Project description
This project is concerned with the recognition of symbols of piping and process equipment together with the instrumentation and control devices that appear on piping and instrumentation diagrams (P&ID). Each item on the P&ID is associated with a pipeline. Piping engineers often receive drawings from clients in PDF format rather than a digital format produced in engineering software packages. This is due to confidentiality reasons and can also be true when dealing with legacy drawings.
From P&IDs, piping engineers are required to generate material take-off (MTOs) documents which will include the list of equipment, their quantities, sizes, etc. The MTO is used to create estimates of costs. To generate MTOs, piping engineers will manually go through every single drawing, identify the symbols, count them, and fill in an MTO spreadsheet. It can take around 36 hours of a highly skilled engineer’s time to generate an MTO for 10 drawings and every project has around 500-1000 drawings. Additionally, the process is prone to mistakes and may require additional checks. Mistakes in quantities, sizes or materials can have huge cost implications.
An AI based approach helps automate the generation of MTOs from P&IDs. It is expected that a human interaction level will still be required for instance in some cases engineers’ expertise may be required to decide on the quantities based on what is presented on drawings. Note that every project has its own legend sheets i.e. symbols differ between projects.

Project Status
To date, with the help of previous interns from Eindhoven and other universities, advanced Deep Learning has been used for text and symbol recognition to identify symbol and text. An application is being built to associate the symbols and their sizes in an MTO.
The remaining scope of work is:
Improve text and symbol recognition models
Extend symbol recognition model to more symbols
Identify lines and their associated line numbers with symbols
Improve application to process 100 P&IDs simultaneously
Add more human interaction features to the application

Desired Skills
Computer vision
Machine learning
Deep learning for object recognition
Python programming language
Text recognition
OpenCV / Template matching
Ability to apply research findings in a timely manner
Proactive in trying and testing new ideas and working with engineers
Any scripts must be written in Python

About McDermott

Please visit https://www.mcdermott.com/ to find out more about McDermott and what we do as a company.

Contact info

Interested students should send their CVs directly to Ya-alimadad, Mina <Mina.alimadad@mcdermott.com>