AbstractThe extremum probability density evolution method (EPDEM) was introduced for the life-cycle risk analysis of the roofs of low-rise buildings subjected to typhoons. This method can use small sample sizes to account for the uncertainties of winds during the entire track of a typhoon and the changes in resistance of structures over time. This study generated typhoons in Shanghai, China using the Yan Meng wind field model. The EPDEM-based wind load analysis combined with a wind tunnel test was used to generate the probability density curve of typhoon wind loads. To improve the computational accuracy, a modified k-means clustering algorithm was introduced in representative point selection. This approach was compared with the Monte Carlo method to show its effectiveness and accuracy for practical applications. Furthermore, risk analysis based on the EPDEM, considering changes in roof resistance within the lifetime of the building, was carried out. The results show that the risk is time-variant and is influenced greatly by the resistance of a structure, and the reliability of the structure declines rapidly after 50 years. A methodology is proposed to predict the vulnerability curves for a new structure at different service times.