Bots for Design: Dessia Uses Bots to Quickly Arrive at Optimum Design and Configuration

Is this an alternative to generative design?

A true solution to design optimization may be coming from a 5-year-old startup in France.

Dessia uses AI on the front end of product design—during the brainstorming process of evaluating the universe of possible solutions to arrive at a few viable ones, which leads to new products. And the company does this in the most extraordinary way: with a small army of bots.

You may remember the analogy for generative design, of a million monkeys banging away on keyboards with the hopes of one producing a Shakespearean verse. But Dessia is different. Instead of monkeys, it uses bots and another analogy: growing a tree, where its decisions are branches and pruning them is eliminating bad decisions.

Rather than provide a service or offer an app, Dessia sells a software
framework with which one can develop bots and execute them.
It’s a novel approach that once again fills me with hope. Could this be the white knight of design engineering, able to bring the true powers of computing to design and fulfilling the promise of CAD—the so-called computer-aided design that delivered only computer-aided geometry definition and not design at all. CAD can make concepts conceived in an engineer’s mind clearer, but it has not been able to come up with the concepts themselves.

But maybe Dessia bots can. Its bots are meant to support and augment the initial brainstorming session, the first part of the process that results in new products, where concepts are offered and weeded out. The free-thinking bots, having absorbed the design knowledge of a company to make a set of rules, blaze through one possibility after another for a new design, evaluating each design against criteria, over and over again.

How It Works

To use Dessia, one must encode the product specifications, domain expertise and either a functional product or a physical description. This is done with Python code using the Dessia SDK. Then the bots go to work, offering as many possible solutions—pruned by self-checking—as time will permit. The results are displayed as 3D CAD models. The CAD model is admittedly crude but can be imported into a more robust CAD program for refinement.

Unlike generative design applications, which make use of finite element analysis (FEA) to validate their shapes, Dessia uses simple, mathematical rules.

Using rules rather than performing a full analysis on every possibility is why we can run through more possible solutions, says Jean-Pierre Roux (JP), VP of sales and a mechanical engineer.

Coding a Prerequisite

To use Dessia, one must know how to use Python, the favored language of AI developers.

“It’s quite common nowadays for engineers to learn Python,” notes JP. “My daughter
heard of Python when she was 14.”

Typically, it is the IT team in a company that develops the bots using Python
and the end users run the bots and analyze data in Dessia’s cloud based no-code
environment, he explains.

“In our development roadmap we aim at bringing a low code interface for
bot development as well,” says CEO and founder Pierre-Emmanuel Dumouchel.

There’s no real competition for what Dessia is doing, JP says. “Our toughest job is in explaining what we do because it is different than how everyone else does it.”

Indeed, no one else employs an army of bots for its product design. Only Mendix comes close with its low-code app development, but Mendix, which was acquired by Siemens in 2018, will keep to the manufacturing side, away from the design side.

Dessia already has some big names as customers: Alstom, Airbus, Renault and Safran, among others. Those are just the companies that have not sworn Dessia to secrecy. The company has also been accepted into Aerospace Xelerated, the tech incubator headed by Boeing.

Dessia in Action

In addition to product shape design, Dessia can also be used for:

  • Systems engineering, with a block diagram of a system, and system examples, such as a power transmission
    and thermal management architecture.
  • Assembly and configurations, where Dessia’s bots will rearrange parts in various ways in an assembly; this brings to mind the most efficient nesting of parts on sheet metal or parts in a 3D printer vat. Dessia suggests that flight cockpit instruments are in an optimum position for pilot ergonomics.
  • Electric harness and wire and cable routing.
  • Fluid systems such as a cooling system or hydraulic brake systems.

 (Image courtesy of Dessia)

(Image courtesy of Dessia)

Renault Group has developed Dessia bots to create viable options for general
design architecture, cooling system channels, electrical system, which it can
evaluate for the hybrid powertrain in development. Dessia Bot was able to find
an optimum design and cooling system in one day when it would have taken its
engineers 30 days to come up with one solution of each—and not necessarily the
optimum one, according to Renault Group.

About Dessia

Dessia was founded by Pierre-Emmanuel Dumouchel (now CEO) and Steven
Masfaraud (now CTO). Dumouchel is a PhD mechanical engineer with 10
years experience at Peugeot-Citroën (now Stellantis) in powertrain department.
Masfaraud has a PhD in Mathematics and AI. The startup had a successful Series A
round of financing in 2021, raising $5.5 million, and was able to employ 30
people. Dumouchel has set the course on growth and customer acquisition in EMEA
and North America. Find out more at
www.dessia.tech
.