Skip to content

Disentangled Representation Learning for Damage Identification

Home / Concepts / Disentangled Representation Learning for Damage Identification

Disentangled Representation Learning for Damage Identification

Auto-generated stub. Edit this file to add more details.

A self-supervised framework using dual-latent autoencoders and invariance regularization to separate damage-related features from operational/environmental noise.

Why It Matters

It introduces a method to separate physical damage signals from operational noise using VICReg and spectral constraints without requiring labeled data, which is highly reusable in industrial anomaly detection.

Evidence

a self-supervised label-free disentangled representation learning framework for robust vibration-based structural damage identification

Metadata & Links