<p>Crack propagation in the top coat is one of the primary failure modes of the atmospheric plasma-sprayed thermal barrier coating system (APS-TBC). However, establishing a clear relationship between it and the overall coating failure remains a challenge. This research proposes a probabilistic statistical method to quantify the risk of top-coat spallation induced by thermally grown oxide (TGO) growth and thermal mismatch. The extended finite element method (XFEM) was used to simulate residual stress and crack propagation behavior under cyclic loading, with a focus on examining the effects of peak temperature, cooling rate, and holding time, as well as the initial length, angle, and location of the crack. Subsequently, a failure criterion and life prediction model for TBCs with top-coat cracks were established based on finite element analysis (FEA) results. The failure probability and reliability were analyzed using the segmented projection method and Monte Carlo theory. Furthermore, the influence of crack distribution on coating failure was analyzed, leading to a proposed method for predicting multi-crack failure probability. Finally, experimental validation confirmed the accuracy of the simulation results, with a prediction bias of 7.5% for the single-crack model and 6.2% for the multi-crack model. The present study provides a new approach for life prediction and reliability evaluation of TBCs.</p>