Rapid Stochastic Response Estimation of Dynamic Nonlinear Structures: Innovative Frameworks and Applications
Li, Bowei
2022
Abstract
Response estimation of structural systems characterized by large numbers of degrees of freedom (DOF) is universally needed in practical engineering for performance evaluation and design. When subjected to extreme events, e.g., destructive earthquakes or hurricanes, such structural systems typically experience significant damage and therefore nonlinearity in their response. This can lead to extremely computationally cumbersome problems. Further, if repetitions of such analyses are required, e.g., in uncertainty propagations or optimizations, the associated computational burden can quickly become intractable. This has created a need for efficient response simulation methods for nonlinear dynamic structural systems. To address this need, innovative mechanics and data-driven approaches are investigated in this research. The mechanics-based approaches address the problem by exploring strategies for increasing the efficiency of the response simulation algorithms. In particular, a novel adaptive fast nonlinear analysis (AFNA) algorithm is developed for solving nonlinear structural systems discretized at the level of the fibers or stress resultants. In the proposed AFNA scheme, algorithm configurations, such as the bases for model order reduction (MOR) and time step sizes, are determined adaptively. Compared to direct integration schemes, the AFNA approach is seen to have comparable accuracy with, however, speedups of an order of magnitude. Further, the solution scheme is embedded into a sampling-based wind reliability analysis framework that enables not only more accurate reliability assessment but also a full range of time history analyses for shakedown and beyond. The data-driven approaches, on the other hand, center on training efficient surrogates of the original high-fidelity model. In particular, the data-driven approaches were developed by leveraging MOR and time series metamodeling. Firstly, a data-driven Galerkin projection is introduced to reduce the response space of the original structural system. Subsequently, techniques based on the multi-input-multi-output nonlinear autoregressive model with exogenous input (MIMO NARX) system identification and long-short term memory (LSTM) deep learning are introduced to capture the dynamics of the reduced system. These approaches are capable of accurately reproducing both the global displacement and local hysteretic curves, with speedups over the high-fidelity simulations of three to five orders of magnitude. As a separate application of metamodeling, a real-time risk forecast framework for hurricane-induced damage in building systems is developed, where metamodels were created to reduce the computational demand to enable the real-time use of high-fidelity performance assessment frameworks. The methodology was used to show the strong potential of metamodeling for informing early emergency response.Deep Blue DOI
Subjects
Rapid nonlinear time history estimation, Metamodeling, Adaptive fast nonlinear analysis, Real-time risk forecast, Uncertainty propagation
Types
Thesis
Metadata
Show full item recordCollections
Remediation of Harmful Language
The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.
Accessibility
If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.