A Novel Kolmogorov–Arnold Attention Allocation Network for Cutting Tool Remaining Useful Life Prediction

Accurately predicting the Remaining Useful Life (RUL) of milling tools based on monitoring signals is crucial for maximizing tool utilization and reducing machining costs. However, tool degradation and working conditions complicate the extraction of features from monitoring data, making the nonlinear mapping from these features to RUL challenging. To address this, a novel Kolmogorov–Arnold Attention Allocation Network (KA-AAN) is proposed, which consists of an Attentional Feature Extraction Network (AFEN) and a KAN regression block. First, the S-transform is applied to multi-sensor signals to create a time-frequency feature dataset. Then, the AFEN allocates attention to extracting attentional features concerning both tool degradation and working conditions by using multiple Kolmogorov–Arnold Attention Blocks and the unique activation function of KAN, which enhances the importance of features at different levels. Furthermore, the KAN regression block maps from attentional features to RUL, learning the most appropriate way to activate and combine features based on the specific circumstances of the tool and machining process. The nonlinear fusion of the attentional features improves the model’s adaptability to different working conditions. Finally, an experiment was designed and conducted to verify the robustness and stability of the proposed method. The experimental results show that the proposed method achieves a mean squared error of 0.97 and a mean absolute percentage error of 8.27% under the same working condition and 2.91% and 18.63% under different conditions (average). Compared to other advanced methods, the proposed method exhibits higher accuracy and adaptability.