FLOWer

In the context of urban climate and urban planning, computer-based models for wind and heat simulations are potentially valuable components that must be integrated into future planning processes for urban climate resilience. In a computer-based flow simulation, real airflows and wind pressure on elements such as façades can be accurately determined, taking into account local effects such as the atmospheric boundary layer or wind turbulence. The essential, time-consuming, and costly part of the simulation lies in solving the underlying physical equations, which is achieved through an approximating numerical method. This is where Deep Flow comes into play. For flows around individual buildings or ensembles of buildings, the highly specialized neural network is capable of replacing the numerical solution. As a result, the computational effort for these applications can be reduced from several hours on a typical high-performance computer to just a few seconds in a web-based app. Simultaneously, using the app does not require any prior knowledge in the field of flow simulations, making it accessible to an entirely new user group.