Complex pipeline analysis using sparse synthetic data
Academic Institution: University of Edinburgh
Industry Partner: ShapeSpace Ltd
Academic Supervisor: Dr Nan Yut
PhD Student: Toa Pecur
Summary
Three dimensional (3D) measurements in real life are complex. Currently, there is a portion of manual work being undertaken in the manufacturing scene when carrying out compliance and quality checks for large scale environments. Digitising some of these checks would streamline decision making processes (a non-exhaustive list: remanufacturing, maintenance, modification, decommission, documentation) using artificial intelligence (AI).
Commonplace methods for measuring large assemblies that are prone to change have revolved around manual measurements. There is an increasing need for utilising the data in a digital format by capturing a digital representation of the physical model. Having a digital model expedites compliance checking for potential conflicting parts earlier in the manufacturing processes found in large scale metrology. Oftentimes, this would entail a complete point cloud scan of the object, or in the case of partial scans a varied amount of manual operator steps to deduce the characteristics captured by a scanner.
This work aims to highlight areas where large rework could be avoided, in part by the detection of potential clashes of components early in the pipeline installation process. With the help of an extracted model in the form of a point cloud generated from a scanned physical model and a digital model representing the nominal geometry, an operator can be made visually aware of potential deviations and component clashes during a pipeline assembly process.
Currently, the field of as-is pipeline comparison to the computer aided design (CAD) and building information modelling (BIM) assemblies during installation is underdeveloped. The majority of assembly deviations are not inspected preventing minor rework. This lack of pipeline comparison accumulates into major rework at later stages in the assembly process which necessitates more costly resources when correcting them.
Key Results/Outcomes
Framework for tackling 3D vision analysis enabling working within compliance when using high accuracy point clouds.
This involved using deep learning for object classification of point clouds, from exclusively synthetically created data.
The final results provided the ability for the end user to better understand if deviations were problematic earlier in the manufacturing process (allowing for correction before major rework would be needed).
Publications
Toa Pecur, Nan Yu, Andrew Sherlock, and Frank Mill. 2022. “Automation development using a digital approach in prototype pipework.” In Proceedings of euspen’s 22nd International Conference & Exhibition, June, Switzerland, Geneva.
Toa Pečur, Nan Yu, Andrew Sherlock, Andrew Brown, and Frank Mill. 2023. “Robust coordinate system alignment using high density point clouds from laser line probe.” In Laser Metrology and Machine Performance XV, March, 103–113, United Kingdom, Edinburgh.
Toa Pečur, Frédéric Bosché, Gabrielis Cerniauskas, Frank Mill, Andrew Sherlock, and Nan Yu, 2024. “Prototype pipeline modelling using interval scanning point cloud”, Advances in Manufacturing (2024). https://doi.org/10.1007/s40436-024-00515-y
Contact information
Dr Nan Yu
Senior Lecturer/Associate Professor, University of Edinburgh
Toa Pecur
toa.pecur@ed.ac.uk