Brisolin, Jacob (2021) Application of Automated Road Fault Detection for Improved Asset Maintenance Planning. [USQ Project]
|
Text (Project)
BRISOLIN Jacob dissertation_redacted.pdf Download (8MB) | Preview |
Abstract
All assets are in a constant state of deterioration, to combat this, a program of inspection and renewal is employed. This program is informed by a series of network condition surveys performed by asset inspectors in order to strive to achieve the strategic levels of service. Currently, these systems are highly reactive simply due to the sheer scale of a road asset network, then when in conjunction with making holistic decisions, this inspection and decision-making requirement increases exponentially. This research project aims to investigate the simplification of the assessment process through the application of automated fault detection and how it can generate condition reports through network inspections. The goal is for these inputs to be generated by not only staff but also submitted by members of the public using smartphones or tablets; this means that faults need to be recognizable from any number of angles and distances. This is achieved by collecting a database of eight hundred and sixty-three images that contain a combination of road faults and undamaged pavement sections, as well as a plethora of background information visible from the road reserve. This database is turned into an alternate database of classified tiled images comprising twelve thousand six hundred images with one thousand two hundred road pavement cracks, with the remainder containing images of a large array of common background items visible from within a residential road reserve. These images were applied to a deep learning program for training through image classification using transfer learning. The result is that this system is able to generate an accuracy of ninety-five percent, which is a successful proof of concept.
From this success, an integrated model of artificial intelligence and an active asset management team is developed, showing the extremely effective relationship that would result, enabling the delivery of more and better outcomes.
|
Statistics for this ePrint Item |
| Item Type: | USQ Project |
|---|---|
| Item Status: | Live Archive |
| Additional Information: | File reproduced in accordance with the copyright policy of the publisher/author. |
| Faculty/School / Institute/Centre: | Historic - Faculty of Health, Engineering and Sciences - School of Civil Engineering and Surveying (1 Jul 2013 - 31 Dec 2021) |
| Supervisors: | Nguyen, Andy |
| Qualification: | Bachelor of Engineering (Honours, Civil) |
| Date Deposited: | 12 Jul 2023 02:35 |
| Last Modified: | 12 Jul 2023 02:35 |
| Uncontrolled Keywords: | asset, road, deep learning, network condition survey, road fault, artificial intelligence |
| URI: | https://sear.unisq.edu.au/id/eprint/52067 |
Actions (login required)
 |
Archive Repository Staff Only |
