[Back]

M. Stadlbauer, C. Hametner, S. Jakubek, T. Winsel:
"Analytic Multilayer Perceptron based Experiment Design for Nonlinear Systems";
Poster: 18th IFAC World Congress, Milan; 2011-08-28 - 2011-09-02; in: "Proceedings of the 18th IFAC World Congress, 2011", 18th IFAC World Congress / Volume# 18 / Part# 1 (2011), ISBN: 978-3-902661-93-7; 4332 - 4337.

Data based modelling requires informative input and output process data gained from experiments in order to parametrize a model. Model based design of experiments is targeted to make experiments as informative as possible so as to reduce experimental effort by simultaneously gathering all relevant information of the process.
In this paper a novel analytic method for optimal model based design of experiments is presented, which can be applied to either nonlinear static or nonlinear dynamic systems.
The analytic optimisation of the proposed method is based on multilayer perceptron networks. The presented DoE approach also considers the incorporation of input signal constraints and its effectiveness is demonstrated by means of a nonlinear static and a nonlinear dynamic simulation example. A comparison with state of the art DoE methods is given.

DoE, A/D - optimality, Fisher information matrix, nonlinear systems, perceptron network, optimal experiment design