Latest news

I am an ISAE-ENSICA[2] graduate, and I’ve always steered my way between industry and research. Digital optimization has been the common theme from the time I did an internship at Altran, then via the subject of my Cifre[3] PhD thesis with Airbus and Cerfacs[4] and until the present day. The idea is to design algorithms and software that improve aircraft performance. I had a taste of more basic research at Edinburgh University in civil engineering. I created my own company for a college project with the development of a software package, production management and customer relations, when the importance of the user became clear to me. The end of my PhD thesis coincided with the setting up of IRT Saint Exupéry’s MDA-MDO project, which I then joined.

The aim of the MDA-MDO project is to demonstrate that multidisciplinary optimization can be applied effectively to industry. It started with an Airbus test case on an engine pylon, a huge part made of titanium that connects the engine to the wing, and it’s designed and manufactured in Toulouse. The goal was to optimize the aerodynamic external fairing simultaneously with the internal structure as a way of minimizing fuel consumption.

 

We developed a software platform called SPIRO[5], whose mathematical core is the GEMS software, which can coordinate a complex process using optimization and coupling algorithms. This process is based on several different tools for simulating structures and aerodynamics developed separately. The ultimate goal is to revolutionize industrial design by combining computer science, math and business lines. Even though we’ve known since the 1990s that MDO can be applied to aeronautics, these algorithms are rarely used industrially because they do not factor in many constraints.

 

My role is to combine the needs of industrial design and applied mathematics. At this interface, I developed and designed the GEMS software, which is now used more widely at IRT Saint Exupéry, outside and beyond our borders. In addition, the strategy followed for its open source implementation allowed us to be associated with the European project Madeleine. I’ve just got back from the AIAA Aviation Forum[6], where we saw that GEMS was acknowledged as a major contributor to MDO methods applied to industry.

The SPIRO platform lets you run distributed calculations on thousands of processors with infrastructure stability issues managed for weeks by automatic restart. Our solution has been validated by the experts at Airbus, with considerable savings in computing time thanks to new algorithms we’ve developed. This means that the integrated IRT team can quickly make the connection between mathematical techniques developed by researchers and real industrial cases. That’s done through PhD theses, post-docs and staff seconded from Cerfacs and Onera, a pioneering MDO laboratory.

 

We have written a dozen publications to consolidate our academic recognition and offer the visibility necessary to initiate new collaborations.

 

Our strategy is to distribute GEMS in open source so we can quickly advance and tap into the best algorithms from the world of research, which is keen on open source, a movement that has gained pace with the development of the cloud and AI. GEMS makes it possible to bridge the gap between industry and academia. For the former, it offers an open and reproducible interface for accessing algorithms to solve their industrial problems. For the latter, it means they can work with their algorithms while accessing representative problems. As for IRT Saint Exupéry, it is positioning itself to provide expertise and transfer in both directions.

What I really like about IRT is the meeting between industry and research, which tallies with my own career This internal knowledge of industrial needs and associated constraints is a strength to work together within the IRT while preserving everyone’s interests. At IRT, we break down lots of barriers.

 

There is an enormous wealth of sharing with experts from other disciplines. I provide my support to various projects in other IRT Saint Exupéry Domains, such as More Electrical Aircraft and Materials, which use GEMS. I assist other teams that use optimization and I co-supervise a PhD student with Airbus and ISAE-SUPAERO. I like the diversity of the practical problems that need to be solved.

I like hiking, mountaineering and climbing, and it turns out that the image of mountains is widely used to help understand the mathematical theory of non-linear optimization. This tries to minimize a function (the optimizer) whose properties depend on directions in space. Imagine that the algorithm is in the mountain and aimed at the lowest point in a valley. Without a map, it can only take measurements. Each measurement takes many hours of calculation, and it attempts to descend as fast as possible to the bottom of the valley. This valley has complicated aerodynamics because it is very flat in one direction and has extremely steep slopes in another, and all in 400 dimensions, i.e. 400 unknowns for the optimizer. One of our techniques can be used to work in a so-called “preconditioned” space, which transforms this flat valley into a half-sphere, which facilitates the search for the lowest point enormously. The cost of solving our engine pylon problem has been reduced by a factor of four!

[1] Generic Engine for MDO Scenarios.

[2] Ecole Nationale Supérieure d’Ingénieurs de Constructions Aéronautiques.

[3] Cifre is an industrial research training agreement, i.e. a scheme for funding theses that helps firms recruit young PhD researchers.

[4] Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique (Airbus/Cnes/EDF/Météo-France/Onera/Safran/Total/CNRS).

[5] Software Platform for Industrial and Research Optimization

[6]  Aviation Forum organized in June (Dallas, USA) by AIAA, the American Institute of Aeronautics and Astronautics, an association of industrialists and research labs from around the world.