Compliant Mechanism Design Using Multi-Agent Large Language Models

The transition from rigid-body to compliant mechanisms offers well-recognized advantages in mechanical design, including reduced part count, improved energy efficiency, and expanded functional integration. However, existing catalogue-based and artificial intelligence–assisted design approaches often rely on abstract functional archetypes, limiting practical applicability and hindering identification of integrated, application-specific compliant solutions. This study presents a methodological framework for extracting compliant mechanism solutions from patent literature based on a structured, multi-agent pipeline employing Large Language Models (LLMs) within a Retrieval-Augmented Generation (RAG) architecture. The proposed approach integrates product decomposition, lexical and technological expansion, patent retrieval, compatibility assessment, and taxonomy-grounded classification using compliant mechanism archetypes. The framework is demonstrated through a case study involving 40 patents related to pedal-actuated waste bin mechanisms, enabling qualitative evaluation against expert judgment. The results indicate consistent identification of compliant alternatives to conventional rigid mechanisms, ranging from localized compliant substitutions (e.g., living hinges and leaf springs) to more integrated compliant transmission architectures. The patent-derived solutions provide application-specific structural embodiments, including geometric details and implementation information suitable for design reuse. Rather than proposing a finalized design tool, the study clarifies the architectural and methodological requirements for leveraging patent data in compliant mechanism design. By grounding design exploration in validated prior art and taxonomy-based reasoning, the proposed framework supports systematic discovery of contextually relevant compliant mechanisms while reducing abstraction and cognitive load in early-stage design.