Thermo-mechanical fatigue (TMF) resistance characterization and life assessment are extremely important in the durability/reliability design and validation of vehicle exhaust components/systems, which are subjected to combined thermal and mechanical loadings during operation. The current thermal-fatigue related design and validation for exhaust products are essentially based on testing and the interpretation of test results. However, thermal-fatigue testing are costly and time consuming, therefore, computer aided engineering (CAE) based virtual thermal-fatigue life assessment tools with predictive powers are strongly desired. Many thermal-fatigue methods have been developed and eventually implemented into the CAE tools; however, most of them are based on deterministic life assessment approach, which cannot provide satisfactory explanation for the observed uncertainties introduced in thermal-fatigue failure data. In this paper, a probabilistic thermal-fatigue life assessment procedure is developed. The life distribution function constructed is expressed as a function of cumulative plastic strain amplitude, peak operating temperature and reliability level. The values of the parameters in the probabilistic distribution function are estimated from extensive V-shape specimen tests and associated data correlation. The developed procedure is then implemented into the CAE environment and applied to the life assessment of exhaust components/systems including a fabricated manifold, and its effectiveness is demonstrated.