CSE Community Seminar | April 19, 2024

Abstract
The ability to efficiently predict unsteady flow fields is imperative for exploration, control, and performance optimization within the design space of a given problem. In many engineering problems, case studies rely on large amounts of numerical simulations or, in some cases, experimental runs, each one of which requires a spatio-temporal evolution of the flow field associated with a different set of design parameters. Data-driven Reduced-Order Models (ROMs) pose an attractive pathway to capture low-dimensional patterns and system dynamics based on a small set of high-fidelity training data. These models can then be used in a design or optimization pipeline for low-cost, high-throughput prediction of flow results. In this work, we developed a ROM to robustly predict the dynamics of fluid flows across a parametric design space. Our approach extends a Long-Short Term Memory (LSTM) neural network with a new design gate, which enables the network to distinguish dynamic patterns associated with different design parameters. The parametric LSTM (pLSTM) can successfully predict the long-time dynamics of the flow field for unseen design parameters while showing exceptional robustness to noise in the initial states. The proposed pLSTM offers a three-aspect ROM approach (space, time, design space) to benefit prediction, optimization, and control problems across parametric flow regimes.