The automatic generation of digital twins of industrial processes requires the integration of several sources of information. If the twin is expected to accurately capture thermo-hydraulic phenomena, dimensions of tanks and other process components as well as detailed pipe routing information is relevant. Such information is not comprehensively captured in PIDs (Piping Instrumentation Diagrams), but it is available from 3D CAD models. However, information about control loops is not available from 3D CAD models, but is available from PIDs. Previous research has demonstrated the extraction of such information from machine-readable PIDs and 3D CAD models and converting this information to graphs. Further research is expected on applying graph matching methods for integrating these separate graphs to a common graph-based data structure that captures all of the desired information. This common model could support further work to develop digital twins. A major obstacle to this is that the graphs that have currently been generated from PIDs and 3D CAD models are at very different abstraction levels, so graph matching methods are not feasible. This article address this obstacle by building on previous work, in which graphs were generated from PIDs and 3D CAD models. The contribution of this paper is several novel algorithms for preprocessing a 3D CAD generated graph, until it is at the same level of abstraction as a PID generated graph of the same industrial process. The algorithms are demonstrated in the context of a laboratory process. © 2020 IEEE.