"Leveraging Machine Learning for Defect Detection in Irrigation Concret" by Mohamed Nabawy
 

Document Type

Article

Publication Date

Winter 12-12-2024

Abstract

Abstract:

Defects during the construction phase of projects pose significant challenges, particularly in terms of

safety, cost overruns, delays, and labor inefficiencies. In irrigation concrete canal lining construction

with mega investments, early detection of defects is critical to ensuring quality and project success.

This study explores the application of machine learning (ML) for onsite defect detection, focusing

on the development and deployment of an object detection mobile application specifically tailored

for identifying visible defects in concrete canal lining construction. Leveraging the machine learning

capabilities of Microsoft Azure, a custom object detection model was trained and validated to

recognize defects such as cracks, honeycombing, and steel reinforcement corrosion in canal lining

structures. The results demonstrated that the application effectively identified defects, though

further enhancements to its accuracy and robustness are needed.To evaluate the potential

project-wide impacts of such technology, the study also examined the effects of implementing MLdriven

defect detection on cost, time, and labor productivity through an online survey of experienced

engineers in the Egyptian construction industry. The findings highlighted the promising potential

of machine learning in transforming defect detection processes, improving project performance,

and addressing key construction challenges. However, the study also identified practical barriers,

such as technology adoption and training, that must be addressed for broader implementation.

This research underscores the transformative potential of ML in irrigation infrastructure projects

and its role in advancing construction quality and efficiency.

Share

COinS