Leveraging Big Data and Deep Learning for the Condition Assessment of Wastewater Pipelines Webcast
The webcast was held in partnership with the Water Environment Federation (WEF). Municipalities routinely inspect the condition of sewer pipelines to determine the optimal approach and timing of maintenance operations. For many municipalities, closed-circuit television (CCTV) is the primary method of inspecting sewers. This manual process of identifying defects can result in errors and inconsistencies, since it relies on a subjective interpretation of images. Furthermore, this process tends to be slow and labor-intensive since it frequently involves rewatching the video recordings. Automated defect identification could improve the speed, consistency, and accuracy of CCTV inspections. This webcast focused on Leveraging Big Data and Deep Learning for Economical Condition Assessment of Wastewater Pipelines (project 4902). The presentation covered the following:
How deep learning and data mining can improve the accuracy and speed of visual inspections of sewer pipe condition.
How an automated system was developed as proof of concept for the use of deep learning in sewer inspection coding.
How this automated system can be applied to 8-inch and 10-inch diameter vitrified clay pipes (VCP), in the context of detecting fissures, root intrusions, and lateral connections.
A discussion of Defect Cluster Analysis (DCA), a methodology that uses big datasets of sewer inspection reports to identify pipe segments that contain clusters closely spaced defects.
Dr. Srinath Shiv Kumar, AI Engineer, Sewer AI
Dr. Dulcy M. Abraham, Professor, Lyles School of Civil Engineering, Purdue University
Walter Graf, Research Program Manager, The Water Research Foundation