AI-Driven Inverse Constrained Design for Turbine Optimization

AI-Driven Inverse Constrained Design for Turbine Optimization

Francesco PORTA, Sergio LAVAGNOLI

Abstract. This work presents an AI-driven framework for the inverse and constrained design of turbine blade sections, aimed at accelerating early-stage aerodynamic optimization. The methodology integrates automated database generation, surrogate modeling, and constrained multi-objective optimization within a unified process. A large set of high-Reynolds-number 2D uncooled blades is automatically generated and simulated through an in-house workflow, providing physically consistent data for model training. Kolmogorov–Arnold Networks are employed as surrogate models to establish a direct mapping between the aerodynamic duty parameters and the corresponding blade geometries and performance indicators. The surrogates are tuned through the Optuna optimization framework to ensure accuracy and robustness over the design space. Once trained, the models are coupled with a constrained optimization environment to identify the Pareto front between kinetic-energy losses and aero- dynamic loading while satisfying geometric feasibility requirements. The resulting framework enables rapid and physically informed exploration of turbine design alternatives without iterative CFD computations, offering a general and reproducible methodology for data-driven aerodynamic design.

Keywords
Turbomachinery, Multi-Objective Optimization, Kolmogorov-Arnold Networks

Published online 7/20/2026, 6 pages
Copyright © 2026 by the author(s)
Published under license by Materials Research Forum LLC., Millersville PA, USA

Citation: Francesco PORTA, Sergio LAVAGNOLI, AI-Driven Inverse Constrained Design for Turbine Optimization, Materials Research Proceedings, Vol. 69, pp 724-729, 2026

DOI: https://doi.org/10.21741/9781644904251-128

The article was published as article 128 of the book CEAS – AIDAA Conference 2025

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 license. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

References
[1] Christopher J. Clark, ”A step towards an intelligent aerodynamic design process”. ASME Turbo Expo 2019: Turbomachinery Technical Conference and Exposition, 2019.
[2] Bellucci, J.; Granata, A. A.; Silei, M.; Giovannini, M.; Spano, E.; Notaristefano, A.; Lengani, D. A Real Time AI-Based Strategy for the Design of a Low-Pressure Turbine Profile. In Volume 12D: Turbomachinery – Multidisciplinary Design Approaches, Optimization, and Uncertainty Quantification; Radial Turbomachinery Aerodynamics; Un-steady Flows in Turbomachinery; American Society of Mechanical Engineers: London, United Kingdom, 2024; p V12DT34A026. https://doi.org/10.1115/GT2024-128809.
[3] Petronio, D.; Paliotta, P.; Lengani, D.; Simoni, D. A ML Strategy for the Identification of Optimal LPT Design Region and Related Blade Shape. Aerospace Science and Technology 2024, 148, 109118. https://doi.org/10.1016/j.ast.2024.109118.
[4] Chen L, Zou L, Fan C et al (2020) Feature weighting network for aircraft engine defect detection. Int J Wavelets Multiresolut Inf Process 18(03):2050012. https://doi.org/10.1142/S0219691320500125
[5] Kulfan, B. M., 2008. “Universal parametric geometry representation method”. Journal of Aircraft, 45(1), pp. 142–158.
[6] Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Soljacic, Thomas Y. Hou, Max Tegmark. “KAN: Kolmogorov–Arnold Networks”. https://arxiv.org/pdf/2404.19756, 2024.
[7] Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. Optuna: A Next-generation Hyperparameter Optimization Framework. In KDD.
[8] Drela, M., and Youngren, H., 1998, “A User’s Guide to MISES 2.53, ”Massachusetts Institute of Technology, Cambridge, MA.