Surrogate Modelling for Structural Dynamics
A public overview of surrogate modelling for transient structural response under extreme fluid-driven loading.
PhD research — public overview · Current
My PhD investigates machine-learning surrogates for transient structural dynamics. The broad aim is to approximate expensive numerical simulation while retaining the behaviour needed for engineering use.
The application direction concerns structures driven by rapid fluid loading, including blast, detonation, and deflagration scenarios. This page intentionally stays at concept level while the research is unpublished.
Coupled high-fidelity simulation can be too costly for repeated design studies, uncertainty analysis, optimisation, or time-sensitive decisions. A useful surrogate would provide faster structural-response estimates while remaining clear about its limits.
The longer-term direction is a surrogate-enabled fluid–structure interaction workflow: finite-element and fluid-dynamics simulations provide the physical data, and learned models accelerate repeated response prediction.
- Abaqus for finite-element analysis and structural data generation
- Machine learning for faster structural-response prediction
- Future CFD data generation for the fluid side of the coupled system