Real-Time Structural Health Monitoring Using Physics-Driven Digital Twins
summary
Claims a working hybrid digital twin that fuses physicsbased models with deep learning to do real‑time damage identification and short‑term prediction, uses physics‑informed ML to cut label needs, and uses generative AI (GANs, VAEs) to synthesize failure data for data‑scarce rare events; demonstrates anomaly detection with isolation forest and diffusion‑based probabilistic detectors (dynamic diffusion models), runs prototype Python/MATLAB examples on sine waves (2,000 training pushes, observed loss ~0.7) and points to ANSYS Twin Builder for deployed twins and a self‑healing material path.
AI-written summary · about
chapters (10)
- 0:00Physics-driven digital twins intro
- 1:53Digital twin foundations
- 10:12Real-time digital twins
- 15:11Live anomaly detection demo
- 21:19Physics-informed ML explained
- 29:55Generative AI for failures
- 49:17Diffusion models for anomalies
- 57:09Digital twin tools
- 1:04:21Self-updating digital twins
- 1:12:48Q&A
related talks
36mPhysics-Informed Spatiotemporal Deep Learning for Emulating Coupled Dynamical Systems
24mDigital Battery Twin
34mData-Driven Solutions in Engineering
10mAI-Powered Digital Twins: Patterns for Smart Manufacturing
02mxDT: Virtual Sensing on Industrial Edge Powered by
27mGeneral Keynote Multi-tiered digital twins with what-if analysis...
30mDataOps for digital twins
59mInnovating power plant operations through digital twins