Thermal cycle parameter effects on the stress state, failure mechanisms and life prediction of thermal barrier coatings

Date of Completion

January 2005


Engineering, Materials Science




The primary objectives of this study are (1) to examine the effects of bond coat composition and processing method on the trends in photoluminescence piezospectroscopy (PLPS) characteristics as a function of cycles for two [(Ni, Pt) Al] bond coated (outward diffusing and inward diffusing) thermal barrier coatings (TBCs), (2) to define the spallation mechanisms for the two TBCs in furnace cyclic tests at three temperatures and two hold times, and (3) use the above data to generate mechanisms-based TBC life prediction methods. ^ For both TBCs evaluated using the 1-hour tests, progressive rumpling of the TGO/bond coat interface is responsible for failure occurring at the YSZ/TGO interface at all 3 temperatures. For the 24-hour tests, failure occurs at the TGO/bond coat interface and is driven by the stored strain energy in the TGO. A mechanism for rumpling in these TBCs is proposed that involves plastic deformation at the maximum temperatures in the cycle and cracking of the ceramic upon cool-down. The single-valued relationship between the TGO thickness and rumpling amplitude, independent of temperature and cycle time, also confirms this notion that the in-plane TGO growth that occurs at the peak temperature directly appears as rumpling displacements of the TGO. ^ The TGO stress, measured by PLPS, exhibits a monotonic decline with cycles. This relationship is independent of cycle temperature for both TBCs, and changes slightly with cycle hold time. There is a constant TGO stress at failure for both TBCs independent of temperature. Correlations between the maximum TGO stress and spallation life and between the slope of the TGO stress versus cycles curve and spallation life are identified: longer life specimens exhibit higher initial TGO stress and shallower slopes. ^ Two-level cyclic tests consisting of multiple temperatures and multiple hold times have been conducted for the first time. The slopes of the stress versus life curves obtained from these tests indicate that while a change in the cyclic temperature amplitude does not affect the normalized stress response of the system, a change in the hold time does alter it. This is consistent with a spallation mechanism that is independent of temperature but changes with hold time. This necessitates that TGO stress versus cycles or TBC life data are generated separately for every cycle with a different hold time. ^ Temperature-blind remaining life predictions were made using three spectral features, the TGO stress, peak width and the standard deviation of stress, using standard regression analyses. (Abstract shortened by UMI.)^