Evaluation of a Generative AI Methodology for Technical System Development: Case Study of an Aircraft Environmental Control System

The development of aircraft systems relies on the structured analysis and reuse of extensive technical documentation. Generative Artificial Intelligence (GenAI) offers new opportunities to support this process by synthesizing distributed knowledge into coherent engineering representations. This paper introduces the SAGE methodology (System Architecture through Generative Engineering), which combines retrieval-augmented reasoning with structured synthesis to generate traceable system documentation and consistent conceptual architectures. The proposed workflow integrates four sequential phases: (1) data preprocessing of heterogeneous literature, (2) recursive retrieval to build a traceable network of question-answer pairs, (3) structured knowledge synthesis aligned with typical SE documentation logic, and (4) generative exploration of alternative architectures under modified constraints. The SAGE methodology is applied to the Environmental Control System (ECS), which is a key aircraft subsystem for pressurization, temperature, and cabin air-quality control, which has undergone a well-documented transition from engine bleed-air architectures to electrically driven, bleed-free systems, as implemented in the Boeing 787. By using a literature corpus on engine bleed-air ECS only, the study examines whether a retrieval-augmented generative process can reconstruct the logical and functional structure of the subsystem and, without exposure to electrical ECS sources, propose coherent bleed-free concepts. Results show that GenAI reliably accelerates documentation and knowledge consolidation up to the logical architecture level with full source traceability, while generated architectures should be interpreted as plausible starting points rather than final designs. The case study suggests a pragmatic division of labor: GenAI supports structured documentation and concept-level reasoning, engineers retain responsibility for quantitative validation and certification.