A hybrid knowledge based and data based approach for efficient clinch joint design

Abstract The design of clinch joints is a cost- and time-intensive iterative process due to the complex relationships between tool and process parameters and the resulting joint properties. To address this, this contribution proposes a novel hybrid workflow that combines knowledge- and data-based approaches. Relationships are categorized based on their knowledge quality and the need for a quantitative prediction. Well-established, generalizable relationships are formalized in an ontology as design guidelines (no quantification required) or SWRL rules (quantification required) to model expert knowledge. In contrast, hard-to-formalize or not-fully-understood relationships are treated with regression models for continuous or classification models for binary criteria. These approaches are combined in a generic user interface (GUI), where the ontology can be accessed using predefined SPARQL queries to select and adapt parameters using expert knowledge. These parameters are then used as input for the metamodels. The developed workflow is evaluated on two exemplary joining tasks to illustrate, how designers can retrieve similar prior joints, adapt parameters using the encoded design rules and predict resulting joint properties under varying process conditions. In summary, the combination of ontology and metamodels facilitates the transition of trial and error into an efficient, documentable design process.