AMEP Workspace

Reduced-Order Modeling with Deep Learning: Flow Reconstruction for External Aerodynamics and Thermal Convection

Inspiration

Objective

Integrate high‑fidelity CFD with data‑driven machine learning (POD → DNN) to create real‑time surrogate models that replicate complex flow physics with <1 % mean‑squared error, reducing simulation runtimes from hours to milliseconds and enabling rapid aerodynamic analysis, design optimization, and control‑oriented predictions on standard hardware.

Results

The Double Pendulum Fractal

Inspiration

Objective

Apply my knowledge of Lagrangian and Hamiltonian formalisms (Physics 311) as well as applied dynamical systems (Math 415/519) to predict when the second pendulum will 'flip'. Unlike the simple pendulum one might analyze in an introductory physics class, the double pendulum is chaotic; hence, I have employed numerical methods to solve the equations of motion which is then validated by my analytical derivations. The final result is a beautiful fractal.

Results