Real-Time Health Monitoring of Gas Turbine Components Using Online Learning and High-Dimensional Data

Sponsor: Department of Energy-University Turbine Systems Research Program

Collaborators: Georgia Tech and Pratt & Whitney

This project will use data analytics to address challenges surrounding “Big Data” sets. Data analytics associated with big data offers unique capabilities to gas turbine operation through its ability to synthesize information from physical sensors and infer additional information from “virtual” sensors. Advanced gas turbine test facilities will be interrogated using state-of-the-art instrumentation techniques to build an open data collection supporting predictive algorithm development for combustors and turbines. Ultimately, these data will be correlated with physics-based models and first-principle relationships to improve component life predictions. Highly-resolved data generated from a combustor test rig and a turbine test rig during both normal operation and with “seeded” faults, will be used as the basis for the Big Data sets. The test conditions in the two test facilities will include common, critical events that occur in the operation of power plants.As one example of this research, the rim seal and under-platform regions are susceptible to damage when hot main gas path fluid is ingested through the rim seal. There is a lack of viable methods for real-time rim seal performance monitoring. As illustrated below, time-resolved pressure sensors (PA, PG) were successfully used to estimate rim sealing effectiveness () and identify faults to the purge flow rate ().