"Quantification of Uncertainties and Response Predictions Based on Bayesian Inference Models"

Abstract:

Structural health monitoring (SHM) is drawing increasing attention among engineers and researchers due to numerous catastrophic structural failures in the past decades. The main objective of SHM is to ensure the safety and reliability of buildings and structures by identifying potential dangers in the changes of their mechanical properties. Probabilistic methods dominate the area of research in that field and a great deal of methods and approaches have been developed to provide an estimate of the structures' performances and predict their behavior. Model updating is known as a way to calibrate the priors on the existing structural model and provide its posterior characteristics by relying on the gathered measurement data from the monitoring. Due to the complexity of the dynamics of the real structures and generally oversimplified computational models, obtaining a robust posterior is not a trivial task and the presence of modeling errors contaminated with measurement noise is inevitable. The correlation of such errors is often ignored and their independence in time is considered for the analysis. However, it can lead to ambiguous results in the posterior uncertainties associated with the model parameters. To that cause, a good probabilistic model must be able to assess such complex data using proper model assumptions and infer reliable structural parameters that can justify the behavior of the monitored structure and give engineers confidence about its future states. To obtain an inference on the posterior various approaches can be used for monitoring data processing. Methods such as Hierarchical model updating and Gaussian process discrepancy models for structural model parameters and their uncertainty quantification will be reviewed for model updating procedures.

Speaker Bio:

Antonina Kosikova is a postdoc researcher in the department of Civil Engineering department within the group of Professor Andrew Smyth. Her research interests lie in uncertainty quantification (UQ) and dynamic prediction for Structural Health Monitoring using Bayesian methods. Before joining Columbia, she received her Ph.D. at the Hong Kong University of Science and Technology (HKUST) advised by Professor Lambros Katafygiotis and Professor Costas Papadimitriou.