.png){kind=icon}
.png){kind=icon}
Return
Re-instrumenting quality function deployment for generative engineering design: a small language model-based adaptive QFD framework
.png){kind=icon}
PRE
AI
DOI：10.1080/09544828.2026.2723799.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
While traditional Quality Function Deployment (QFD) remains effective at translating stakeholder requirements into engineering characteristics, its static formulation is poorly suited to contemporary generative design, where rapid synthesis can bypass requirement traceability and constraint governance. This paper proposes an Adaptive QFD (AQFD) framework for generative engineering design in which a domain-specialised Small Language Model (SLM) re-instruments QFD as an active governance mechanism. The framework comprises three phases: (1) Insight, where the SLM transforms stakeholder inputs, standards, and design documentation into a requirement-constraint knowledge graph (RCKG) for requirement extraction, prioritisation, and conflict identification; (2) Integration, where an Adaptive House of Quality(AHoQ) derives traceable, constraint-aware engineering targets; and (3) Ideation, where these targets become structured generative conditions guiding concept synthesis within a governed design space. When validated requirement evidence, priorities, regulations, or constraints change, affected RCKG/AHoQ elements and downstream target bundle are recomputed. In an aerospace monitor-console redesign case study, the proposed approach reduced human-active iteration time and increased concept-level constraint compliance and expert-rated engineering plausibility relative to a prompt-only generative baseline. The findings suggest that re-instrumented QFD can integrate concept generation into requirement-driven engineering workflows while preserving traceability and explicit concept-level validation boundaries.
Keywords:
Small language model
generative AI
Quality Function Deployment (QFD)
engineering design
Knowledge Graph
Journal
J
IF:
3.4
Papers:
160
Citations:
0