The state function based on fatigue accumulation model has a great influence on accuracy of fatigue reliability analysis of components. A nonlinear fatigue accumulation model considering the interaction of loads is proposed in this paper. The square ratio of front and back loads is presented as a new load interaction factor and embedded into the conventional toughness exhaustion model; thus, this improved model can reflect the load sequence and interaction effect simultaneously. Moreover, the fatigue state function using the improved model is constructed and analyzed by the probability density evolution method. Then experimental data of four materials are used to verify the proposed model, and results show that the residual life fraction predicted by the proposed model is accurate. Furthermore, accuracy and efficiency of the timevarying fatigue reliability analysis using the proposed model is validated using experimental data from a carbon steel test and a helical compression spring test.