Repository logo
 

Leak Detection for Waterproof Enclosures Using Schlieren Optics

Files

Thesis(9.5 MB)

Date

2018

Authors

Supervisor

Gschwendtner, Michael
Lowe, Andrew

Item type

Thesis

Degree name

Master of Engineering

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

In the search for a method of detecting leak location without danger of damaging the product tested or harming the user, Navico approached Auckland University of Technology (AUT) with the proposition of developing a visual method to detect leaks. Schlieren optics showed promise when a tracer gas with a different refractive index is used. Helium was chosen for its inertness as well as small molecule size for leak penetration. A double pass Schlieren optical assembly was built, development of the optical assembly iteratively increased the sensitivity to small leaks until a leak that would pass Navico’s production test criteria could be seen. Once small leaks could be seen the next stage in the development of the Schlieren optical assembly was increasing the range of leak sizes that could be viewed, with the optical assembly tuned to smaller leaks, it caused larger leaks to be less visible. Automation of the knife edge adjustment allowed for photos with different cut-off amounts and orientations to be captured. These photos were then combined on a computer giving an image with sensitivity to a full range of leak sizes. Full automation was achieved by computer control of the camera, knife edge and a rotating product table that allowed multiple view orientations. Finally, artificial intelligence was used to increase the ease of detection by unexperienced users, region detection neural networks and Semantic Segmentation networks showed good success at identifying and highlighting leaks from video taken from the Schlieren optical assembly.

Description

Keywords

Schlieren Optics, Leak Detection, Artificial Intelligence, Image Processing

Source

DOI

Publisher's version

Rights statement

Permanent link

https://hdl.handle.net/10292/11881

Collections

Masters Theses