Modern large frame F, G, & J class gas turbine flow path component design requires the complex integration of multiple design disciplines that traditionally reside with distinct specialists. A traditional design system for an actively cooled gas turbine blade includes aerodynamicists, heat transfer engineers, structural engineers and a failure or lifetime prediction engineers passing information through a manual process with a small number of iterations between the disciplines. Design or manufacturing engineers can also be involved ensuring manufacturability and policing best practices in a predominantly deterministic design system. Over the past few decades robust design or probabilistic design philosophies along with cluster computing advancements have accelerated the release of commercial software that allows for the automation of multiple analytical evaluations at off design points. These software codes allow for process automation of several independent codes executed multiple times at various conditions for automated design of experiment (DOE), and reliability analysis using Monte Carlo or other advanced approximate probabilistic methods across the entire design system.
In this paper, the authors are presenting a novel approach of using a commercially available process integration tool to fully integrate a series of other commercially available tools for a root cause failure analysis of a F class turbine blade rather than a new make design. The analysis incorporated a computational fluid dynamics model (CFD) to define inlet temperatures and pressure profiles, a fully conjugate thermal analysis interacting with a finite element (FEA) solver linked to a proprietary creep lifetime prediction model. A DOE was executed to define the meaningful parameters and help rule out potential causes of failure such as loss of coating or compromised cooling system as a contributing factor of the failure which greatly reduced the amount of time and money needed for the investigation. A probabilistic failure analysis was then executed and surrogate models created for quick probabilistic assessment for different operating conditions. This allowed for validation against fleet history to explain the single engine failure not previously observed. Setting up the non-deterministic models initially allowed for rapid redesign in less than 1 month time with confidence that the true root cause was identified and mitigated. It further allowed for feedback and calibrations to the traditional design system methodology.