Deep learning techniques for the design and optimization of centrifugal compressor families

Driven by global carbon neutrality goals, the rapid development of centrifugal compressor families for applications such as LNG has become increasingly critical. To overcome the computational challenges of high-fidelity 3D aerodynamic simulations across both design and off-design conditions, this thesis leverages artificial intelligence and surrogate modeling to accelerate multi-objective optimization. Ultimately, it presents a validated methodology for the fast, robust, and automated generation of transonic centrifugal impellers using experimental data from Baker Hughes Florence.