Digitization of engineering diagrams

We develop AI methods that extract the topology and data of flowsheets and P&IDs from images and PDF files and store them in machine-readable graph formats. This research is the basis of the Digitization Companion (DigiCo).

Challenge

Most engineering diagrams in the process industry, such as process flow diagrams (PFDs) and Piping and Instrumentation Diagrams (P&IDs), exist only as images or PDF files. Their information cannot be searched, checked, or used by software without manual re-drawing, which is slow and error-prone. Machine-readable diagrams are a prerequisite for most applications of AI in process engineering.

Our approach

The digitization of flowsheets combines several object detection algorithms with a pathway exploration algorithm. In the object detection step, machine learning models identify the position and type of unit operations and other symbols. In the pathway exploration step, the connectivity between them is explored. Text and tables are recognized and digitized as well. We have built a large dataset of labeled flowsheets and P&IDs from various sources and train deep learning models on it.

The result is a graph of the diagram that can be exported to standard formats such as DEXPI. Such machine-readable Smart P&IDs are the basis for our work on ChatP&ID, autocorrection, and AI-assisted HAZOP.

From research to practice: DigiCo

This research started more than five years ago and is now growing into the Digitization Companion (DigiCo), prototype software for Smart P&IDs. A team of software engineers is turning the research into industry-grade software. A first beta release is planned for January 2027. You can sign up at digitization-companion.com.

Recognition

  • The work on flowsheet digitization received the IChemE Senior Moulton Medal 2024 (news).

Key publications

  • Theisen, M. F., Flores, K. N., Schulze Balhorn, L., & Schweidtmann, A. M. (2023). Digitization of chemical process flow diagrams using deep convolutional neural networks. Digital Chemical Engineering, 6, 100072. doi:10.1016/j.dche.2022.100072 · Details
  • Schulze Balhorn, L., Gao, Q., & Schweidtmann, A. M. (2022). Flowsheet recognition using deep convolutional neural networks. Computer Aided Chemical Engineering, 49, 1567–1572. doi:10.1016/B978-0-323-85159-6.50261-X · Details
← All projects