Data-driven modelling (identification) of non-stationary systems

We develop robust methods to capture time-dependent and nonlinear dynamics, including quantification of uncertainties derived from experimental and operational conditions.

We specialize on Bayesian statistical time-series methods, which can, in an implicit or explicit manner, integrate the effects of environmental and operational variables on the dynamic properties of structures.

Want to know more?

We specialize in the development of system identification methods based on non-parametric and parametric Bayesian time-series models for different types of non-stationary behaviour, including:

• Fast, continuously varying non-stationary dynamics (rotor dynamics, robotic manipulators, wind turbines)
• Slow, continuously varying non-stationary dynamics (mid and long-term changes in bridges, maritime structures, wind energy facilities)
• Switching non-stationary dynamics featured in systems with changing operational regimes, including damaged conditions
• Estimation and tracking methods for slowly varying systems