Prediction of remaining useful life of milling tool using a low-cost data acquisition system

This study addresses a critical challenge in smart manufacturing: monitoring the condition of end mill tools and accurately predicting their Remaining Useful Life (RUL). The development of machine learning enabled tool condition monitoring systems is essential for advancing unmanned and automated machining operations, where early fault detection can reduce tool failure, unplanned downtimes, and overall manufacturing costs. In this work, a novel, cost-effective, and compact Data Acquisition (DAQ) system was developed using an Arduino microcontroller, MPU6050 accelerometer, and the PLX-DAQ Excel add-in. The low-cost hardware configuration makes the system particularly suitable for adoption by Micro, Small, and Medium Enterprises (MSMEs), supporting scalable and accessible predictive maintenance solutions. After collecting real-time vibration signals during milling, features were extracted from the time, frequency, and time–frequency domains. Significant features were selected using the Least Absolute Shrinkage and Selection Operator and used to train various regression models for predicting both tool wear and RUL. Additionally, a fusion of feature approaches was introduced to enhance the results. A total of ten regression models, including ensemble approaches, were evaluated using performance metrics such as Coefficient of Determination (R²), Relative Squared Error (RSE), Mean Absolute Error (MAE), Relative Absolute Error (RAE), and Root Mean Squared Error (RMSE). Among these, the CatBoost Regressor with fusion of features outperformed others by achieving the lowest prediction errors and highest accuracy. This research contributes toward achieving multiple United Nations Sustainable Development Goals: SDG 9 (Industry, Innovation and Infrastructure) by promoting affordable digital transformation in manufacturing, and SDG 8 (Decent Work and Economic Growth) by enabling more efficient, less labour-intensive operations.