Implementing Large Eddy Simulation to Numerical Simulation of Optical Wave Propagation
In this study, we want to simulate long-range laser propagation in atmospheric turbulence. The numerical simulations are carried out to study the impact of strong atmospheric turbulence in spatial, temporal, and related spectral domains. The first section of this study will be concerned with modeling this numerical simulation in Kolmogorov and non-Kolmogorov spectrum. To validate our numerical simulation, we will compare the statistical parameter to theoretical approximation in both Kolmogorov and non-Kolmogorov spectrums. Once the code is validated, we want to integrate Large Eddy Simulation (LES) turbulence modeling. LES simulations allowa us to study strong fluid turbulence and can predict advection-dissipation at various energy spectrums. Therefore, this choice of modeling allows us to characterize turbulence at outer and inner regimes important for correcting the effects of optical turbulence over long-distance laser beam propagation. LES simulations are computationally expensive and need high wall time and computational power. To overcome that, we implement a conditional generative adversarial network (cGAN) that applies phase unwrapping to a wrapped phase screen. The cGAN framework establishes competition between two distinct players in the model, the discriminator and the generator. The network is trained on 512x512 wrapped and unwrapped images and learns the mapping to conduct phase unwrapping. This portion of the study focuses on employing neural networks to improve the accuracy of current simulation. The focus of this study is to accurately simulate laser propagation through the atmosphere via our knowledge of fluid mechanics and machine learning.
Jimenez, Diego Alberto Lozano, "Implementing Large Eddy Simulation to Numerical Simulation of Optical Wave Propagation" (2018). ETD Collection for University of Texas, El Paso. AAI13424314.