LAUSR

Response surface method (RSM) provides an efficient way to performance reliability analysis for structures. However, the locations of samples are very important due to its great influence on computational accuracy and efficiency of RSM for reliability analysis. To further improve computational accuracy and efficiency of RSM, new methods of selecting samples are proposed based on a new starting center point, 2n + 1 directions and a linear interpolation. The objective of the new methods is to find samples which are close to limit state function (LSF) around design point, thus the fitting precision of response surface function (RSF) to LSF can be improved, and a quadratic polynomial without cross terms is employed as the RSF in each iteration. Then improved iterative RSMs are formed. Two mathematical examples and a truss structure are employed to demonstrate the accuracy and efficiency of the proposed RSMs. Results show that the proposed RSMs can improve the fitting precision of RSF to LSF and achieve more accurate results with relatively high efficiency.