Model Updating of Structural Systems with Incomplete Information: Global Identifiability, Optimal Instrumentation and Uncertainty Quantification

Columbia University, New York NY

Investigators

Abstract

Civil infrastructure systems form the backbone of the economic well-being and progress of any nation. The field of Structural Health Monitoring (SHM) addresses the proper maintenance and timely rehabilitation of such systems, thereby ensuring their continued functionality. In SHM, having a reliable model of the structure is essential. Usually, Finite Element models are used to predict how actual systems, e.g. bridges and buildings, behave under various conditions, from day-to-day service conditions (e.g. traffic) to extreme events (e.g. earthquakes, hurricanes). However, there is a certain level of uncertainty inherent, for example, to the modeling assumptions used by the engineer. In addition, a system changes its characteristics with time due to aging and seasonal effects, and/or sudden extreme events. The aim of model updating is to correct the physical parameters of a model of a structure using information obtained from the real structure so as to get an updated model accurately describing the current condition of the system. This updated model can then be used for predictive, health monitoring and reliability analyses purposes. The importance of this project is in its treatment of constraints related to practical implementation, e.g. not having a sufficient number of sensors, which could have serious implications and could consequently invalidate the results of the entire representation of the model. In the project, a model updating strategy will be conducted according to the following steps. First, tests for global ascertaining and consequent constraints on instrumentation requirements will be formulated, providing minimal instrumentation set-ups to guarantee unique identification. Next, a mode shape normalization scheme will be developed to estimate the complete normalized modal information, which will then be used to estimate the system's physical parameters. Efficient modal parameter comparative measures will be used for mode matching purposes and ways to account for unobserved modes will also is investigated. To account for the uncertainties introduced by uncertain information about the system, environmental fluctuations and the inevitable presence of measurement noise, the algorithm will be extended into a stochastic framework using Bayesian inference as well as an alternative sensitivity based approach to characterize the uncertainty propagation at different steps. The goal of the project is to represent the structure in its present condition for analysis.