Development of a Predictive Model for Tool Wear in Cutting and Drilling for Multipurpose Machining
Main Article Content
Abstract
Accurate prediction of tool wear is important in machining to ensure product quality and operational efficiency. However, the conventional methods for monitoring the tool wear are generally slow, inefficient, and fail to deliver the real-time accuracy necessary for industrial environments, limiting their effectiveness in modern manufacturing environments. Therefore, this study introduces a novel framework for real-time flank wear prediction, leveraging high-frequency vibration data collected from an accelerometer mounted on a locally manufactured multipurpose machine. The collected dataset, covering cutting and drilling operations, was used to train and test two deep learning models: Long Short-Term Memory (LSTM) and transformer networks. The novelty of this work lies in the detailed comparative analysis and optimization of these models, designed specifically for machining environments. By using several calibration techniques, the optimized LSTM showed higher performance through experiments with a Root Mean Square Error (RMSE) of 0.1417, Mean Absolute Error (MAE) of 0.076, and coefficient of determination (R²) of 0.987, outperforming the transformer model. This finding gives a predictive maintenance approach, addressing the challenges of real-time tool wear monitoring and offering insights into model selection for industrial applications. The proposed framework sets a new benchmark for integrating adaptive predictive techniques into manufacturing, enhancing tool life, minimizing downtime, and supporting the flexibility of predictive maintenance in manufacturing settings.