Conference Agenda

Siphons are one of the most difficult sewer pipelines to maintain as they are designed to be continously full of water, often have little redundancy, and typically have very high flow rates. They are also located deep underground as they are designed to carry wastewater under roadways, channels, and water bodies. These factors make maintenance work, such as cleaning and inspections, very difficult and often costly. The construction and operation of inverted siphons will be explained, then an in-depth discussion on methods that can be used to dewater, bypass, clean, and inspect sanitary sewer siphons will be presented. Proactive maintenance best practices will be discussed, along with design ideas to improve the future construction of sanitary sewer siphons.

Learning Objectives:

1. Explain the construction and typical operation of sewer siphons

2. Explain cleaning methods based on siphon configuration

3. Explain inspection methods for sewer siphons

4. Present design ideas for building better siphons.

Condition assessment of sewer systems is critical to maintaining system structural integrity and functionality and to identifying pipes requiring rehabilitation before they deteriorate past the point of renewal. Inspection of wastewater collection systems is typically completed using closed-circuit television (CCTV) cameras to provide visual inspection of the underground infrastructure. Trained technicians then review the videos, identify defects, and provide a condition rating of each pipe that has been inspected.

The recent development of artificial intelligence (AI) tools have the potential to advance the state of the practice of sewer condition assessments. AI algorithms are being developed to automatically identify defects from inspection footage. AI can also be used to identify poor quality videos so that the pipes can be reinspected. AI algorithms for defect autocoding have the potential to improve the accuracy of defect coding and reduce the time required to complete defect coding and pipe scoring. However, the use of AI for autocoding defects is not widespread and the benefits have not been documented beyond a handful of pilot studies.

As part of the City of Salem’s Wastewater Collection System Master Plan, the City evaluated the ability of AI algorithms to automatically code CCTV video and obtain sewer condition assessment data. Two CCTV autocoding vendors were chosen for the pilot study. The pilot study included selecting a wastewater basin with available fully coded CCTV data, establishing testing parameters, and comparing the autocoded results to the City’s own coding. Results of the autocoded CCTV videos from each vendor were compared to results provided by the City in the following categories:

1. Recall: AI able to find any defect within one foot (plus or minus) of City defect.
2. Precision: AI able to find same defect within one foot (plus or minus) of City defect.
3. Accuracy: AI matches the same grade level of City defect.

This presentation will describe the approach taken to evaluate if defect autocoding is a viable option for their inspection of their sewer system. Results of the comparison along with lessons learned and recommendations for implementation of AI for CCTV autocoding will also be presented.