Graph-Based Patent Mining for Mechanical Designs

Patents represent a rich source of design innovations, prompting the application of different technologies. Machine learning, text and data mining, similarity scoring, and evolving ontology methods are among the various approaches applied in the literature. This study introduces a schema-free graph data modelling of Functional Analysis Diagrams (FAD) extracted from Patents and their associated Auto-CAD models. It aims to represent mechanical design patents semantically. The schema-free graph model allows for a flexible evolving ontology of known geometries, interactions, and functions. This evolution enables comprehensive queries and ensures efficient storage that is compatible with visualisation libraries. The developed PatMine SolidWorks Add-in ©streamlines CAD design comparison with stored patents' FAD annotations by highlighting shared concepts. It also enables comprehensive full-text and semantic search queries, similarity scoring, and efficient storage compatible with visualisation libraries. The extraction of functional analysis and geometric features empowers the patent database with capabilities for seamless integration with graph analytics and machine-learning approaches for future endeavours.