Enhancing Gas Turbine Component Life Predictions Using Fleet Data and Probabilistic Methods

Flexible operation of Industrial Gas Turbines is becoming more commonplace as a solution to reduce life cycle cost and increase sustainability. OEMs are developing digital twin technology to understand the effect of this variable operation on component life. This technology can also be used to accurately forecast turbine operation to a potential future maintenance period or end of service interval. The consideration of available fleet machine data is required to manage the risk and enable these future predictions of damage in life-limiting components. Probabilistic methods, utilizing Monte-Carlo approaches, can be leveraged to model physically realistic operation, material variation, and uncertainties in important parameters such as metal stress and temperature. A key technology in leveraging Probabilistic methods is the ability to translate engine operation parameters into component stress and temperature variables using Reduced Order Models (ROMs). This in conjunction with damage models, representing mechanisms such as creep, fatigue, oxidation, and corrosion, provides the mathematical structure necessary to produce many hundreds of thousands of potential damage results within a reasonable computation time. This paper aims to outline techniques to characterize a subset of fleet data to provide fleet models of operation. The integration of these fleet models into Probabilistic frameworks and how they are used to manage data driven upper bound damage predictions where operational data is not available will also be discussed.