Renumics automates CAE processes with the aid of artificial intelligence – the innovative software platform makes complex simulation technology accessible and easy to use
Computer-aided engineering (CAE) with the aid of simulation is extremely important in lightweight technology for optimising structures on the PC. The problem is that it is usually extremely expensive, time consuming and personnel intensive. So why not just automate simple numerical simulations with the aid of machine-learning methods and artificial intelligence and then let a clever algorithm do the work? That is precisely what the three founders of the Karlsruhe-based company Renumics GmbH thought before developing a software platform which not only makes it a easier but also considerably less expensive to use the simulation technology.
The Development Agency for Lightweighting Baden-Württemberg presents this innovation in the December edition of its ThinKing. Leichtbau BW GmbH uses this label to showcase excellent lightweight technology products or services from Baden-Württemberg every month.
Crash tests play an indispensable role in the development of a new motor vehicle. However, there are many simulations to perform on a computer before a prototype is subject to a crash test and impacts a solid wall in a test laboratory. This requires that the CAD model is transferred to a simulation model, which can very quickly make use of several hundred hours of work, because the procedure is extremely complex: there are numerous manual steps and individual “clicks” that have to be made on the PC to assign the specific physical properties of the components and materials to the model. A huge amount of know how is also required and a complex simulation consequently needs to be performed by numerical analysts.
“The large-scale input of human resources and financial costs involved meant that in many cases simulations were only performed in large corporations. SMEs tended to do without them”, explains Dr. Stefan Suwelack, one of Renumics’ three founders. The software platform created by the Karlsruhe-based start-up is specifically intended to enable small and medium-sized enterprises to make use of simulation technology. “Our form of machine learning involves people teaching the computer how to understand geometries, for example, or how particular workflows must be performed so that previously manual processes can later be automated,” says Suwelack.
The computer takes on the processing of geometry data
This not only takes some of the strain of numerical analysts. The really big advantage which Renumics’ software solutions offers is that automation enables significant time and cost savings for the development of a simulation. “We are also making these processes and the technology more accessible for SMEs,” continues Suwelack. The software can be used by non-experts who do not have a specialist background in simulation. “Our philosophy for the Renumics platform is to provide a kind of modular system. The software currently in use is usually highly complex. We prefer to use far more streamlined simulation apps which are specialised for a specific application field.”
Digitally mapping production processes
It is possible, for example, to teach the algorithm to interpret measurements and to decide whether they are “good” or “bad”. Dr. Stefan Suwelack believes that the software also has great potential for design to cost and the digital mapping of entire production processes. Simulations can be used on the PC to determine and monitor quickly and simply how manufacturing costs, for example, change when a particular material or component is swapped for another one. It could also conceivably be used for topology optimisation in mechanical engineering. Lighter components not only allow manufacturing costs to be reduced, the drives of moving parts could also be made smaller if less mass needed to be moved. This would reduce both the amount of energy consumed and the amount of CO2 emitted.
Optimisation at the touch of a button
“I also have additive manufacturing in mind: for example, our software could help to calculate automatically whether a model is suitable at all for additive manufacturing or whether there is perhaps a better way of processing the material – fully automatically and, as it were, at the touch of a button,” explains the Renumics’s founder. Optimisation through simulation is now indispensable in lightweight engineering. “As we often come up against the limits of both the material and the structure in lightweight engineering, it is essential to keep on testing and trying out whether the parts really hold so that the results can be fed back into the original design process. This brings about incremental improvements in design structures. Digitalisation enables this process and the requisite feedback loop to be sped up enormously,” explains Dr. Wolfgang Seeliger, Managing Director of Leichtbau BW GmbH.
Suwelack believes it would also be perfectly conceivable for the software to be used to simulate the entire product lifecycle, such as the maintenance of components. The measurement values may be compared with the “digital twin” and the values from the virtual sensor to identify any hidden damage, such as cracks or vibrations. “The big difference is that the algorithm later works in precisely the way it has been trained. Users all train the algorithm with their own data so that user A ends up using a different version than user “B”, says Suwelack. Several pilot customers, including from the automotive and plastics industries, are already using the Renumics software. “The subsequent processing of the simulation results of vibration analyses has been automated for the first time, for example,” explains Suwelack.
Renumics GmbH was founded in February 2017. The name is an acronym of “rethink numerics”. The three founders have known each other from the day when they all worked on medical technology research at the Karlsruhe Institute of Technology (KIT). Today the team draws on experience and research in the field of numerical simulation and cognitive computing and has specialised in software development and CAE.