AI, or rather Deep Learning, can be combined well with simulation.
Simulations can be very costly in terms of modeling and computing time in individual cases. If a product that is manufactured in a small number of units has computational costs that exceed the sales price, individual simulations do not make sense.
However, if simulation results of several, different individual cases are available, an AI can be trained to predict the computational results for new variants quickly and with minimum effort.
If, for example, one shows an AI the flow behavior of water in the case of a dam break, the AI can, after appropriate training, "calculate" the flow behavior of the flow in the case of a break of other dams in a fraction of the time required for a simulation, which is difficult to distinguish from the physical result of the real simulation.
However, this requires a great deal of training data.
Applications for this are still in their infancy.
The topic of batch size 1 is particularly exciting in medical technology and building services engineering.
Merkle & Partner is strongly interested and already sustainably active in research and development of the possibilities to develop simulation technologies here even further and to link them with meaningful applications from AI.
Join us in making the unthinkable possible!
Prostheses, implants, exoskeletons, the human-machine interface is always an individual case. But these individual cases have things in common.
If you combine the possibilities of simulation with empirically determined values from databases, you have everything you need for a successful AI project.
Let's consider a piece of clothing. The shape of the body can be scanned, trying on the clothes gives information about the comfort and the wearing comfort and the simulation summarizes the subjective feeling in key figures. If you start optimization runs (variants) in the simulation, you generate a lot of data that is of enormous value for training an AI. The effect: precisely fitting clothes.
Without simulation, one has only a subjective database, there is no explanation for the why, just as little about how optimization can be done. The potential is not exploited, new approaches and ideas for better comfort do not emerge.
In the long run, this purely empirical approach may work, as our evolutionary history of man shows. But it takes a long time, too long, we think. And simulation is the shortcut.
Buildings, apart from terraced houses and prefabricated buildings, are also individual cases.
For larger buildings, even this individual consideration is worthwhile, especially if 3D data is available via BIM.
Air conditioning can be optimized with the help of simulation. The optimization generates data. The indoor climate can be subjectively evaluated after the building is completed.
AI thus has an initial database that learns as the number of computationally optimized buildings increases and is eventually able to pre-assess a design.
The problem currently is that the driver is missing. The architect doesn't care, it has to look good. The heating and air conditioning contractor is paid by weight and the client, who would later benefit via comfort and energy costs, doesn't know.
The use of AI in building technology gives you a comprehensive advantage!
Merkle & Partner has been active in the field of engineering calculation with FEM and CFD simulations for over 30 years.
In hundreds of projects each year, we make invisible physical phenomena visible, help to better understand and optimize structures and products. And deliver meaningful approaches to solutions. With added value.
Merkle & Partner not only has a lot of experience. We work on the basis of the latest technologies, are involved in research projects and maintain close relationships with universities but also strategic partners from the industry. Take advantage of highly qualified know-how to advance your products and your company.