Abstract

This paper provides a brief review of artificial intelligence (AI) methods for sustainable aerospace systems, focusing on predictive and generative models that enable innovation in Industry 4.0 and Industry 5.0. Predictive AI models are analyzed in terms of their capacity to estimate remaining useful life (RUL), optimize maintenance planning, and enhance safety management of critical aerospace components, such as turbofan engines and aircraft bearings. Generative models, including GANs, VAEs, and diffusion-based approaches, are examined as enablers of aerodynamic design optimization, structural reliability assessment, and digital twin integration, significantly reducing costs of physical testing and accelerating certification. The review highlights how physics-informed neural networks (PINNs) integrate physical laws into machine learning, ensuring reliability and interpretability in real-world aerospace applications. Key challenges such as data scarcity, algorithmic transparency, and compliance with certification standards are addressed, alongside perspectives for hybrid AI approaches combining symbolic reasoning, neural networks, and digital twins. The findings demonstrate that the integration of predictive and generative AI methods not only supports sustainable development goals by reducing emissions, resource consumption, and downtime, but also strengthens the competitiveness of the aerospace sector in the European digital transition.