Response surface methods are often used to solve reliability analysis problems of complex engineering structure. But the solved reliability index of these methods is mostly geometric reliability. Some errors exist in geometric reliability index compared with the reliability in practical engineering which need to be improved. The importance sampling method in V-space was lead into this paper to transform the basic random variable space to V-space by Rosenblatt transformation and linear orthogonal transformation basing on the designed checking points and its nearby response surface, the complex structure reliability value was solved out by using Monte Carlo importance sampling method. The comparative analysis of several numerical examples’ results have showed that the method put forward in this paper is of high precision and suitable for the reliability solution of complex structures with strong nonlinear performance function.