Physically inspired models are a useful means for designing better machines and vehicles. They allow for a faster control design and verification of the correct operation of the machine/vehicle through virtual tests. But how can we incorporate the different behavior of multiple machines of the same type in a single model? Or how can we rely on a model, if we do not know to what extent its behavior deviates from reality? For example, what guarantee do we have that a model-based control design will produce the expected performance on the actual machine? Or in an end-of-line test, can we use a model to evaluate whether the observed machine behavior is within acceptable tolerances? So an understanding of the reliability of the model and its parameters is necessary.
In the Relimod project, an approach is developed and demonstrated that allows quantifying the reliability (uncertainty) of the output of physically inspired models and their parameters based on a certain set of experiments performed on a realization of the mechatronic system.
Date & Place: Thursday June 12 2014
@ FMTC, Celestijnenlaan 300, 3001 Leuven, Meeting room: Baekeland 00.207
Time: 9:00 a.m. – 13:00 p.m.
What to bring: a PC with Matlab (no particular toolbox required)
Target audience: engineers in modeling & control, reliability engineers, design engineers, test & validation engineers, technical directors, researchers in / interested to the field of modeling and reliability.
9 am: Welcome and introduction by Erik Hostens, projectleader
9.30 am Modeling uncertainty - basis concepts (Fisher information matrix; Jacobian propagation of input and parameter uncertainty) by Bert Lenaerts, senior project engineer
10.30 am Hands on Session by Erik Hostens and Bert Lenaerts
11.30 am Coffee Break
12.00 am Modelinguncertainty with the Parameter Estimation Toolbox
Participation is free of charge but registration is mandatory by sending an email to email@example.com, before June 6.