A digital twin-oriented data-physics-AI integrated framework for RUL prediction of gas turbine blade

As one of the most high-fidelity and high-end turbomachines, gas turbine (GT) has been broadly applied to generate power in various industries such as energy, oil & gas and transportation, because of its unique characteristics of high efficiency, low-emission, and flexible in startup and shutdown. It is often subject to the long-term harsh environment under high pressure, high temperature and high corrosion so that many failure modes like crack, creep, and fouling are inspected in different key components. In the past years predictive maintenance has become a powerful approach to ensure the safety and reliability of a gas turbine, which seamlessly integrates the novel digital technologies such as big data, statistical modeling and artificial intelligence analysis. This paper presents a data-physics-AI integrated framework to predict remaining useful life (RUL) of the gas turbine blade, with the ultimate purpose of digital twin-oriented predictive maintenance of the system. This framework includes data-driven machine learning, physics-based lightweight modeling, and model tuning for fatigue-based RUL prediction of a turbine blade. A comparison study with the tradition methods is conducted to show the advantage of the proposed framework using the data from a real-world gas turbine.