LLM Agent-Based QFD-TRIZ Integration for Contradiction Resolution Through Compliant Mechanisms

The increasing complexity of product development requires methodologies capable of managing heterogeneous and conflicting requirements while supporting innovation-oriented design. Although Quality Function Deployment (QFD) and TRIZ are widely used to translate customer needs into engineering solutions, their integration still provides limited support for connecting abstract contradictions to physically implementable concepts. This study proposes an integrated framework combining QFD, TRIZ, compliant mechanism-based design and large language models (LLMs) within a unified workflow. A multi-level House of Quality is used to identify critical technical parameters and localize contradictions within the product architecture, while a structured prompting strategy guides the LLM toward compliant mechanism-based solution concepts. The methodology is demonstrated through a case study on an adaptive computer mouse characterized by a conflict between shell geometric variability and structural stiffness. The framework generated concept-level proposals including living-hinge shells, bistable structures, anisotropic kirigami geometries and compliant lateral zones combined with rigid load-bearing regions. Results indicate that the AI effectively supports requirement formalization, product decomposition and conceptual exploration, while still requiring expert supervision for embodiment validation and interface consistency.