Improvement and Demonstration of a Machine Learning-Based Digital Twin Framework for Rotorcraft Life Assessment Using In-Service Operational Data

Abstract:
Traditional safe-life methodologies for rotorcraft structural components rely on deterministic safety factors to account for uncertainty in loads, material properties, and operational usage. While effective for ensuring safety, these approaches lead to early retirement lives and reduced aircraft availability. This paper presents an updated digital twin-based probabilistic framework for rotorcraft component fatigue life assessment that integrates a probabilistic stress–life (S-N) material model, machine learning-based load estimation from flight data, and Monte Carlo uncertainty propagation. The approach is demonstrated for a critical location on the CH-146 Griffon main rotor yoke. Compared with earlier work, the present study advances the framework through independent validation of the load-estimation model and application to available in-service flight data from multiple mission categories. A probabilistic sensitivity analysis is used to examine the separate and combined effects of material variability and load-estimation uncertainty on fatigue life, cumulative probability of failure, and hazard rate. For the CH-146 demonstration case, the results indicate that the material fatigue strength uncertainty has a major impact on the lower tail of the life distribution and the corresponding reliability-based life, whereas load-estimation accuracy uncertainty has a secondary influence on risk metrics. The application of the digital twin framework to operational, search and rescue, and training mission data further shows that mission-specific usage variability plays an important role in the evolution of fatigue damage accumulation and structural risk. Overall, the proposed framework provides a more informative basis for risk-based rotorcraft life assessment by explicitly quantifying uncertainty and incorporating aircraft-specific operational data. The study is intended as a step toward validation of the framework rather than a completed operational deployment.

Did you attend Forum 82? Click the preview below to access the full paper.