Using artificial intelligence to support the drawing of piping and instrumentation diagrams using DEXPI standard
Jonas Oeing, Wolfgang Welscher, Niclas Krink, Lars Jansen, Fabian Henke, Norbert Kockmann
Abstract
The design and engineering of piping and instrumentation diagrams (P&ID) is a very time-consuming and labor-intensive process. Although P&IDs show common patterns that could be reused during development, the drawing is usually created manually and built up from scratch for each process. The aim of this paper is to recognize these patterns with the help of artificial intelligence (AI) and to make them available for the development and the drawing process of P&IDs. In order to achieve this, P&ID data is made accessible for AI applications through the DEXPI format, which is a machine-readable, manufacturer-independent exchange standard for P&IDs. It is demonstrated how deep learning models trained with DEXPI P&ID data can support the engineering as well as drawing of P&IDs and therefore decrease labor time and costs. This is achieved by assisted prediction of equipment in P&IDs based on recurrent neural networks as well as consistency checks based on graph neural networks.