Sewer Pipe Blockagr Prediction

October 27th, 2016

Background

Wastewater pipe networks are valuable urban infrastructure assets that are responsible for reliable wastewater collection. However, a considerable number of blockages can occur each year on wastewater pipes due to various reasons, e.g., wet wipes, grease, debris, tree roots and etc. Blockages can cause significant social-economic costs, and hence have become the primary challenge for water utilities. Preventative maintenance is the key to tackling the challenge in a financially viable way. The basic idea of preventative maintenance is to identify the high-risk (high probability of blockage in future) pipes proactively, and then maintain them in time to prevent potential blockages. Therefore, the critical component of such strategy is the risk prediction model that can accurately predict future pipe blockages.

Technologies

ExByte proposed a data-driven machine learning-based method for blockage prediction. The model considers not only the physical characteristics of pipes (e.g., material, laid date, length, etc.) but also environmental attributes (e.g., tree coverage, weather condition, demographic features, etc.) for prediction. The effectiveness of the model is validated on real-world wastewater pipe network, showing significant potential for reducing social-economic costs caused by blockages.

  • Spatial-temporal factor analysis for identifying key influential factors.
  • Trend analysis for predicting the total number of blockages in future.
  • Stochastic point process-based prediction model for short-term/long-term blockage risk prediction at pipe level.

Outcomes

The outcomes of the project:

  • Factor analysis – Identify key influential factors for different types of sewer blockages. The discovery can provide guidelines for utilities to optimize their current operational practice.
  • Trend prediction – Predict the total number of sewer blockages in six-month advance based on weather and soil condition for strategic planning.
  • Blockage prediction – Predict future blockage likelihood for each individual pipe with significant accuracy improvement compared with the state-of-the-art approaches.
  • Decision support – The predicted risk likelihood can be incorporated with consequence cost together to make decision support for preventative maintenance.

5