Font Size: 

The issue of pipeline failure has garnered considerable interest from variousresearch communities due to its notable repercussions on the worldwide economy,as well as the risks associated with leaks, explosions, and expensive periods ofdowntime. This paper aims to build a model for classifying and predicting thecorrosion degradation of a pipe used to transport water in mines by the QuebecMetallurgy Center. To this end, two types of models were developed: three binaryclassification models: SVM, RF, and KNN, yielding F1-measurements of 0.968, 0.969, and 0.945 respectively, and a time series model, LSTM, which, with a lossof less than 0.01, was able to predict average variations in pipeline thickness for 63 days.

Machine Learning, Classification, Prediction, Pipeline Corrosion