Aplicación de técnicas de machine learning en la predicción de la corrosión en las operaciones petroleras de Ecuador.

Abstract

The presence of corrosion in oil extraction pipes not only generates economic losses for Ecuador, but also has harmful effects on both people's health and the environment. This Master's thesis proposes the use of Machine Learning algorithms to estimate the level of corrosion based on data collected regarding production and the chemical processes present in the extraction lines. In this work, the Support Vector Machine, Random Forest and XGBoost classification models are implemented, the same ones that are evaluated based on a set of metrics, that is, the comparison of the results of the accuracy, precision, sensitivity, specificity and f1-measure, and according to these results, a model is selected so that later a hyperparameter tuning is made to improve the performance of this model. In addition, a feature engineering stage is considered, which includes the normalization of the data, the elimination of variables with high correlation and the balancing of the data set. Also, a dashboard-type web application is included so that the end user can make the prediction in a friendly way. It is worth mentioning that this entire project was developed using the R programming language.