High Dimensional Interval Fields: spatial uncertainty modeling for mechanical reliability (HiDIF)

Dealing with uncertainty is one of the key-competences of a mechanical engineer. When designing new machines, automobiles, airplanes, ... little is certain about the actual loading the design will encounter over its lifetime. Nevertheless, the designer has to come up with a deterministic design that, when produced according to his specifications, will put no (or at most very few) lives at stake. Whereas this was classically accounted for by applying safety factors, ensuring reliability is becoming far less obvious in a world where every component is engineered up to the limits of its performance. Therefore, uncertainty quantification is currently finding its way to the virtual computer aided engineering environment through the application of non-deterministic simulation techniques. One of the challenges we are facing here relates to uncertainty in properties that vary within a single component, as e.g., geometric variations due to tolerances, or variable material properties caused by the manufacturing process. These spatial uncertainties are commonly present, but hard to identify, and, because of the highly detailed models currently used, even harder to account for in numerical simulation using the already available techniques. Therefore, this project envisages the development of a new efficient and intuitive formalism for the representation and identification of ‘HD’ uncertainty in numerical models used for mechanical design assessment.