Engineering topics
Publications and source records attributed to Kur, Eugene.
DeepOpt is a simple and easy-to-use library for performing Bayesian optimization, leveraging the powerful capabilities of BoTorch. Its key feature is the ability to use neural networks as surrogate functions during the optimization process, allowing Bayesian optimization to work smoothly even on large datasets and in many dimensions. DeepOpt also provides simplified wrappers for BoTorch fitting and optimization routines.
Deep neural network (DNN) surrogates of expensive physics simulations are enabling a rapid change in the way that common experimental design and analysis tasks are approached. Surrogate models allow simulations to be performed in parallel and separately from downstream tasks, thereby enabling analyses that would be impossible with the simulation in-the-loop; surrogates based on DNNs can effectively emulate diverse non-scalar data of the types collected in fusion and laboratory-astrophysics experiments. The challenge is in training the surrogate model, for which large ensembles of physics simulations must be run, preferably without wasting computational effort on uninteresting simulations. Here, in this paper, we present an iterative sampling scheme that can preferentially propose simulations in interesting regions of parameter space without neglecting unexplored regions, allowing high-quality and wide-ranging surrogate models to be trained using 2–3 times fewer simulations compare to space-filling designs. Our approach uses an explicit importance function defined on the simulation output space, balanced against a measure of simulation density which serves as a proxy for surrogate accuracy. It is easy to implement and can be tuned to find interesting simulations early in the study, allowing surrogates to be trained quickly and refined as new simulations become available; this represents an important step towards the routine generation of deep surrogate models quickly enough to be truly relevant to experimental work.
Herein we report on the experimental measurement of specular reflection (“glint”) of laser beams off the hohlraum wall in inertial confinement fusion experiments at the National Ignition Facility. In a hohlraum, glinted light can escape the opposite laser entrance hole of the hohlraum and is a potential laser energy loss mechanism. The total measured glint on the inner cones of beams is measured to be less than 8 TW (when using the full National Ignition Facility laser), which is <2% of incident peak power. The simulated x-ray flux exceeds the measurement by 10%–20%, and glinted laser light is unable to account for this discrepancy. Similar inner beam glint was measured for ρ = 0.3 and 0.6 mg/cc gas fill hohlraums, but no glint was detected for 1.2 mg/cc densities. Inner beam glint is dominated by the lowest angle 21.5 beams within a 23.5 quad, and it is at most 30% sensitive to different quad polarization arrangements.
A novel structure-preserving algorithm for general relativity in vacuum is derived from a lattice gauge theoretic discretization of the tetradic Palatini action. Here, the resulting model of discrete gravity is demonstrated to preserve local Lorentz invariance and symplectic structure.
Beam spray measurements suggest thresholds that are a factor of ≈2 to 15× less than expected based on the filamentation figure of merit often quoted in the literature. In this moderate-intensity regime, the relevant mechanism is forward stimulated Brillouin scattering (FSBS). Both weak ion acoustic wave damping and thermal enhancement of ion acoustic waves contribute to the low thresholds. FSBS imparts a redshift to the transmitted beam. Here, regarding the specific possibility of beam spray occurring outside the laser entrance holes of an indirectly driven hohlraum, this shift may be the most concerning feature owing to the high sensitivity of crossed-beam energy transfer to the interacting beam wavelengths in the subsequent overlap region.