A Feature Selection–Based Intelligent Framework for Predicting Maximum Depth of Corroded Pipeline Defects

Abstract

Corrosion is one of the most common defects of buried pipelines. Accurate prediction of the maximum pitting depth of corroded pipelines is conducive to assessing the remaining strength of the pipeline. In the context of big data, machine learning has been proved to have good results. However, previous studies have less consideration of feature selection in modeling, so that the interpretation of corrosion mechanism in machine learning model is not clear enough. This paper aims to develop a novel intelligent framework to accurately predict the maximum pitting depth of buried pipelines. The framework utilizes correlation analysis to extract features from many factors related to corrosion depth, and then uses a hybrid machine learning tool to predict the maximum pitting depth. Through empirical analysis, it is found that the pipe age of buried pipelines is the leading factor for pipeline corrosion. The prediction model proposed in this paper uses an improved gray wolf optimizer to optimize the support vector machine, and compares the prediction results with eight other benchmark models. It is concluded that the proposed model has the best prediction accuracy and stability. Finally, this paper discusses the influence of feature analysis on the prediction results, showing that this operation can effectively improve the model’s prediction performance and enhance interpretability.