Intelligent chatter detection in turning operations with imbalanced data using conditional generative adversarial networks

Turning is a widely used machining process in the industry. However, chatter-induced vibrations limit cutting efficiency and damage tools and workpieces. Despite advances in chatter detection, most methods assume balanced training datasets, which is rarely the case due to the difficulty of collecting chatter signals. This study introduces a deep learning-based chatter detection approach that addresses data imbalance by using one-dimensional conditional generative adversarial networks to synthesize chatter data. A low-cost sensor collects turning data, and the complete ensemble empirical mode decomposition with an adaptive noise algorithm is applied for the first time. The reliability of synthetic data is evaluated under different conditions, and its effectiveness is compared to data generated by variational autoencoders. The model, trained on synthetic data, is tested with real-world scenarios. It shows its ability to detect chatter with classification accuracies between 98.5% and 100%. By addressing data scarcity with generative AI and incorporating improved signal processing, this work delivers a practical, low-cost, and adaptable AI solution for real-time chatter detection in industrial turning applications.