Events Calendar

CEE Oral Dissertation Defense

Speaker: Miguel Hernandez-Garcia, Astani CEE Ph.D. Students

Talk Title: Analytical and experimental studies in modeling and monitoring of uncertain nonlinear systems using data-driven reduced-order methods

Abstract:
Most of the available data-based methodologies developed for system identification and health monitoring of complex nonlinear systems can be considered to be deterministic in nature. These approaches use experimental measurements to characterize the complex systems by means of nominal mathematical (e.g., parametric or non-parametric) models, while neglecting the effects of aleatory and epistemic uncertainties that can be present in real structures. The inherent stochastic nature of the systems’ components (i.e., randomness in structural, geometric and material properties); the variability in environmental and operational conditions; and the uncertainties associated with the modeling, measurement and data analysis process can lead to unreliable description and characterization of complex nonlinear systems. Consequently, in order to develop robust and reliable models of nonlinear systems, it is imperative to address the issues of quantification and propagation of uncertainties.
This dissertation compiles analytical and experimental studies focused on implementing, analyzing, and validating promising and robust data-driven methodologies to build high-fidelity reduced-order models of uncertain complex nonlinear systems. These data-based reduced-order methodologies were used in structural health monitoring applications, and in the modeling of critical structural components. Experimental datasets from dynamic tests performed in a small-scale lab structure at Los Alamos National Laboratory (LANL); a re-configurable test structure designed, built, and tested at University of Southern California (USC); a scaled-down six-story steel-frame laboratory structure at the National Center for Research in Earthquake Engineering (NCREE); and a seven-story full-scale reinforced-concrete structure at the UCSD-NEES facilities, were used to evaluate the effectiveness and reliability of the data-based reduced-order models for detecting, locating and quantifying structural changes. In addition, sensor fault-detection and identification techniques based on statistical monitoring using latent-variable techniques, were implemented and evaluated for detecting and identifying faulty sensors using measurements from an actual cable-supported bridge in the metropolitan Los Angeles (CA) region. Finally, a general methodology for developing probabilistic reduced-order models of critical structural components from experimental measurements was proposed. This methodology was used to develop a probabilistic data-based reduced-order model to characterize the mechanical behavior of a particular type of lap bolted joints with an inclined interface directly from experimental data obtained from dynamic tests performed at Sandia National Laboratories (SNL).

Location: Kaprielian Hall (KAP) - 209 Conference Room

Audiences: Everyone Is Invited

Contact: Evangeline Reyes