An AI tool cuts pipeline defects detection costs by up to 50 percent

Summary:

An AI tool allows utilities to automate the identification, annotation, grading, and reporting of pipeline defects with high accuracy. This enables the workforce to concentrate on making timely maintenance decisions rather than performing routine inspections. Additional benefits include cost reduction, decreased error rates, and minimized service disruptions.

Client: 

DC Water operates as an independent authority within the District Government, maintaining a distinct legal status. It provides drinking water and handles wastewater collection and treatment for over 672,000 residents and 17.8 million annual visitors in Washington, D.C.

Problem Statement: 

As part of their cost-benefit analysis, DC Water sought methods to reduce sewer pipe faults and lower inspection costs. Traditional inspections involve challenging and time-consuming manual tasks, prompting DC Water to seek more efficient solutions for pipeline inspections.

Results: 

• Detects 50 specific anomalies in wastewater utility infrastructure.
• Analyzes 60 minutes of inspection video and produces a report in just 10 minutes, compared to the 75 minutes needed for manual analysis.
• Cuts anomaly detection costs by up to 50 percent, allowing funds to be redirected from maintenance to capital improvements.
• Achieves a 90 percent accuracy rate, reducing human errors caused by fatigue and distraction, and improves detection reliability by up to 20 percent.
• Frees up staff to focus on complex issues by using AI for routine inspections.
•  Offers utilities a potential ROI of up to 350 percent over a three-year period.
• Provides enhanced long-term monitoring by storing pipeline data and metadata.

AI Solution Overview:

DC Water, in collaboration with an IT services provider Wipro, developed an AI tool called Pipe Sleuth, utilizing Intel® Xeon® Scalable processors and the Intel® Distribution of OpenVINO™ toolkit. Pipe Sleuth automates the identification, annotation, scoring/grading of pipeline conditions, and reporting of pipeline defects.

It leverages a dataset of 26,600 annotated images from documented pipe inspection videos. A ML model, trained using TensorFlow and optimized with the Intel® Distribution of OpenVINO™ toolkit, allows Pipe Sleuth to compare new pipe inspection videos with known anomalies.

The solution rapidly detects quality issues, eliminating the need for manual review and coding of underground sewer pipeline video scans. It generates comprehensive inspection reports that integrate with commonly used asset management systems to automate maintenance work orders. Additionally, Pipe Sleuth interfaces with geographic information systems to map precise task locations and facilitates easy retrieval and viewing of specific scenes in pipe inspection videos.

References: 

1. DC Water: Streamlined Sewer Pipe Inspection Analysis. https://www.intel.com/content/www/us/en/customer-spotlight/stories/dc-water-customer-story. html
2. Government Technology Solutions for Positive Mission Outcomes. https://www.intel.com/content/www/us/en/government/public-sector-solutions-overview.html

Industry: Public Services

Vendor: Wipro

Client: DC Water