Graph models for engineering design: Model encoding, and fidelity evaluation based on dataset and other sources of knowledge

Abstract Automatically extracting knowledge from small datasets with a valid causal ordering is challenge for current state-of-the-art methods in machine learning. Extracting other type of important but challenging multiple engineering fields where data are scarce and difficult to collect. This research aims address this problem by presenting learning-based modeling framework leveraging the available fundamental units variables recorded samples, develop parsimonious, explainable, graph-based simulation models during early design stages. The developed approach exemplified using an case study spherical body moving fluid. For system interest, two types intricated generated (1) automated selection (2) combining extraction supplementary about functions dimensional homogeneity associated system. effect design, data, model, specifications on model fidelity investigated. discusses interrelationships between levels, variables, functions, knowledge. contributes development measurement theory premises standardized, transforming into measurable level fidelities produced models. shows that structured building focus can support reasoning decision making example analysis conceptual (DACM) framework.