Parametric uncertainties in dynamic models for aeronautical structures

Different EADS divisions are in the process of improving their capabilities in terms of numerical prediction of vibration levels for large structures. It still is a challenging task even in the relatively low frequency range of 0 to 50 Hz, because of the large size and hence very high modal density (meaning many modes in 0-50Hz range) of the structures at stake. If very simplified models are validated in the 0-15 Hz range, the next step involves the description of many more structural parts, especially structural details (such as secondary structures that are relevant dynamically and the physical size and properties of junctions) and also a lot of relevant mass items (such as passengers and cargo and fuel and systems heavy enough to have an impact on the overall dynamics). In addition to deterministic modelling difficulties (meaning modelling things you know), there is at any development stage a lot of uncertainty in the knowledge of many parameters, and it would be very valuable to account for it for two reasons: first of all, this variation in dynamic properties is a reality and a reliable prediction should include this lack of knowledge. furthermore, it is not always meaningful to give as a prediction result deterministic dynamic responses, but often more interesting to produce response distributions with uncertainty bounds. In what follows, the use of an open-source platform OpenTURNS in connection with finite element solvers will be illustrated on several examples. The examples are taken from current studies on space and aircraft systems. The employed methods belong to a class of parametric uncertainty management methods, and are state of the art, but the aim of this paper is to assess their relevance when used in relation to realistic and full scale dynamic models, where computation time is an issue.