Probabilistic lifetime model for atmospherically plasma sprayed thermal barrier coating systems

•Calculated lifetimes reflect experimental data also with regard to data scattering.•Stress field inversion due to TGO growth was identified as main failure mechanism.•The probabilistic lifetime model concept is based on FE stress field analyses.•Subcritical crack growth is taken into account by fracture mechanical calculations. Calculations of atmospherically plasma sprayed thermal barrier coating durability were facilitated by the development of a numerical lifetime model including probabilistic fracture mechanical analyses of thermally induced topcoat stress field evolutions. The stress distributions were determined in finite element analyses taking into account oxide scale growth and topcoat sintering as transient degradation effects. The influence of interface microstructure was investigated by implementing two different interface approximation functions. Subsequent fracture mechanical analyses of subcritical crack growth were performed at numerous different and permanently assigned abstract crack positions. A comparison of the transient energy release rate to its critical value, which depends on crack length and therefore position, results in statistical distributions of system lifetime as a function of simulated thermal cycling conditions. The model was calibrated by presetting an experimental lifetime distribution which was determined in thermal cycling experiments performed at a burner rig facility. The associated cycle-dependent calibration parameter reflects the effect of fracture toughness increase for increasing crack lengths. Experimental reference values for system lifetime were found to be reproduced by the lifetime model. The stress field inversion directly correlated to oxide scale growth rate was identified as the main failure mechanism. The expectation values and standard deviations of the calculated lifetime distributions were found to be in accordance to the experimentally obtained lifetime data and the data scattering typically observed in thermal cycling.