Tool Wear Prediction in the End Milling Process of Ti-6Al-4V using Bayesian Learning

Titanium alloy is one of the most popularly used materials in various industries because of its desirable mechanical properties. However, it is also a difficult-to-cut material because of its low specific heat and thermal conductivity. This influences an end milling process of titanium alloy through tool wear that affects not only product quality but also productivity. Therefore, accurate tool wear prediction is required to improve titanium alloy quality and replace the tool at an appropriate time. Furthermore, since the effects of tool wear prediction on the overall machining process are significant in terms of cost and time, the uncertainty of tool wear prediction should be taken into account. In this work, a deep learning-based uncertainty-aware tool wear prediction model using Bayesian learning approach is proposed. The proposed probabilistic model could produce a predictive distribution over estimated tool wear with uncertainty awareness. Experiments with real-world end milling processes have proven the effectiveness of the proposed method. The results indicate that the proposed method outperforms existing comparative tool wear prediction models.