Cloud computing and storage make AI applications like these possible. “A great advantage of Industry 4.0 and cloud computing is that it allows all sizes of factories to deploy very quickly without any installation or production downtime. They can start tracking their assets, parts and materials, collecting a lot of information without having to worry about where to store it,” he adds. “On top of that, you can get all of the computing power for those AI algorithms, whether we are talking about machine learning or other tools that would require significant computing power.”

The beauty of cloud computing is that it gives companies the flexibility and the ability to use these resources quite inexpensively and only where they need it. Additionally, if they wish to scale up production, it’s easier to do once they are on the cloud.

Algorithms for Optimization

With the availability of increased computing power, some companies are employing mathematics to help composites manufacturers overcome production challenges.

“Our job is translating customers’ challenges into problem statements which we can solve mathematically,” explains André Wilmes, CEO of Rafinex. “We target our high-end algorithms directly to their specific pain points, giving users answers upon which they can act, rather than overloading them with raw simulation results.” Challenges can range from reducing the time to produce a part to improving the precision of a tool.

Rafinex’s services include stochastic topology optimization (shape generation), random-variable risk modeling, composite fiber direction optimization and early-design manufacturability analysis. Stochastic topology optimization enables manufacturers to produce safer and more robust designs for composites used in high-performance sectors like automotive, aerospace and some sports. The stochastic process accounts for the real-life variability in load directions and in material properties for a design. It then tests them in thousands of simulations, simultaneously and near instantaneously. The result is a single design that is optimized for the customer’s requirements, while also remaining safe across best- and worst-case scenarios, thanks to risk quantification.

Rafinex’s stochastic risk analysis algorithm uses the same mathematical approach to provide insights, in one single simulation step, into the global performance of a concept-stage design across all real-life operating conditions.

The company selectively uses AI to speed up these simulations and optimizations. “As the simulations progress, the intermediary data can be used to accelerate the algorithms so that they finish in a couple of hours without losing accuracy and thus design quality,” says Wilmes. “That allows us to scan a much wider design space in our optimization.” Previously, it used to take up to three or four weeks to run a stochastic topology analysis on any realistic industrial design problem.