Application of advanced regression methods for wear prediction of superalloys

Abstract

Analytical models able to predict the tool wear can provide companies instruments to optimize the cutting processes. The focus of this thesis is to accomplish a study of the tool wear process in the turning process of superalloys, including its dependence on multiple factors related to the characteristics of the workpiece and machinery used for turning. As a natural extension of this study we propose the application of some statistical and machine learning techniques to address the prediction of the tool wear. Data corresponding to different tests carried out as part of the European project called Himmoval is used. The process of prediction involves selecting features from the variables acquired by different sensors that characterize the machining process. Additionally, several machine learning algorithms are implemented and applied to analyze the data from the wear experiments. Among these algorithms, Gradient Boosting Regressor predominates over the rest of regression methods evaluated.