Abstract: A sample selecting method was proposed based on polar transformation， and reliability analysis of rolling bearing clearance was performed combining the method and sparse response surface. In every step of response surface construction， the critical samples were added and fitted according to the clustering feature and distinguishability of safe and failure classes of samples in a plane after polar transformation. The most significant terms of polynomial response surface were chosen according to error prediction criteria and cross-validation method. The limit state function of rolling bearing clearance reliability problem was built comprehensively considering the factors of interference fit， temperature variation and centrifugal force， and the working clearance was calculated employing the finite element method. Finally， the reliability of bearing clearance was calculated based on the explicit style of limit state function. This study provides theoretical basis for rolling bearing design and an efficient way for analyzing implicit reliability problems with multiple dimensions.

Key words: rolling bearing clearance, polar transformation, error prediction, sparse response surface, finite element