Efficient numerical methodologies for uncertainty quantification and optimisation

After four years of research in robust design methodologies for uncertainty quantification and optimisation for industrial applications the EUFORIA consortium is ready to present its results and developed methods. On 25 April Brussels is the place to be for an inspiring conference.

Under the name EUFORIA ('Efficient Uncertainty quantification For Optimisation and Robust design in Industrial Applications') a consortium of academic experts from VUB, KU Leuven, UGent, VKI, Stanford University and Sirris, was brought together to tackle scientific challenges and to realise industrial innovation. The aim of the project has been to develop efficient methodologies for the optimisation process 'under uncertainties'.

Optimisation process 'under uncertainties'

Today it is possible to perform optimisations on all kind of structures through Computer-Aided Engineering (CAE). Computer-Aided Engineering has grown strongly over the last few decades, and in the future it will put its mark even stronger on the process of product design. However, during the CAE design process only an average value of many input parameters is used, while in real conditions the actual input parameters will differ from this average value. Also the models upon which the optimisation process is based are in most cases a 'best guess' method and geometrical tolerances can influence the result significantly. Understanding these uncertainties will enhance the reliability of the models and make them a powerful tool within the optimisation process.

For uncertainty management and optimisation of engineering products, it is necessary to include all sources of uncertainty in the simulations. This in turn can dramatically increase the computational cost and memory requirement ('the curse of dimensionality'), i.e. with increasing number of uncertain parameters the number of simulation calls increases exponentially. From an industrial point of view, developing efficient reduced order models to reduce the computational cost is therefore of great interest.

EUFORIA methodology

The EUFORIA methodology includes a number of mathematical approaches to propagate uncertainties - both on input and design - and to converge as efficient as possible to a global optimum with a minimised number of samplings. This will result in a Pareto front of global optimal designs that serves a certain product strategy. To achieve this goal, a combination of different mathematical algorithms have been studied. These are called 'building blocks'. Each building block is a piece of software code - a script - with a specific function, that wraps around an existing code and improves its efficiency and cost. The developed methodologies enable to achieve a reduction of the computational cost by a factor 4 and optimised designs that are as insensitive as possible to parameter fluctuations, to uncertainties, to manufacturing tolerances, etc.

Inspiring conference

On 25 April a conference will take place at Bluepoint Brussels, at which the results of this 4-year research project will be presented. The aim is to share upcoming methods and techniques both on academic and industrial case studies, together with partners Numeca, Noesis, Bekaert, Siemens, Bosal, 3E and Keppel Seghers.

Would you like to know more? You can find all further information in our agenda!  

This research project is made possible thanks to the support of VLAIO.

Tags: