CFD Schuck Ingenieurgesellschaft won the Digitaler Mittelstands-Award (DIMA) in the category “Innovation” for their use of artificial neural networks (ANNs) to forecast product characteristics and reduce costs. With the help of ANNs, tendencies for various target dimensions can be forecasted, such as the calculation times of CFD simulations or the performance of a cooling element.
In an interview with Managing Director Andreas Schuck, he reveals how the idea came to life, which industries can profit from ANNs and what he hopes to gain by winning the award.
Mr Andreas Schuck, what triggered you as an engineer to venture into the topic of artificial neural networks?
As a service provider, we have been developing virtual prototypes for 30 years now, with a focus on fluid mechanics. Within this process, many cost- and time-intensive calculations were undertaken to date. The first idea regarding AI-supported analyses of existing data, with the aim to make product development more efficient, came about in 2018. The realisation of this then followed within the scope of a project from the German Central Innovation Programme for small and medium-sized enterprises (ZIM). It has existed for quite some time now as open-source software, which provides the base for such artificial neural networks. The challenge was to formulate this code in such a way that our idea could be realised, namely the reliable forecasting of product characteristics. Now we can concentrate in advance on just a few promising models out of a large number of design variants.
Which industries can profit from this innovation?
This technology is useful across all industries. For many years, we focused on the automotive industry, as well as on aviation and aerospace technology, but the opportunities for use can be found throughout the entire product development sector where technical parameters are combinable with target dimensions: from the development of cooling elements to filling systems to electric mobility.
Based on a concrete, real-life example: how do ANNs help to develop products?
Let’s take the example of a cooling element to showcase the application. Reliable forecasts about the performance of so-called cooling ribs could be made. Initially, known parameters were imported via an intuitive Web-based user interface, such as length and number of ribs, their position and distance to one another. Afterwards, the learnt ANNs calculated the respective tendencies about the predefined target dimensions, in this case the cooling performance, and delivered the respective output. The results and the parameters applied were then provided in a single file. On top of this, new calculations could be performed until the optimal variant was determined. New fluid simulations are now no longer necessary.
What do you hope to gain now by winning the Digitaler Mittelstands-Awards?
First and foremost, I expect to see a real boost in motivation among my employees. With this award, they see that joint project development can make a big difference. What’s more, we are still in the marketing phase. Winning the DIMA certainly helps then to increase awareness, so that our customers or potential new customers can perceive us in a different way and so that we can develop such projects together in future.