Assessing the influence of land use on concrete sewer pipe corrosion through AI-based predictive modeling

Abstract

The corrosion of sewer pipes is greatly affected by land use and population density, which in turn impact the types and amounts of organic pollutants present in wastewater. Yet prior research has not extensively examined the connection, between these variables and the risk of corrosion. This study aims to fill that gap by forecasting corrosion risk through the use of land use information, population density and physical attributes of pipes including length, diameter and age. Machine learning techniques, XGBoost, LightGBM and SVM were created and assessed. SHapley Additive exPlanations (SHAP) analysis along with permutation feature importance were utilized to interpret model predictions and evaluate the influence of each factor. The XGBoost model exhibited the performance attaining an accuracy of 0.87 a recall of 0.72 for corrosion instances and 0.88 for non-corrosion instances. Findings indicated that agricultural land use exerted the effect, on corrosion likelihood around 77% trailed by commercial/residential and government land uses. In comparison green belts and conservation areas did not exhibit any impact on the risk of corrosion. These findings provide insights for prioritizing sewer inspection and upkeep. Identifying regions susceptible, to damage enables infrastructure managers to adopt a more proactive strategy and equips environmental engineers with the data necessary to implement focused actions that minimize corrosion and prolong the lifespan of sewer systems. Practically, the findings assist planners and engineers in targeting high-risk pipes for timely inspection and maintenance.