Probabilistic Optimisation Of Composite Structures: Machine Learning For Design Optimisation

Kwangkyu Alex Yoo, Omar Bacarreza Nogales, M H Ferri Aliabadi

ISBN 10: 1800616848 / ISBN 13: 9781800616844

Editore: World Scientific Europe Ltd, GB, 2025

Riguardo questo articolo

This book introduces an innovative approach to multi-fidelity probabilistic optimisation for aircraft composite structures, addressing the challenge of balancing reliability with computational cost. Probabilistic optimisation pursues statistically reliable and robust solutions by accounting for uncertainties in data, such as material properties and geometry tolerances. Traditional approaches using high-fidelity models, though accurate, are computationally expensive and time-consuming, especially when using complex methods such as Monte Carlo simulations and gradient calculations.For the first time, the proposed multi-fidelity method combines high- and low-fidelity models, enabling high-fidelity models to focus on specific areas of the design space, while low-fidelity models explore the entire space. Machine learning technologies, such as artificial neural networks and nonlinear autoregressive Gaussian processes, fill information gaps between different fidelity models, enhancing model accuracy. The multi-fidelity probabilistic optimisation framework is demonstrated through the reliability-based and robust design problems of aircraft composite structures under a thermo-mechanical environment, showing acceptable accuracy and reductions in computational time.

Codice articolo LU-9781800616844

Dati bibliografici

• Titolo: Probabilistic Optimisation Of Composite Structures: Machine Learning For Design Optimisation
• Editore: World Scientific Europe Ltd, GB
• Data di pubblicazione: 2025
• Lingua: Inglese
• ISBN 10: 1800616848
• ISBN 13: 9781800616844
• Rilegatura: Hardback
• Condizione: New

Descrizione articolo

Riassunto

This book introduces an innovative approach to multi-fidelity probabilistic optimisation for aircraft composite structures, addressing the challenge of balancing reliability with computational cost. Probabilistic optimisation pursues statistically reliable and robust solutions by accounting for uncertainties in data, such as material properties and geometry tolerances. Traditional approaches using high-fidelity models, though accurate, are computationally expensive and time-consuming, especially when using complex methods such as Monte Carlo simulations and gradient calculations. For the first time, the proposed multi-fidelity method combines high- and low-fidelity models, enabling high-fidelity models to focus on specific areas of the design space, while low-fidelity models explore the entire space. Machine learning technologies, such as artificial neural networks and nonlinear autoregressive Gaussian processes, fill information gaps between different fidelity models, enhancing model accuracy. The multi-fidelity probabilistic optimisation framework is demonstrated through the reliability-based and robust design problems of aircraft composite structures under a thermo-mechanical environment, showing acceptable accuracy and reductions in computational time.