Dynamic physics-based digital twin for supervisory, condition monitoring and fault diagnosis of industrial turboshaft engines

This study presents a digital-twin (DT) framework for real-time condition monitoring and fault diagnosis of industrial turboshaft engines. The proposed architecture employs physics-based semi-empirical models as the core of the twin, which are calibrated using operational data collected under both nominal and off-design conditions. Given that gas turbines are prone to diverse faults with varying onset times, magnitudes, and signatures, a time-windowed residual embedding (TWRE) method is introduced to enhance the temporal representation of diagnostic features. In addition, a data-driven dynamic adaptive thresholding scheme is applied to improve robustness against environmental variations and uncertainties. By providing a comprehensive fault table, feature spaces are systematically linked to distinct fault scenarios using pattern recognition classifiers. Finally, a decision fusion mechanism is implemented to integrate diagnostic outputs based on fault types and their severities. To assess the performance of the proposed system, four sample faults are considered for simulation studies. The results indicate that the system reliably tracks fault progression and accurately detects both evident and latent anomalies, providing clear fault timelines for early alarm generation. The effectiveness of integrating semi-empirical modelling and advanced signal processing in a digital twin framework is confirmed for real-time fault monitoring and diagnosis of industrial gas turbines.