Onderzoek in Vlaanderen

Methods for the multi-objective identification of nonlinear dynamic models consisting of local linear models are being investigated. The tradeoff between global model accuracy and interpretability of the local models is explicitly considered by introducing weights on the criteria for local model accuracy. A strategy has been proposed to tune the local weights in order to achieve a similar tradeoff for each local model. In this way, the model generalization is improved. The multi-objective identification algorithm has been applied to predict the engine load of an off-road vehicle (a combine harvester) operating under varying working load conditions. The analysis tools have proven useful for the construction of an accurate and robust engine load prediction model. The resulting model can directly be used in model-based control algorithms in automatic tuning systems that explicitly deal with constraints on the working region.

Figure 4: Multi-objective nonlinear identification was used to construct a model to predict the engine load of a combine harvester.