Research And Application Of Structural Time-variant Reliability Analysis Method

As is known,there are many uncertainties inherent in the engineering structures,and the deterministic analysis methods are difficult to measure the influence of these uncertainties on the structural behavior.Structural reliability analysis provides theoretical basis and calculation approaches for the propagation and quantification of uncertainty in engineering structural systems.In addition,due to the complexity of the environment in which the engineering structure is in service,the structural parameters and loads are constantly changing over time,resulting in the time-variant structural reliability.Therefore,time-variant reliability analysis of engineering structures has received increasing attention.In the structural time-variant reliability analysis methods,the extreme-value distribution-based method can easily transform the time-variant reliability analysis problem into the corresponding time-invariant reliability problem,which is extensively implemented in engineering practice.However,the double-loop approach is usually adopted in the traditional extreme-value distribution-based structural time-variant reliability analysis method,which cannot balance the accuracy and efficiency when solving strong nonlinear and high-dimensional problems.In order to further develop the extreme-value distribution-based structural time-variant reliability analysis method and enrich the structural time-variant reliability theory,a single-loop structural time-variant reliability analysis method based on a decoupling strategy and the reconstruction probability distribution is proposed in this thesis.In fact,the input random process involved in the structural time-variant reliability problem should be specified as a function of time and random variables.Therefore,in the proposed method,the expansion optimal linear estimation is first employed to discretize the stochastic process as a set of random variables.Then,by analyzing the internal connection between the principle of first-passage failure and the extreme value response,as well as based on the extreme-value distribution-based structural time-variant reliability analysis method,a decoupling strategy that decouples the stochastic process and degenerate processes is proposed.This decoupling strategy can obtain the equivalent extreme limit state function(EEV-LSF),and construct a single-loop approach for structural time-variant reliability analysis.To improve the accuracy and robustness,the Box-Cox transformation is applied to get a transformed EEV-LSF,and the improved fractional exponential moment-based maximum entropy method is employed to robustly derive the probability distribution of the transformed EEV-LSF.Since a large number of random variables could be always involved,a weighted sampling method is applied for fractional exponential moments assessment with accuracy and efficiency.Once the probability distribution of the transformed EEV-LSF is captured,the structural time-variant failure probability can be readily computed by integrating the probability density function over the failure domain.This thesis conducts time-variant reliability analysis and parametric analysis on three classical numerical examples and one engineering case.The analysis results of traditional structural time-variant reliability methods are also given for comparisons.The research results demonstrate that the structural time-variant reliability analysis method proposed in this thesis can balance the accuracy and efficiency,and can provide guiding suggestions for engineering practice.