Machine learning model for corrosion detection and assessment for pad mounted equipment

Corrosion of electrical utility infrastructure represents a major operational challenge that currently relies on labor-intensive manual inspections. This project aims to develop an automated machine learning system to detect and assess corrosion using publicly available street-level imagery, enabling frequent, low-cost monitoring across the distribution network. The proposed methodology will leverage recent advances in object detection and semantic segmentation. Upon successful development, the corrosion detection system will be integrated into EPCOR’s asset management workflow. Automated analysis of public imagery can provide frequent, low-cost monitoring to prioritize field inspections and maintenance activities. Reducing operational costs while extending asset life represents major potential cost savings for the utility industry.

Faculty Supervisor:

Zhigang (Will) Tian

Student:

Partner:

EPCOR Utilities Inc.

Discipline:

Engineering

Sector:

Utilities

University:

University of Alberta

Program: