Published Paper Details:

Tool wear monitoring is a critical requirement in CNC machining to ensure dimensional accuracy, surface quality, and uninterrupted production while minimizing tooling costs and downtime. Conventional monitoring techniques largely rely on manual inspection or rule-based signal processing, which limits their effectiveness for continuous and autonomous operation. In recent years, artificial intelligence (AI) has emerged as a viable solution for data-driven tool condition monitoring using indirect sensor measurements. This study explores deep learning-based approaches--namely Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and a hybrid CNN-LSTM architecture--for vibration-based tool wear classification. The proposed hybrid model integrates convolutional feature extraction with sequential temporal learning to capture both localized signal characteristics and long-term wear evolution. Signal preprocessing strategies, including down-sampling, batch normalization, and sequence standardization, are applied to improve training stability and model robustness. Experimental results obtained from a benchmark brownfield CNC milling dataset demonstrate that the hybrid CNN-LSTM model consistently outperforms standalone CNN and LSTM models, indicating its suitability for practical predictive maintenance applications in Industry 4.0 environments.