Component bench testing is a basic method to validate the component fatigue life. However, the component bench testing takes long time and is costly. With the development of more powerful computer and CAE simulation techniques, virtual CAE simulation method becomes more important in the component design, optimization, and validation due to its efficiency and low cost. Fatigue life of components of exhaust system is a critical characteristic and it is not deterministic but statistical phenomenon. Thus, a probabilistic approach is necessary. Variations and reliability of fatigue life can be considered in physical testing by testing more samples. However, how to account variations from manufacturing and testing in virtual CAE simulation is a big challenge. In this paper, a virtual CAE fatigue life prediction of components of exhaust system by probabilistic approach is studied and proposed. Different methods such as the traditional stress and fatigue life analysis and the Verity method are used to predict fatigue life mean value. Weibull distribution is used to calculate reliability of the predicted fatigue life. The analysis results by traditional stress method and the Verity method are compared and correlated to the test results. Key parameters of Weibull distribution are also discussed for proper representation of manufacturing and testing process.