Poster Presenters
- Carter Montag | UA PhD Applied Mathematics Graduate Student
- Title: Machine-Learned Inverse Pulse Solving for NIF Pulse Shape Prediction
Abstract: At the NIF, the master oscillator room (MOR) input required to produce a desired target pulse shape is typically found with an iterative inverse solver. Because this process can take minutes to converge, it can impact the shot cycle and limits offline pulse-shape design studies. This challenge becomes even more significant in the design and operation of future inertial fusion energy systems, where shot rates near 10 Hz would make minutes-scale iterative pulse design impractical. A fast, data-driven surrogate that maps desired laser entrance hole (LEH) waveforms to MOR input pulses was developed, reducing solve time by supplying either a high-quality initial guess for the inverse solver or a candidate waveform for forward-simulation validation.
- Ed McDugald | UA PhD Applied Mathematics Graduate Student
- Title: Learning the Equations of Stripe Patterns as Defects Form
- Abstract: Stripe patterns appear across physics and engineering, from convection rolls and fluid instabilities to crystal growth and materials microstructure. Their large-scale behavior is often described by reduced models for the local wavevector, but these models are not expected to hold where the pattern’s topology changes, for example when a smooth bend in the stripes forms dislocations. We present a pipeline that extracts order parameters (wavevector, amplitude, and phase) directly from simulations of the Swift–Hohenberg equation, validates them against known analytic solutions, and then uses weak-form sparse regression (weak SINDy) to identify the equations these fields obey. Every identified law is read beside synthetic controls with known answers, so that features of the physics can be separated from artifacts of the extraction. The data recover the classical phase theory where the pattern is smooth and stationary, and show where and how the Swift–Hohenberg fields depart from it as defects form. We discuss what this suggests about which reduced descriptions can capture topological change, and how equations learned from data can be held to the standard of physical theory.