Product evolution model (PEM): a functional semantic knowledge-driven framework for product technology landscape construction

Purpose The construction of a product technology landscape remains heavily reliant on patent-based methods despite its recognized value in supporting the systematic product planning and technology opportunity identification. These patented approaches, however, are typically constrained to document-level operations such as retrieval, clustering and keyword analysis, creating a critical gap in their capacity to support the generation of component-level solutions. To address this gap, this study aims to propose a product evolution model (PEM), a functional semantic knowledge-driven framework that integrates patent-derived component knowledge, function–object–property (FOP) representation, TRIZ-based evolution trend label mapping, multi-objective optimization and TOPSIS-based decision ranking. Design/methodology/approach First, patent claims are transformed into product-component and component-FOP knowledge using the structured information extraction. Moreover, functional community detection and topic modeling are then used to identify representative components, while evolution trend labels are mapped to candidate components and incorporated into a component-evolution trend association matrix. The PEM formulates product technology landscape construction as a three-objective optimization problem that maximizes implementation effect, maximizes functional coverage and minimizes component count. Findings Using a new energy vehicle case study, the feasibility of the framework has been further demonstrated: from 873 initial component candidates, 654 valid components are retained after the semantic filtering and domain inspection, and 22 representative components are identified through the Louvain community detection and latent Dirichlet allocation topic modeling. Under the TOPSIS preference setting of 0.5 for the implementation effect, 0.2 for functional coverage and 0.3 for component count, the PEM selects a 19-component technology landscape solution with an implementation effect of 17.4375, a functional coverage of 0.9528 and a TOPSIS score of 0.6288. Originality/value The case study demonstrates the feasibility of PEM for supporting component-level technology landscape construction by connecting patent-derived functional semantics with evolution-oriented optimization and preference-sensitive decision ranking.