← Back to Projects

AI-Engineering Diagrams

Overview

Engineering drawings and diagrams are commonly used across different industries such as oil and gas, petrochemical, construction, transportation and others. Many legacy documents exist only as paper copies, scanned images or in pixelated format. Digitising these drawings is becoming increasingly important to enable conversion of information to a format which is compatible with other digital information relating to the same facility. In this ongong project we are developing an end-to-end solutions for processing and analysing Piping and Instrumentation Diagrams (P&IDs), Construction and electrical diagrams.

Collaborators

  • DNV GL
  • Fieri Analytics Inc
  • Robert Gordon University

Impact

The developed methods significantly improve the accuracy and speed of digitising engineering diagrams compared to traditional manual methods. This enables organisations to efficiently convert legacy documents into digital formats, facilitating better data management, analysis, and decision-making processes. The automation of symbol recognition and diagram interpretation reduces human error and operational costs, leading to enhanced productivity and reliability in engineering workflows. For example, P&ID sheets which contains more than 180 symbols can be processed in under five minutes with an accuracy of over 94%.

Funding

The work was funded by the Oil & Gas Innovation Centre, the Data Lab Innovation Centre, DNV GL and Fieri Analytics Inc 500,000.

Selected Publications

[1] I. Ekeke, C. F. Moreno-García, and E. Elyan, “KD-LSRED: Knowledge distillation for lightweight symbol recognition in engineering diagrams,” in Proc. Int. Conf. Document Analysis and Recognition (ICDAR), 2025, ser. Lecture Notes in Computer Science, vol. 16027. Cham, Switzerland: Springer, 2025, doi: 10.1007/978-3-032-04630-7_28.

[2] L. Jamieson, C. Moreno-García, and E. Elyan, “A review of deep learning methods for digitisation of complex documents and engineering diagrams,” Artificial Intelligence Review, vol. 57, art. no. 136, 2024, doi: 10.1007/s10462-024-10779-2.

[3] E. Elyan, L. Jamieson, and A. A. Gombe, “Deep learning for symbols detection and classification in engineering drawings,” Neural Networks, vol. 129, pp. 91–102, 2020, doi: 10.1016/j.neunet.2020.05.025.

[4] C. Moreno-García, E. Elyan, and C. Jayne, “New trends on digitisation of complex engineering drawings,” Neural Computing and Applications, vol. 31, pp. 1695–1706, 2019, doi: 10.1007/s00521-018-3583-1.

[5] E. Elyan, C. Moreno-García, and C. Jayne, “Symbols classification in engineering drawings,” in International Joint Conference Neural Networks (IJCNN), 2018, doi: 10.1109/IJCNN.2018.8489087.

Team

The project’s team include two Postdoctoral Researchers, two academics, one PhD student, and several domain experts.