This cumulative dissertation consists of four variants of deep learning methods for physics field predictions using convolutional neural networks. First, a data-driven method to characterize film cooling flows in a rocket combustor, second, a combined data- and physics-driven method for heat conduction predictions, third, a physics-driven approach to solving fluid mechanics problems obeying Navier-Stokes equations, and finally, a generative adversarial network with physical evaluators framework to employ prior knowledge of the involved physics for spray simulations.
«
This cumulative dissertation consists of four variants of deep learning methods for physics field predictions using convolutional neural networks. First, a data-driven method to characterize film cooling flows in a rocket combustor, second, a combined data- and physics-driven method for heat conduction predictions, third, a physics-driven approach to solving fluid mechanics problems obeying Navier-Stokes equations, and finally, a generative adversarial network with physical evaluators framework...
»