Vibration-Based Fault Diagnosis of Multi-Point Tool Inserts in Vertical Machining Centers Using Decision Tree, Random Forest Classifiers and Hybrid Stacking Model: A Machine Learning Approach

The condition of multi-point tool inserts is crucial for optimal machining performance, impacting surface finish, power consumption, and process stability. Early detection of tool faults can signifi-cantly enhance machining efficiency and prevent potential failures. This study presents a vibration-based real time fault diagnosis approach for triangular multi-point tool inserts using Machine Learning (ML) techniques. Time-domain vibration data is collected from various both defective and fault-free configurations of six tool inserts while performing surface milling operations on a Vertical CNC Milling Machine (VMC). An event-driven method that is developed in a Visual Basic Environment (VBE) is used to extract statistical information. The most significant features are selected using the Decision Tree (DT), Random Forest (RF) classifiers and Hybrid Model (DT + RF Stacking) for effective classification. The study evaluates the performance of these two su-pervised tree-based ML algorithms with Ensemble Learning Approach to determine the most accu-rate classifier for predicting tool insert faults. Major Findings: Results indicate that the DT classifier achieved a classification accuracy of 98.75 % using 10-fold crossvalidation, while the RF classifier achieved 97.53 % and hybrid stacking achieved accuracy of 99.25 %. This approach facilitates early detection of tool defects, improving machining process-es and reducing downtime.