Abstracts – Browse Results
Click on the titles below to expand the information about each abstract.
Viewing 1 results ...
Wang, J (2004) Hybrid genetic algorithms for reliability assessment of structural system, Unpublished PhD Thesis, , City University of New York.
- Type: Thesis
- Keywords: accuracy; failure; genetic algorithms; reliability; deterioration; learning; data mining; bridge
- ISBN/ISSN:
- URL: https://www.proquest.com/docview/305204407
- Abstract:
Although the theory of structural system reliability has greatly matured over the last decades, widespread implementation of reliability methods in engineering practice has not yet taken place. One main reason for the lag between the theoretical developments and implementation is attributed to the limitations of most available reliability analytical techniques in their ability to account for one or more of these factors: (1) Accurately model the behavior of structural systems at high loads; (2) Consider different performance criteria; (3) Identify multiple equally important failure modes; (4) Account for load combinations; (5) Solve time dependent problems; and (6) Provide accurate solutions in a computationaly efficient manner. To help resolve some of these perceived deficiencies, this Ph.D dissertation develops flexible yet efficient simulation-based methods that can be easily adapted for routine application when solving various types of structural reliability problems that are encountered in engineering practice. Two hybrid Genetic Search Algorithms are developed to efficiently determine the probabilistically dominant failure modes of complex structural systems and determine their reliability index values. One of the proposed hybrid methods combines the benefits of the Gene Expression Messy Genetic Algorithm (GEMGA) and the Shredding Genetic (SGA) operator to improve the efficiency of the search for failure modes through their linkage learning processes. The other proposed algorithm takes advantage of the pattern identification ability of Data Mining (DM) techniques to supplement the capacity of GA operators to explore new significant search domains. New data analysis schemes including an exploitation process based on the Tabu local search procedure are introduced in the algorithms to obtain accurate reliability index values and quantify the contributions of various random variables to the dominant failure modes. The efficiency and accuracy of the proposed GA methods are verified by applying them to solve a range of benchmark reliability problems. By linking the proposed Genetic Algorithms to general-purpose finite element programs, the reliability of any structural system with any type of material behavior can be solved. This dissertation demonstrates the applicability of the proposed methods for solving realistic structural problems by performing the reliability analysis of cable stayed and suspension bridges subjected to combinations of loads and accounting for the geometric nonlinearity and the time-dependent deterioration of structural members.