Industrial AI Laboratory
Research
Research Area
I Physics-guided AI for Digital Twins
The IAI Lab focuses on the development of novel physics-guided artificial intelligence (AI), aiming to develop neural computational methods with unprecedented performance. We study diverse methods for hybridizing the AI framework and our engineering knowledge with expansive physics models, providing the fidelity and adaptability of data-driven methods for a variety of domains. Our research offers the potential for versatile digital-twin applications that require real-time, accurate, interactive, and interpretable analyses. In particular, our lab places a strong emphasis on deep learning-based reduced order modeling (ROM) approaches to address the computational bottlenecks of high-fidelity simulations. By embedding domain-specific physical constraints into data-driven ROM frameworks, we seek to achieve fast and scalable functionalities that enable the extension of analysis domains beyond traditional boundaries and further facilitate modeling of complex systems.
I Mechanism-inspired Deep Neural Networks
We develop novel deep neural networks that are intrinsically inspired by the underlying mechanisms of engineering systems. Our aim is not only to achieve high predictive performance but also to enhance interpretability in a way that aligns with physical phenomena. This innovative method advances the field of neural network research and opens up new possibilities for creating more intuitive and explainable applications across a range of scientific and engineering disciplines.
