Intelligent Product Conceptual Design: A Comparative Review of Paradigms, Challenges, and Future Trends

<title>Abstract</title> <p>Intelligent product conceptual design (IPCD) is evolving from predominantly symbolic reasoning systems toward data-driven, generative, and hybrid AI architectures, yet the field lacks a coherent understanding of how these paradigms redistribute design knowledge, reasoning, automation, and human agency. This review examines 247 studies published between 1990 and 2026 using a PRISMA-informed selection protocol and a five-dimensional comparative framework covering knowledge representation, reasoning mechanism, automation level, scalability, and human–AI interaction. The findings reveal not a simple replacement of earlier paradigms by machine learning, but persistent structural trade-offs. Symbolic methods provide explicit and interpretable reasoning but scale poorly; evolutionary approaches expand design-space exploration but remain dependent on representation and evaluation; and ML/DL methods offer strong scalability and generative capacity but lack reliable explainability and physical-feasibility guarantees. Hybrid architectures are increasingly prominent, although current evidence remains insufficient to establish their general superiority over well-designed single-paradigm baselines. Only 5.7% of the reviewed studies include human-designer baseline comparisons, while reproducibility and CAD/PLM integration remain limited. The review argues that IPCD is moving toward distributed design intelligence in which human judgment, symbolic knowledge, generative search, and physically grounded reasoning perform complementary roles. A research agenda is proposed for open FBS-oriented datasets, feasibility-aware generative methods, standardized human–AI benchmarks, reproducible experimentation, and industrial workflow integration.</p>