DEPS Abstract

Guest >> Sign In

DIRECTED ENERGY PROFESSIONAL SOCIETY

Abstract: 24-Symp-116

UNCLASSIFIED, PUBLIC RELEASE

Modeling and Feed-Forward Predictive Neural Network Control of the Wavefront of High-Energy Pulsed Lasers

Pulsed lasers produce a broad range of superior DE lethality effects, however, they also involve more difficult control of the output wavefront. Specifically, it is well known that temperature variations in the laser gain material produce optical aberrations in solid-state lasers. These aberrations arise from three sources. First, aberrations arise due to the dependence of the index of refraction on temperature within solid-state materials. Stresses produced by temperature variations in the gain material also produce index-of-refraction variations due to photoelastic effects. Finally, temperature-induced strain can result in deformation of the faces of the laser material. For CW lasers the laser quickly comes to thermal equilibrium so that the thermal aberration is constant. However, for pulsed lasers, energy is continually added during pumping pulses. Also, cooling of the gain material occurs between pulses and after pulse bursts, so that the exact temperature distribution in the rod and hence thermal aberration, can be highly variable and unpredictable.
In this presentation, we will describe the development of a mathematical model for the thermal-induced optical aberration in a cylindrical laser rod. The model is based on a finite-difference solution of heat conduction in the laser rod, and includes calculation of the absorption and diffusion of pump radiation, energy extraction by the laser emission, and the effects of index of refraction dependence on temperature, photoelastic effects, and deformation of the rod ends on the wavefront of the emitted laser beam. Validation of the model against experimental data will also be presented. Following this, methods will be described for using the mathematical model to generate a database for training a neural network that can rapidly predict the thermal-induced laser wavefront for a variety of operating conditions including pulse rate and energy, cooling flow, etc. The design and training of a neural network will also be described for handling intermittent and random on/off cycles of the laser, as would occur as new targets are acquired. The resulting neural networks could be used to rapidly provide mirror shapes to a deformable mirror that would be used to correct the outgoing laser wavefront.

UNCLASSIFIED, PUBLIC RELEASE