Core Concepts
Surrogate deep learning models offer faster vibration predictions with trade-offs in accuracy.
Abstract
Introduction: Noise reduction through vibration control in mechanical structures.
Vibration Patterns: Influence of excitation frequency and resonance on vibration patterns.
Design Modifications: Impact of damping elements and beadings on vibration energy.
Finite Element Method: Computational simulation challenges in predicting vibrations.
Deep Learning Models: Introduction of Frequency-Query Operator for accurate vibration prediction.
Dataset and Benchmark: Description of the vibrating plates dataset and evaluation metrics.
Architecture Variations: Comparison of different encoder-decoder architectures for vibration prediction.
Baseline Methods: Evaluation of kNN, DeepONet, and Fourier Neural Operators as baselines.
Experiments and Results: Performance comparison of different methods on the dataset.
Related Work: Overview of machine learning applications in acoustics and scientific fields.
Stats
Surrogate deep learning models offer faster vibration predictions with trade-offs in accuracy.
Quotes
"Vibrating structures radiate sound into the surrounding air." - Jan van Delden et al.
"Our method outperforms DeepONets, Fourier Neural Operators." - Jan van Delden et al.