Research on chatter monitoring of ultrasonic milling of thin-walled parts based on ICS-CNN network

In modern industry, thin-walled components have become critical elements in high-end manufacturing due to their unique structure and lightweight properties. However, their production process is prone to chatter, which severely impacts machining quality and efficiency. Although ultrasonic vibration grinding technology can partially suppress chatter, the issue remains unresolved. To address this, this paper proposes an online chatter monitoring method based on an improved convolutional neural network (ICS-CNN) and the Sparrow Optimization Algorithm (SSA). This approach enhances key information capture through multi-scale feature extraction and attention mechanisms, while incorporating residual connections and a feature pyramid structure to strengthen the model's ability to identify subtle chatter characteristics. Input signals undergo frequency domain analysis and filtering to improve data quality. The SSA algorithm is further employed to optimise network parameters, constructing the SSA-ICS-CNN intelligent monitoring model. Experimental results demonstrate an identification accuracy of 98.37 % with a decision time of merely 147 milliseconds, while visualisation techniques validate its discrimination precision. Compared to conventional convolutional neural networks, this approach achieves significant improvements in both recognition accuracy and response speed, effectively overcoming limitations inherent in traditional methods reliant on manual feature extraction and dynamic response delays.