My research interests include structural health monitoring, digital twin, Bayesian inference, deep learning, and hybrid modeling.
1. Digital twin
Building digital twins for large-span bridges and wind turbines based on Bayesian model updating, Bayesian filtering, and hybrid modeling, for damage identification, input load (wind loads, traffic loads, etc.) estimation, and structural response prediction.
2. Physics and Data-driven hybrid modeling
Integrating physics-based modeling techniques (finite element models, ordinary/partial differential equations, state-space models, etc.) and data-driven modeling methods (supervised learning, unsupervised learning, reinforcement learning, and adversarial learning, etc.) to build hybrid models and improve prediction accuracy, generalizability and interpretability.
3. Bayesian system identification
Applying Bayesian inference and Hierarchical Bayesian method for modal identification, model updating, data fusion, and uncertainty quantification; Developing Bayesian filtering and smoothing methods, including Kalman filter, nonlinear Bayesian filter, general Gaussian filter, partial filter, and RTS smoother, to accurately identify structural parameters, input loads and system states.