Physics-Informed Neural Networks in Aerospace: A Structured Taxonomy with Literature Review

Purpose. This study aims to develop a structured four-tier taxonomy that systematically organizes aerospace engineering tasks suitable for the application of Physics-Informed Neural Networks (PINNs), while validating this classification through a literature review and identifying opportunities for future research. Design / Method / Approach. The methodology involves grouping tasks into four distinct tiers—Physical Modeling, Dynamic Analysis, Functional Assessment, and System-Level Assessment—based on their physical, operational, and systemic characteristics. This framework is subsequently populated with real-world examples derived from the analysis of 145 peer-reviewed studies. Findings. The reviewed literature confirms a balanced distribution of PINNs applications across all tiers. Contrary to initial assumptions, studies were identified even in areas previously presumed underrepresented, such as acoustic modeling, optical simulations, and environmental impact assessment. This outcome reveals the broader applicability of PINNs and calls for a reassessment of current assumptions regarding underexplored domains. Theoretical Implications. The proposed taxonomy offers a coherent framework for structuring interdisciplinary PINNs applications by integrating physics-based modeling with machine learning across aerospace engineering contexts. Practical Implications. It provides engineers and researchers with a practical roadmap for selecting PINNs methods tailored to specific problem types, potentially improving computational efficiency and enhancing predictive accuracy in aerospace design and analysis. Originality / Value. The study’s originality lies in its empirically validated, four-tier taxonomy that synthesizes the fragmented body of literature on PINNs in aerospace, offering a unified perspective for researchers and practitioners. Research Limitations / Future Research. While the taxonomy covers a wide range of existing applications, future studies should consider extending it with new tiers—particularly related to manufacturing-aware modeling—and pursue methodological standardization to ensure reproducibility and scalability. Article Type. Review.