Artificial Intelligence, Real Solutions: Identifying Sewer Defects with AI
Pipe inspection as we know it today began in the 1950s using closed-circuit television (CCTV) video with cameras that traveled through sewers recording video used by engineers to code defects as part of their quality control process. Over the last 60 years, there have been upgrades to pipeline defect standardization and inspection equipment, but not to the process.
The scope of pipe inspections has grown with the addition of maintenance of traffic, locating manholes and interacting with residents. Contractors are expected to manage and review the video recordings using more than 200 NASSCO codes to report defects as well. Coding is often the most time-consuming piece and can be open to interpretation as two different people may view a code differently. After working on over 23,000 sewer segment inspections, B&N recognized the inefficient use of time inherent with the inspection and review process and looked to technology for a solution: artificial intelligence (AI)
Introducing BNtelligence: PipeAId™
After two years of development and testing, B&N released BNtelligence: PipeAId, our first AI application. PipeAId streamlines the pipe inspection process by uploading CCTV video to the program and PipeAId scans it to determine if there is a defect. If there is a defect present, the AI models determine which NASSCO code will most likely apply and assigns a level of accuracy for each defect prediction.
B&N engineers verify the defects with a low likelihood of accuracy and spot-check the remaining data. As our engineers correct and verify data, the program gains a deeper understanding of what each specific defect looks like. The more the program is used, the smarter and more accurate it becomes.
All the defects and codes are tied to the associated frame of the CCTV video and stored in a database for our clients to use in their asset management and capital improvement planning.
What are you gAIning?
As PipeAId “watches” the CCTV video to identify and code any defects, it works to reduce the time required to review sewer inspection data and increases the accuracy of the inspection by verifying all codes are correct and identifying incidences that may have been overlooked during the initial evaluation. This results in cost and time savings and more reliable data. The more it’s used, the smarter it becomes, providing a more reliable the dataset for decision making.
PipeAId’s efficiencies could eventually cut CCTV video review time in half! Because the contractor would no longer need to code defects that the AI is trained detect, they can quickly move through pipe inspection and focus on identifying critical defects. The final database is AI-verified so that utility providers can use to make quick, transparent and repeatable data-driven decisions.
The Human Effect
While AI reduces time spent reviewing video, people are still needed to interpret and verify the results. AI does not replace the need for human intelligence but instead supplements and streamlines the involvement. B&N’s AI solution provides one level of quality assurance/quality control (QA/QC) for sewer inspection. But engineers are still needed to spot check and verify the accuracy of the identified defects. The defects are checked on three different levels:
- PipeAId – The AI functionality captures and codes the defects using CCTV video
- Contractor – An engineer confirms that the defects identified by the AI are correctly coded
- Designated QA/QC Personnel – A final check confirms again that the defects identified are correctly coded
If a contractor or designated QA/QC personnel find an incorrectly coded defect, they correct it and retrain the AI. This is how the AI learns and becomes smarter and more efficient with each use. Through the combination of machine and human intelligence, the dataset gathered is assured to be reliable and useful.
AI: The Future of Sewer Inspection
Today, technology is changing more rapidly than ever to enhance everyday tasks. This long-awaited update to the sewer inspection process will help address the growing needs more efficiently. Instead of replacing human intelligence, it helps automate some of the sewer inspection process so that the technical expertise of engineers is used most effectively to verify and interpret results to provide recommendations. To talk with the humans behind B&N’s artificial intelligence, please contact Josh Ford or John Swartzbaugh.
