MSc. Thesis Defence: Sevim Defne Ünlü

Machine Learning-Based Prediction of Cutting Parameters in Orthogonal Cutting

Sevim Defne Ünlü
Manufacturing Engineering, MSc. Thesis, 2025

Thesis Jury

Prof. Dr. Erhan Budak (Thesis Advisor)

         Dr. Arash Ebrahimi Araghizad (Thesis Co-advisor)

Assoc. Prof. Dr. Lütfi Taner Tunç

Assoc. Prof. Dr. Kemal Kılıç

Asst. Prof. Dr. Faraz Tehrenizadeh

Date & Time: July 17th, 2025 –  11:00 AM

Place: FASS G049

Keywords : Orthogonal Cutting, Machine Learning, Cutting Force Coefficients,

Cutting Parameter Prediction, Digital Manufacturing

Abstract

This study investigates the use of machine learning techniques to predict cutting force coefficients in orthogonal metal cutting operations with minimal experimental data. Unlike traditional studies that rely on large datasets or repeated testing for each new condition, this work demonstrates how models trained on a limited set of experiments can generalize to unseen materials, feed rates, and cutting speeds, as well as to coated tools and lubricated cutting conditions, using only minimal additional input. This is critical for accelerating process design and supporting intelligent manufacturing. It focuses on integrating material properties and cutting conditions into predictive models to estimate key parameters such as tangential and feed cutting constants, edge force coefficients, shear stress, shear angle, and friction angle. Orthogonal cutting experiments were performed on four materials under varying coating and lubrication conditions. Three datasets were prepared with varying input complexity to assess model performance under different levels of data availability. Random Forest and XGBoost models achieved high accuracy in predicting shear angle, shear stress, and especially when force or chip data was included. While accuracy declined for coefficients like and under limited input or unseen materials, the models still captured general trends effectively. This study shows that ML can be used to transfer knowledge across different cutting conditions and surface treatments, reducing the need for repeated experiments. By bridging experimental data and predictive modeling, the study supports data-efficient, scalable, and intelligent manufacturing aligned with Industry 4.0 goals.