Data-Based Dynamic Modeling

Please refer to the LSF​​​​​​​ for the most recent information.

The course which is taught in English covers the following topics:

Identification of simple models from step responses. Parameter identification: Basic idea, mathematical description of sampled systems, AXR, ARMAX and OE estimation. Modeling using nonlinear black box models (perceptron neural nets, radial-basis-function nets), training. Structures of dynamic nonlinear black box models, quality of neural net models. Model errors: Sources of errors, limits of model accuracy, model accuracy and controller performance.

The students can identify the dominant dynamics of a process from step responses and can apply modern methods and algorithms to identify the parameters of linear process models from measured data. They understand the concept of sufficient excitation and the sources of errors in parameter estimation. The students understand the structure of nonlinear black box models and can judge the quality and the limitations of data-based models.