Deep Learning Framework for Gas Turbine Performance Digital Twin and Degradation Prognostics from Airline Operator Perspective

•A data-driven Performance Digital Twin is developed just based on the gas turbine operational data rather than the physics-based performance model, which is generally unavailable for the airline operator.•The derived multi-dimensional health features based on PDT improved the RUL prediction accuracy by enhancing the input feature space of the CNN- based prognostics network. To the best of our knowledge, the proposed framework outperforms the state-of-art approaches evaluated on the NCMAPSS data set.•This purely data-driven framework is developed from an airline operator perspective, where only operating conditions descriptors and physical sensor measurements are available in practical application. Therefore, the signals from virtual sensors and health status parameters or other information derived from the physics-based models are excluded from the analysis. Various expert knowledge is incorporated into the framework to improve the prognostics performance.