Politecnico di Torino (logo)

Machine Learning Application for Tool Wear Prediction in Milling

Vinay Nagabhushana Rao

Machine Learning Application for Tool Wear Prediction in Milling.

Rel. Giulia Bruno, Franco Lombardi. Politecnico di Torino, Corso di laurea magistrale in Automotive Engineering (Ingegneria Dell'Autoveicolo), 2020

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution.

Download (3MB) | Preview

Milling is one of the most versatile processes used in the manufacture of various components. With this, the milling tool usage has gained momentum, so as the research on its wear phenomenon. Flank wear has been considered as, one of the most commonly observed and an unavoidable phenomenon in metal cutting process, which is also a major source of economic loss resulting due to material loss and machine downtime. With the aim of implementing a predictive maintenance for the milling process, so as to avoid unnecessary cost and wastage of time due to sudden failure of cutting tool, and also to maintain the best product output quality, one of the applications of Machine Learning has been presented in this thesis by giving due importance to Tool Condition Monitoring. By highlighting the usage of model based maintenance method, the study presents the implementation of the framework of predictive maintenance which has been proposed extensively by many research papers. This thesis presents a method to apply Machine Learning in the prediction of the tool wear, thereby assessing the remaining useful life of the tool for best performance with respect to cost, quality and time. The work here presents the methods of Data cleansing, manipulation of data to extract and select features, utilization of the features in training various machine learning models and testing them to conclude in finding the possible tool wear severeness and to assess the Remaining Useful Life based on wear results. The study also presents the possible tools that can be used to carry out the regression analysis in order to train and test machine learning models like Linear Regression, Bayesian Ridge Regression, Kernel Ridge Regression, Neural Network and so on. Cross validation has been carried out in order to narrow down on the machine learning model, which has been further improved using Hyperparameter tuning. This has enabled to arrive at the best possible results for wear prediction and Remaining useful life of the tool.

Relators: Giulia Bruno, Franco Lombardi
Academic year: 2019/20
Publication type: Electronic
Number of Pages: 110
Corso di laurea: Corso di laurea magistrale in Automotive Engineering (Ingegneria Dell'Autoveicolo)
Classe di laurea: New organization > Master science > LM-33 - MECHANICAL ENGINEERING
Aziende collaboratrici: UNSPECIFIED
URI: http://webthesis.biblio.polito.it/id/eprint/14677
Modify record (reserved for operators) Modify record (reserved for operators)