Inverting shock-wave temperatures via artificial neural networks

Abstract

Temperature is one of the most important parameters for characterizing the thermodynamic state of matter in extreme conditions. However, there is as of yet no universal and accurate way to measure the temperature associated with a shock wave propagating in an opaque material, let alone an inversion method for determining how this temperature evolves. Based on the current strong generalization and learning abilities of artificial neural networks, this paper proposes using an artificial neural network to determine (i) how the shock-wave temperature in a material evolves and (ii) the surface temperature of the interface between the material and vacuum when a shock wave propagates through the material. Data generated using a one-dimensional numerical hydrodynamic simulation are used to train the artificial neural network by applying backpropagation and optimization to many datasets. Once the artificial neural network is trained sufficiently, it becomes an excellent approximator that can estimate the shock-wave temperature from a given streaked-optical-pyrometer image and other known information from the experiment. The paper ends with various possible extensions to the present research.

Physics-informed neural networks for inverse problems in supersonic flows

“…Apart from PINNs, the other frameworks which employed deep neural networks for tackling shock wave problems can be found in [24,25,26,27,28]. Inverse problems in supersonic compressible flows are often encountered in designing specialized high-speed vehicles.…”