Empirical Data-Based Condition Prediction for Stormwater Pipelines with Machine Learning
Jingyi Qi, Michael Smith, Nicole Barclay
Abstract
Aging compounded with the urbanization encroachment and economic pressures have rendered underground infrastructure a great challenge to municipal agencies in terms of maintenance. Reliable condition prediction can alleviate this burden for underground pipelines by providing decision support on optimal renewal, replacement, and maintenance. However, municipalities are challenged on the management of stormwater pipelines, as it is usually constrained by limited budget and time. Different from traditional mathematical models, machine learning shows better strength in processing large datasets amidst some missing or "noisy" data. This paper proposes a framework for a novel data-driven model for predicting stormwater infrastructure conditions to identify at-risk pipelines and culverts using existing data inventory.