CFD-Informed Methods for Condition Monitoring and Diagnostics of Turbomachinery

Turbomachinery plays a central role in industrial energy conversion systems, where reliable operation and early detection of performance degradation are essential for ensuring efficiency, safety and reduced operational costs. This thesis develops methods informed by Computational Fluid Dynamics (CFD) for condition monitoring and diagnostics of turbomachines operating under realistic industrial constraints. The research was conducted within an industrial PhD programme in collaboration with the Istituto Italiano della Saldatura (IIS) and Cranfield University. The research follows a progressive methodological development combining CFD, signal analysis and data-driven methods. Initially, diagnostic techniques are developed for an industrial radial compressor using time-resolved CFD simulations. In this context, aerodynamic instabilities, such as surge, are considered from a maintenance viewpoint as fault mechanisms that can cause significant performance deterioration. Machine learning models are developed to reconstruct internal flow behaviour from limited pressure measurements and to classify compressor states approaching instability. Expanding on these results, a second approach utilises approximate entropy (ApEn) as a metric to identify the monitoring locations most sensitive to surge onset and provides a quantitative basis for sensor placement optimisation. The expertise developed in this controlled numerical environment is subsequently transferred to wind turbines, a complementary class of turbomachinery for which experimental validation was possible. Digital Twin (DT) architectures are developed for a small-scale horizontal-axis wind turbine, integrating CFD-derived performance models with operational measurements. These DTs enable the detection of aerodynamic performance deviations, including yaw misalignment caused by sensor faults and blade surface degradation, while providing maintenance-oriented feedback to support operational decision-making. Overall, the thesis demonstrates how CFD-informed diagnostic methodologies can evolve into DT frameworks capable of supporting condition monitoring and maintenance strategies for turbomachinery operating in complex industrial environments.