Dr. Alice Cicirello (University of Oxford, UK)

Uncertainty Models in Structural Dynamics
When Feb 06, 2017
from 02:00 PM to 03:00 PM
Where LR8
Contact Name
Contact Phone 01865-283446
Add event to calendar vCal

Much industrial interest has been focused on rapidly explore the performance of a design to dynamic loading by building virtual prototypes. However, at the design stage there are uncertainties associated with (i) an inherent variability of the properties of the system because of the manufacturing process, and with (ii) a lack of knowledge of the analyst with respect to the system properties which are fixed. Therefore, the challenge is not only to develop a mathematical model able to capture the physics of the problem, but also to account for these uncertainties, which might be associated with the geometry and mechanical properties, loading, boundary conditions and structural joints.
The most direct approach to modelling uncertainties is to describe system parameters by means of a probability density function (pdf) and to propagate the uncertainties to yield the reliability of a system and/or the response moments. However, a limited set of data or vague information are often available, therefore specifying a single pdf for describing the uncertain variable becomes a very challenging task. This has led to the widespread of non-probabilistic uncertainty frameworks, such as intervals, convex and fuzzy descriptions, and several strategies have been developed to efficiently propagate these uncertainty models through the equations of motion to yield a bounded description of the system response. Other approaches introduce uncertainties in the probabilistic assignments by considering sets of probability distributions instead of a single distribution. However the application of these approaches to structural dynamics is very limited, mainly because of the computational burden associated to their propagation through the equations of motion. Alternatively, non-parametric models of uncertainty have been used to predict the response statistics of random structures as in Statistical Energy Analysis (SEA) method, Hybrid Finite Element/SEA method, and Random Matrix Theory.
In this talk, I will give an overview of these uncertainty models, with particular focus on recent advances on (i) the combination of parametric and non-parametric uncertainty models, and (ii) on imprecise probability.