UNCLASSIFIED, PUBLIC RELEASE Modeling and Feed-Forward Predictive Neural Network Control of the Wavefront of High-Energy Pulsed Lasers 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
|