Automated Detection of Sewer Pipe Structural Defects Using Machine Learning

In its 2014 Sewer Master Plan Supplemental Study (Study), the Long Beach Water Department (LBWD) discovered inconsistent structural condition scoring between closed circuit televisions (CCTV) Operators who would inspect the same sewer pipeline. Repetitive workload and the manual processing of pipe condition assessments, introduces multiple errors in scoring such as defects being overlooked and miscoded. The discrepancies ultimately affect LBWD’s goal of developing a comprehensive sewer Capital Improvement Projects (CIP) to efficiently plan and cost our projects. LBWD has teamed with California State University, Long Beach (CSULB) to develop patent-pending software to assist Operators in their everyday practice of pipe inspections. The software utilizes machine learning to identify defects within a pipe. This project is an example of a partnership between local resources working together to solve problems for its community.