Implementation of Machine Learning Techniques for Prediction of the Corrosion Depth for Water Pipelines

Water pipelines are crucial components of water supply systems that safely supply water produced by water purification plants to consumers. Metal pipelines deteriorate over time owing to various physical, environmental, and operational factors; in particular, corrosion occurs inside and outside the pipelines due to the characteristics of the pipeline material. In this study, models were developed using machine-learning algorithms to predict internal and external corrosion depth. The hyperparameters of each model were determined through Bayesian optimization, and model training, validation, and prediction were performed. The proposed machine-learning techniques for predicting the corrosion depth of water pipelines can overcome current limitations, such as the prediction of deterioration and residual life of water pipelines and the selection of the diagnostic points of the pipelines. These models may be increasingly valuable with changes in the technological paradigm of diagnostic methods.