Lecture
This lecture explores the concept of faults as equivalent to labels in intelligent maintenance, emphasizing the challenges in learning faulty conditions. It delves into topics such as the Denoising Sparse Wavelet Network, Koopman operator theory, and the Koopman-inspired Degradation Model. The presentation also covers the application of Graph Neural Networks in modelling interactions of dynamical systems and the benefits of Multi-System Dynamics Learning. The instructor discusses the importance of generalization, the use of physics-based modelling, and the potential of Graph Neural Networks as simulators for inferring physics-consistent interactions from particle trajectories.