Experimental Study of a Sensor Based Toolwear Detection Using Machine Learning Models

Tool wear detection is essential for optimizing machining processes, maintaining product quality, and reducing production downtime. Continuous monitoring of tool condition helps in preventing unexpected tool failure and improves overall manufacturing efficiency. This project presents a sensor-based tool wear detection system integrated with machine learning techniques for accurate classification of tool conditions. Real-time data is collected during turning operations on a lathe machine using multiple sensors mounted near the cutting tool. A vibration sensor measures changes in vibration caused by tool wear, a temperature sensor monitors heat generated at the tool–workpiece interface, and an ultrasonic sensor detects variations in reflected signals due to surface irregularities. The acquired sensor data is transmitted to a microcontroller for preprocessing and then forwarded to a processing unit for analysis. Relevant statistical features are extracted from the collected signals and used to train machine learning models such as Support Vector Machine (SVM) and Random Forest. These models classify tool conditions into normal, moderate wear, and severe wear categories. The proposed system enhances predictive maintenance, reduces operational costs, and supports intelligent decision- making in modern manufacturing environments.