<p>Considering the nonlinear dependence and the diversity dependence structures between failure modes, a time-variant reliability prediction method based on R-vine copula and an optimal pair-copula model selection method based on tail dependence are proposed. Firstly, by considering the nonlinear dependence between failure modes, the optimal R-vine copula model for the performance functions of each failure mode is established. Secondly, the failure probability of the structural system is derived.A numerical example and a bridge engineering project are used to verify the proposed methods. The results show that diverse dependence structures among failure modes, with failure probabilities overestimated when nonlinear dependence is ignored. When considering dependence among failure modes, using a single dependence structure overlooks certain features, resulting in failure probabilities closer to those under the assumption of independence and higher than those obtained with multiple dependence structures. This study provides a new approach for real-time dynamic reliability prediction and analysis of complex dependence structures with multiple monitoring points.</p>