Neural Network Determination of Optical Phase Correction in a Plane Shear Layer Using Parallel Optoelectronic Image Processing and Global Optical Flow Diagnostics
Abstract
Neural networks that allow aero-optic phase correction to be made using localized measurements without an external probe beam are being developed in an information-rich laboratory environment. The neural networks are trained to relate phase corrections to low-order modal descriptions of a plane shear layer obtained by a proper orthogonal decomposition (POD) applied to index of refraction data. Optical measurements are taken in a plane shear layer between two uniform streams with different temperatures. The training sequence will use actuators to influence the flow, thus providing the networks with a broader operational range. Elements critical to the training include development of a three-dimensionally interconnected hig frame rate optoelectronic smart camera, extraction of the velocity field using two scalars, and determination of modal coefficients in a low-order description of the flow. The ultimate objective is to simultaneously determine teh three-dimensional index of refraction field and the resulting optical phase front distortion in the plane shear layer, thereby providing real-time correlation between index variation and optical phase shift to the neural networks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1998
- Accession Number
- ADA351961
Entities
People
- Ari Glezer
- Martin Brooke
- Nan Marie Jokerst
Organizations
- Georgia Tech