Structural Health Monitoring

Exploring how machine learning accelerates CFD simulations, from physics-informed neural networks to neural operators for fluid flow prediction.

How neural networks can model and predict complex dynamical systems in engineering applications, from Neural ODEs to operator learning.

This paper investigates the performance of Fourier Neural Operators (FNOs) for structural dynamics, highlighting their strengths in linear systems and limitations in non-linear systems, and introduces a spectrogram loss to mitigate issues.

A hybrid physics-informed machine learning (PIML) framework that adaptively optimizes finite-difference mesh nodalization to achieve real-time spacecraft thermal simulation with high accuracy and reduced computational cost.

A novel Neural ODE architecture that enforces physical consistency between displacement and velocity states, enabling interpretable latent dynamics and virtual sensing.

A comprehensive review of the PEML spectrum—ranging from white-box Bayesian filters to dark-gray physics-encoded learners—demonstrated on a Duffing oscillator.

A framework combining Variational Autoencoders with Physics-informed Neural ODEs and modal decomposition to reconstruct full-field structural dynamics from sparse sensor data.

A probabilistic framework combining deep learning with Extended Kalman Filtering to learn latent dynamics and noise covariances from structural response data without explicit physical equations.