PIPE AI – AI for Pipeline Condition Assessment

by Rebecca Shanahan
Wave Blue
Feb 07, 2021

The Issue at Unitywater

Unitywater is a south-east Queensland, Australian utility that supplies water and sewerage services to a population of approximately 791,000 residents across a 5,223km2 geographical spread. With over 12,000km of water and sewerage mains to manage and maintain at all times, Unitywater struggled to maintain consistency in the process of pipeline condition assessment, which had been performed manually for decades.

In the past, Unitywater trialed various solution sets to detect pipe anomalies although had not been successful. Historically, it was found that extracted analysis reports (data from WinCan) was not accurate and held inconsistencies from different operators.

In response to this issue, Unitywater wanted to apply standard rule sets on the classification and detection of anomalies using CCTV. Unitywater not only wanted to automate defect identification, but also the process for inspection prioritization and renewal prioritization.

The Solution

PIPE AI Pty Ltd is a registered Australian Proprietary Company, comprised of a partnership between PEAKURBAN and BlackbookAi.  The company provides a customizable toolkit that utilizes artificial intelligence to review CCTV footage of pipe conditions and checks for anomalies in the pipes including cracks, roots, blockages, etc.

The PIPE AI solution comprises of multiple components including predictive failure of Assets (PIPE AI – Predict) and inspection of footage for the identification of anomalies (PIPE AI – Review). This Pilot focused on using the Review product. PIPE AI can consistently review and identify the anomalies and output a detailed report of the defects. Once the process is completed, this output can be reviewed by staff who can make any changes to the defect classification (if required), to further improve the model’s detection rate over time.

The Pilot

The project was commissioned as an “open-ended pilot” which commenced in November 2019 and completed in March 2020, with ongoing activity.

The solution was deployed on Amazon Web Services (AWS) to rescan the video footage and classified faults consistently using AI, machine learning, and industry best practices. The pilot was delivered in two stages:

Phase 1 – Proof of Value, involving:

  • Building the AI toolset to automatically analyze CCTV footage and identify defects in line with the Industry standard Defect coding to assign a condition score for maintenance and renewal prioritization.
    • Testing the software to quantify the accuracy against previously manually scored videos to identify false positives and learnings.
    • Applying manual interventions to further improve the accuracy of the tool and enable accuracy levels above 95%.

Phase 2 – Operationalising, involving:

  • Front end software work.
  • Asset inspection prioritization module based on asset attributes and failure rate.
  • CCTV Analysis and reporting – AI.
  • Asset risk and investment prioritization.
  • Renewal and maintenance budgeting.
  • Formal reporting.

As more videos were consumed and faults identified and processed the AI capability increased its accuracy over the pilot term to achieve 100% accuracy based on report comparisons.

The Results

The pilot was deemed successful by Unitywater. PIPE AI was identified to be on a pathway to deliver a seamless service where defects are identified faster and with more accuracy than humans including:

  • Reduction in Adhoc CAPEX spending.
  • Potential for savings in Unitywater of up to 1 x FTE’s being reallocated to higher-value activities.
  • Potential to accelerate the process of data capture, resulting in a significant reduction in risk.
  • Enhanced consistency in interpretation and accuracy in data – more targeted renewals and reduce renewal spend.
  • Increased compliance in Development Services – where developers provide CCTV of their PVC pipes, Unitywater has greater compliance and ability to validate pipe configuration

Further Development

Unitywater is working with PipeAI to advance the technology to the next stage to create the solution as a corporate approach. Removal of the laptop and have appropriate specified and supported infrastructure.

······

About PIPEAI

PIPE AI offers an end-to-end solution for pipe condition management that is a game-changer for the industry. Powered by Artificial Intelligence, PIPE AI  informs your business of predictive failure works and schedule inspections by priority. Its platform automatically assesses CCTV footage to identify cracks and anomalies in pipes. Let PIPE AI do the tedious tasks so you can spend more time on high-value work.

Learn more about PipeAI: www.pipeai.com.au