A general framework to design structural systems for a system-reliability goal is proposed. Component-based structural design proceeds on a member to member basis, insuring acceptable failure probabilities for every single structural member without explicitly assessing the overall system safety, whereas structural failure consequences are related to the whole system performance (the cost of a building or a bridge destroyed by an earthquake) rather than a single beam or column failure. Engineering intuition tells us that the system is safer than each individual component due to the likelihood of load redistribution and al- ternate load paths, however such conservatism cannot be guaranteed without an explicit system-level safety check. As a result, component-based structural designs can lead to both over-conservative components and a less-than-anticipated system reliability.
System performance depends on component properties as well as the load-sharing network, which can possess a wide range of behaviors varying from a dense redundant system with scope for load redistribution after failure initiates, to a weakest-link type network that fails as soon as the first member exceeds its capacity. The load-sharing network is characterized by its overall system reliability and the system-reliability sensitivity, which quantifies the change in system safety due to component reliability modifications. A general algorithm is proposed to calculate modified component reliabilities using the sensitivity vector for the load-sharing network. The modifications represent an improvement on the structural properties of more critical components (more capacity, better ductility), and provide savings on less important members which do not play a significant role.
The general methodology is applied to light steel-framed buildings under seismic loads. The building is modeled with non-linear spring elements representing its subsystems. The stochastic response of this model under seismic ground motions provides load-sharing, system reliability and sensitivity information, which are used to propose target diaphragm and shear wall reliability to meet a building reliability goal. Finally, diaphragm target reliability is used to propose modified component designs using stochastic simulations on geometric and materially non-linear finite-element models including every individual component.
This material is based upon work supported by the National Science Foundation under Grant Nos. 1301001 (Virginia Tech), 1301033 (University of Massachusetts, Amherst) and 1300484 (Johns Hopkins University). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily re ect the views of the National Science Foundation. The author is grateful to the industry partner, the American Iron and Steel Institute, for their cooperation. / Ph. D. / This research proposes methods to design engineering networks for acceptable overall safety. Some examples of engineering networks include electrical systems, transportation systems and infrastructural systems. When any such system is designed, the properties of every individual component (size, capacity etc.) are assigned according to cost and safety requirements. However, it is typically very difficult to reliably quantify the overall safety of the entire system, which is technically known as ‘system reliability’. As a result, there are limited options for engineers to adjust the individual component designs within a system to achieve a pre-specified ‘targeted’ system reliability . This dissertation proposes computational and statistical methods to achieve this.
The proposed methods are applied to a specific engineering system, namely a two story building subjected to ground shaking resulting from an earthquake. Computer models are developed for different scales of the building, beginning from the full building structure, then its individual floors and walls, and finally the individual components that make up each floor and wall. These models are verified with experimental results spanning all three scales. The verified models are then used to both compute the overall system reliability of the building subjected to earthquake ground shaking, as well as to modify its design component-by-component to achieve a targeted system reliability which is different from the system reliability of the original design.
The results indicate that the as-designed reliability of the building system is adequate, but this reliability results from features of the building that are not expected to provide additional safety. The research demonstrates means to obtain this additional safety by redesigning the core functional building components, without relying on the unexpected added safety from ‘non-structural’ components (such as partition walls inside a building). The methods developed herein can be applied to redesign the components of various engineering system networks such that a targeted overall system reliability can be satisfied, resulting in improved performance and life-safety, potentially even at reduced costs.
| Identifer | oai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/74879 |
| Date | 31 January 2017 |
| Creators | Chatterjee, Aritra |
| Contributors | Civil and Environmental Engineering, Moen, Cristopher D., Leon, Roberto T., Wu, Xiaowei, Eatherton, Matthew R., Arwade, Sanjay Raja |
| Publisher | Virginia Tech |
| Source Sets | Virginia Tech Theses and Dissertation |
| Detected Language | English |
| Type | Dissertation |
| Format | ETD, application/pdf |
| Rights | In Copyright, http://rightsstatements.org/vocab/InC/1.0/ |
Page generated in 0.0022 seconds