What is the best architecture for the gearbox of a hybrid car? If an engineer had to answer that question, he would consider a handful of possibilities based on what already exists on the market. But the startup DessIA takes a whole different approach. Its artificial intelligence algorithms enable it to consider billions of different architectures to find the optimum configuration. The software developed by the young company digitally builds all the possible structures using the necessary components. The performance and the feasibility of the architectures built using this method are assessed, the design space is therefore intelligently explored to reduce the number of architectures physically tested. The automated, smart sorting keeps only the best architectures. In addition to the possibility of analyzing considerably more models than a human could, DessIA’s advantage is that the layouts created with its components are radically different from what already exists. “When we present our approaches to manufacturers, many of them say this is exactly the way they want to work, but they have no idea where to start,” say Pierre-Emmanuel Dumouchel and Steven Masfaraud, co-founders of the startup incubated at ParisTech Entrepreneurs.
For now, DessIA is specialized in subjects related to the transmission of mechanical power. It can work on both on gearboxes for cars and systems for transferring energy between a helicopter’s turbines and blades. The field itself is vast, and reflects the experience of its two founders, former employees of PSA. The issues can even include the mechatronic systems of complex electrically motorized mechanisms. The startup’s applications are limited to this subject because the algorithms’ work must be controlled by a thorough knowledge of the sector. Still, the two founders are not ruling out the possibility of someday moving towards providing assistance in the design of electrical or hydraulic systems. But not until a few years from now.
By remaining focused on mechanical systems, many opportunities have opened up for the young company. DessIA’s objective is to go beyond the mere optimization of architectures. Once the best structure has been determined, the ideal solution would be to have a very simple way of obtaining a 2D industrial plan, or even the 3D CAD model to directly integrate into the computer aided design software. The two founders intend to achieve this outcome by the end of 2018. If they succeed, they could redefine how mechanical systems are designed at the industrial level, from the reflection phases to drawing the part.
Pierre-Emmanuel Dumouchel worked at PSA for 10 years. After supervising Steven Masfaraud’s thesis for three years, they decided to partner together to create DessIA. They aim to simplify the design process for engineers through a breakthrough approach based on artificial intelligence.