Giuseppe Giordano
Multi-Agent Generative AI for Sustainable Design for Manufacturing.
Rel. Alessandro Simeone, Jian Zhang, Domenico Augusto Francesco Maisano. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Della Produzione Industriale E Dell'Innovazione Tecnologica, 2026
Abstract
Manufacturability is still assessed after the design is fixed, and the trade-off among environmental impact, cost and the physical strain borne by assembly operators is resolved late, implicitly and by separate tools. Generative artificial intelligence promises to close that gap, but its use in engineering is constrained by a specific difficulty: a language model can produce a plausible number without being able to say where the number comes from. This thesis presents a five-layer multi-agent framework that validates the manufacturability of a mechanical component and, in the same act, determines how that component should be produced. Constraints are retrieved from a corpus of twenty standards and technical documents and become numeric limits that carry their own provenance; a structured design specification is parsed into a log of critical features, each classified against a cited norm; three specialised agents - sustainability, economic and human-centric - negotiate a balanced configuration over fourteen indicators computed by formulas anchored to those sources; a discrete-event simulation of the resulting production line verifies that the negotiated configuration can in fact be built; and the verdict is returned to a human operator with the epistemic status of every figure declared, distinguishing what a standard prescribes from what an industry benchmark suggests and from what the corpus cannot ground at all.
The production line is an output rather than an input: the forming process follows from the material family, the automation level is obtained by inverting the ergonomic models until the lifting index returns within its limit, and the number of workstations follows from the fastener count and the takt time
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Informazioni aggiuntive
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
