Structural Reliability Analysis And Optimization Based On Kriging Technique

This thesis mainly focuses on studying efficient methods for structural reliability analysis and reliability-based optimal design, especially developing a semi-analytical method and a simulated importance sampling method based on Kriging technique for structural reliability analysis. These methods are used in the reliability analysis of an offshore Jacket platform. Also, a new reliability-based optimization method is developed which incorporates both kriging reliability analysis and semi-infinite programming algorithm.In chapter 1, basic concepts and methods for structural reliability analysis are introduced, and traditional methods for reliability-based optimization are surveyed. Finally, a semi-variable interpolation method—kriging is discussed.In chapter 2, a semi-analytical method for reliability analysis is proposed. It is based on a semi-parameter interpolation technique, kriging model. At first, general semi-analytical methods are outlined. Response Surface Method (RSM) is often used to construct the limit state function. Due to the restriction of polynomial functions, RSM can not attain satisfactory accuracy for highly non-linearity problems. To avoid the influence of different limit state functions, our method combines the kriging model with the optimization method and can calculate the reliability index with good accuracy. Moreover, it could take advantage of some commercial structural analysis software.In chapter 3, a reliability-based optimization method based on kriging reliability analysis and semi-infinite programming is developed. At first, existing algorithms for reliability-based optimization are surveyed. Generally, this problem is a bi-level optimization problem. Through combining the kriging simulation reliability analysis with semi-infinite programming, the proposed method treats the random and non-random variables equally and accomplishes the integration of structural optimization and reliability analysis, whereby improving the computational accuracy and efficiency.In chapter 4, a simulated importance sampling approach is presented for structural reliability evaluation. This technique combines the importance sampling method with kriging simulation technique in structural reliability analysis, and can greatly decrease the sample number of Monte Carlo method.In chapter 5, Victory CB11F offshore platform is taken as the example and kriging reliability analysis and simulation importance sampling methods are respectively applied to evaluate its structural reliability. For Jacket platforms, the Limit State Function (LSF) of structure is impossible to have explicit expressions. In this paper, two methods have high efficiency as certain improvements of RSM and importance sampling method, and avoid their respective limitation. The methods in the present paper may provide a new way for reliability analysis of offshore engineering.