Automatic extraction of cable connection information from 2D drawings for electrical outfittings design in shipyards

This study proposes to automate the analysis of wiring diagrams to generate cable lists by using machine learning for text classification and pre-trained Deep Neural Network (DNN)-based image classification to detect cable routes. In shipyards, many drawings are produced for each ship, and analyzing these drawings, especially wiring diagrams, to generate cable lists for the Bill of Materials (BOM) can be a time-consuming and error-prone task. This process is performed manually by reading the cable routes and entering the information into a spreadsheet. To address these challenges, this study aims to automate the information extraction from wiring diagrams. The process involves extracting text from the PDF document and classifying it using machine learning, followed by using DNN-based image classification to trace cable routes and identify the relevant annotations. The developed algorithm was tested on three PDF wiring diagram samples and achieved an average accuracy of about 90%, confirming its effectiveness.