1. Uncertainties in Thermal Barrier Coating Life Prediction by Karl A. Mentz A Thesis Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements for the degree of MASTER OF SCIENCE IN MECHANICAL ENGINEERING

2. Thermal Barrier Coating (TBC) Life Overview: Thermal barrier coating life is usually predicted based on oxidation life of the bond coating. The effect of thermal fatigue on coating life has been studied but is generally not included in life predictions due to the complex crack dynamics and interactions with the coating and substrate. An attempt to quantify the effect of different thermal cycling on coating life is made using test data. Method:Coating life data was obtained from multiple sources and tested in multiple test cells. Data analysis is first done to determine a coating life for the base test cycle (short cycle). This is to quantify the manufacturing and test variation and give an accurate life number for comparison. The data for the long cycle will then be compared to the base short cycle coating life to determine if a difference in coating life can be detected. Sample Information Material:Cobalt base superalloy Bond Coat:NiCoCrAlY. Applied by low pressure plasma spray (LPPS) TBC:ZrO2 partially stabilized with 7wt % Y2O3. Applied by air plasma spray (APS)

3. Thermal Barrier Coating (TBC) Life TBCs reduce substrate metal temperatures by being an insulating layer between the hot gas on the outside of a component and the base metal. The coating consists of a ceramic coating to provide thermal protection and a metallic bond coat to provide oxidation protection and provide a bonding surface for the ceramic. Temperature gradient in a TBC coated part [1]. Coating structure on a substrate material. [2].

4. Thermal Barrier Coating (TBC) Life TBC failure is believed to be due to a combination of oxidation damage and crack damage from thermal cycling. Failure Modes Oxidation: A thermally grown oxide layer (TGO) forms at the interface between the bond coat and the ceramic coating. Spallation of the ceramic coating occurs when the TGO reaches a critical thickness where the internal stresses exceed the TGO properties and delamination of the ceramic occurs. Thermal fatigue: Cracks propagate through the ceramic and TGO layers, driven by thermally induced cyclic stresses until the cracks link together and the coating spalls. Test Cycles Short cycle:Typical cycle used for testing aerospace components. Thermal fatigue damage should decrease life compared to pure oxidation failure mode. Long cycle:Test cycle used for industrial gas turbine engines Oxidation failure mode should dominate. Expect longer life due to decreased cycling.

5. Test Method Burner Rig Testing Burner rig schematic [3] Test bar geometry [3]

6. Short Cycle Test Data Coatings produced at five different sources. 1)Secondary production source 2)Development process at source 3 3)Main production source 4)Main production source 5)Development process at source 4 Testing Done in 4 different test cells. Different coating properties Coating life by source. “Hot time” is the amount of time the test specimen spends at the maximum test temperature of 1950F. Hot time is used to compare life from different test cycles and can be directly used to calculate TGO growth. Typical failed test specimen. Failure defined as spallation of ~50% of the hot zone area. Hot zone area is the face of the test bar that is exposed to the burner rig [3].

7. Relative coating life by coating source

8. Data Fitting Failure data fitted using Weibull distribution. Can mimic behavior of normal, exponential, and other distributions Will not predict negative lives Functions from multiple sample sets can be combined to predict an overall population life. Can be used to fit small data samples. Alpha: Scale parameter Beta: Shape parameter x: time

9. Weibull Plot. Source 4 Coating Life

10. Relative life data by test cell and source.

11. Short Cycle Test Data Coating Life by Test Cell. Failure data was fit with a Weibull distribution. Weibull parameters used to predict the average coating life for each source and test cell. Data shows source 3 has a higher life than source 4 Test cell 3 produces higher coating life results than test cell 2.

12. Weibull plot of Source 4, Test cell 2 data

13. Long Cycle Test Data Long cycle test data fitted with Weibull distribution and average life predicted. All data was from source 4. Test cell 2 showed results that agreed with expected trends, the long cycle testing produced longer coating life. Test cell 4 showed drastically reduced life for the long cycle testing. Sources of error in the data; 1.Small amount of data available for long cycle tests (5 samples each) 2.Little information on test cell 4 and no test data to determine the impact of the test cell on life results. 3.Extremely short life indicates an anomaly in coating specimens or test conditions for samples in cell 4. Relative life for long cycle testing by test cell.

14. Conclusions Coating life is extremely variable. Coating life may vary by a factor of 3x or more within coatings produced by the same source. Test conditions can introduce significant variation in addition to the actual coating life variation by source. Differences in coating life due to thermal cycle damage is small compared to production and testing differences. Life predictions for thermal barrier coatings are limited by the consistency of production and testing.

15. References 1.Dynamic-Ceramic LTD.Crewe, England http://www.dynacer.com/coatings.htm 2.IMR Test Labs. Ithaca, NY. http://www.imrtest.com/what_we_do/Thermal_Spray_Coatings/ http://www.imrtest.com/what_we_do/Thermal_Spray_Coatings/ 3.DeMasi, J.T., Sheffler, K.D., Ortiz, M. “Thermal Barrier Coating Life Prediction Model Life Development.” United Technologies Corporation, 1989. NASA Contract NAS3-23944