ICON is a global weather and climate system that allows seamless predictions from a local to a global scale. With the addition of the ART submodule, that is developed by KIT, ICON is further supplemented by including emissions, transport, gas phase chemistry and aerosol dynamics in the troposphere and stratosphere. A drawback of ICON-ART is that the simulation of aerosols and chemical subprocesses is at the moment computationally expensive. A way to reduce complexity in ICON-ART is with the use of deep neural networks. These networks would act as approximators of the underlying physics and as a result heavily accelerate computation.