HomeNews and blogs hub

Aircraft Geometry and Surrogate Modelling

Bookmark this page Bookmarked

Aircraft Geometry and Surrogate Modelling

Posted on 6 December 2014

Estimated read time: 3 min
Sections in this article
Share on blog/article:
Twitter LinkedIn

Aircraft Geometry and Surrogate Modelling

Improving release management and community engagement

Aircraft in flight

Modern aerospace engineering design synthesis relies upon two key ingredients: a flexible, parametric description of the object being designed e.g., the external surface geometry of an aircraft, and a measure of merit capable of evaluating the performance of each candidate geometry. A performance optimization process then searches the broad range of geometry parameter combinations and, upon computing the measure-of-merit corresponding to each, it converges towards the best design. Over the last decade, András Sóbester and Alexander Forrester, within the Faculty of Engineering and the Environment at the University of Southampton, have developed and honed a range of software tools to assist with this design process.

Combating the curse of dimensionality

Searching large parameter spaces can easily become prohibitive if the parameterisation of the geometry model or the evaluation of measure of merit is inefficient. Combating this "curse of dimensionality" is essential to make this process affordable.

The tools developed by András and Alexander support geometry models that are flexible, yet concise and robust, and statistical models that are capable of learning the response of expensive measure-of-merit calculations. Both the geometry and statistical models are code-based. Geometry models are code whose inputs are design variables and output is geometry models in various formats. Statistical models are general-purpose low computational-cost surrogates of high-fidelity physics-based analyses. Each of the tools complement text books written by András and Alexander.

The software tools include:

Each of these has been released under the GNU General Public License and GNU Lesser General Public License.

While most of the applications of these supervised learning codes are related to aircraft design, they have been applied in diverse fields including razor design, swimming analysis and hydro-turbine engineering.

Objectives

For the past few years, András and Alexander have polished the numerical performance and effectiveness of their codes. However, they now feel that further development is hindered by limitations in how they manage releases and manage engagement with, and contributions from, their user community.

We will provide recommendations both on how software updates and releases can be managed and community engagement improved. This will be based upon a review of how András and Alexander manage their software development and community engagement at present.

 

 

Share on blog/article:
Twitter LinkedIn