Structural Time-variant Reliability Analysis Method Under Single Failure Mode

Structural reliability analysis has attractive in both academia and industry.Compared with traditional static reliability analysis method,the time-varying reliability analysis method considers the effect of time-varying factors such as dynamic load,structural degradation and other time-variant uncertainties on structural performance.Therefore,the time-variant reliability analysis method has high applicability in engineering and is more consistent with the actual situation of mechanical structure that the performance is changing with time.In addition,in practical engineering,the structural time-variant performance function of structure is always implicit function.In this case,the traditional time-variant reliability analysis methods are difficult to be applied.To address the issue,the corresponding surrogate model can be constructed to improve computational efficiency.In this thesis,the time-variant reliability analysis method of structures under single failure mode is studied carefully.The main research contents and contributions are as follows:(1)Time-variant reliability semi-analytical extremum method based on mean value point expansion.The existing structural time-variant reliability analysis methods can be roughly divided into analytical method and extremum method.The analytical method has high efficiency but low precision.The computational efficiency of extremum method cannot be guaranteed under keep the accuracy level.Therefore,the analytical method and extremum method are combined to transform the time-variant performance function into a time-invariant performance function,and the approximation is carried out at the mean value of points.Under the condition of ensuring accuracy,the computational efficiency and robustness are improved.(2)Time-variant reliability analysis method based on Kriging surrogate model and parallel computing.Traditional time-variant reliability analysis method can effectively solve time-variant reliability problems for structures with explicit performance functions,but are difficult for implicit performance functions.Existing methods use surrogate model to represent time-variant performance function,but there are still defects in computational efficiency.To address the issue,the parallel strategy and new sample learning function are bridged,and a time-variant reliability analysis method based on Kriging surrogate model for a single failure mode is proposed.Since the strategy of parallel computing is adopted in the updating process of Kriging surrogate model,the computing efficiency can be significantly improved by searching multiple sample points at one time.(3)Time-variant local reliability sensitivity analysis method based on Kriging surrogate model.When performance function is implicit function,the reliability sensitivity analysis is computationally expensive.In addition,there is still lack of an effective reliability sensitivity analysis method for implicit performance time-variant function.To address the issue,the traditional time-invariant sensitivity analysis method is improved by considering the time-variant characteristics of variables.Subsequently,a local sensitivity analysis method is established for systems with implicit-time-variant performance functions.