Researchers have developed new physics-informed machine learning strategies to improve global sensitivity analysis (GSA) by effectively combining physics-based models with experimental data. The study explores two machine learning techniques, deep neural networks (DNN) and Gaussian process (GP) modeling, and two methods for integrating physics knowledge: enforcing physics constraints in loss functions and using simulation data for pre-training followed by experimental data for updating. Results indicate that DNN-based models offer smaller bounds on sensitivity estimates compared to GP models, with applications demonstrated in additive manufacturing and lake temperature modeling. AI
IMPACT This research could lead to more accurate engineering system analyses by better integrating simulation and real-world data.
RANK_REASON The cluster contains an academic paper detailing novel research methods.
Read on Hugging Face Daily Papers →
- arXiv
- Berkcan Kapusuzoglu
- DNN
- Additive Manufacturing
- Deep Neural Networks
- Experimental data
- Gaussian process
- Hugging Face
- Information Fusion
- lake temperature modeling
- machine learning
- Physics knowledge-based transfer learning between buildings for seismic response prediction
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
