- Date: Tue. 14.10.2025, 4 pm
- Room: MW 1501 (5505.01.501)
- Presenter: Johannes D. Schmid
- Title: Physics-Informed Neural Operators for the Prediction of Structural Intensity from Laser Doppler Vibrometry Measurements of Plates
Abstract:
Structural intensity provides valuable insight into how vibrational energy propagates through plate structures, but estimating it from experimental data is notoriously difficult due to the noise sensitivity of numerical differentiation. This work introduces a physics-informed deep operator network that predicts structural intensity directly from noisy laser Doppler vibrometry measurements. By learning a smooth, differentiable representation of the displacement field, the method enables stable computation of higher-order derivatives via automatic differentiation, avoiding the limitations of conventional approaches. The framework is validated on both an analytical benchmark and experimental plate measurements, showing high accuracy in capturing magnitude and directional energy flow across a wide frequency range. The results highlight the potential of physics-informed machine learning for robust structural dynamics analysis and noise control applications.