With the increasing automation of production lines, chip residues generated during computer numerical control (CNC) machining may result in tool damage, equipment malfunctions, unscheduled downtime, and increased maintenance costs if not promptly identified and managed. To address this problem, we propose a synthetic data generation framework that integrates 2D image-based modeling with the NVIDIA Omniverse simulation platform to rapidly produce large-scale annotated datasets of cutting tool chips. These synthetic datasets were subsequently employed for training a You Only Look Once (YOLO) object detection model. A comparative analysis was performed between synthetic images generated by Omniverse and manually annotated real images, evaluating their detection performance and labeling costs. Experimental results demonstrated that synthetic data significantly reduces manual labeling efforts while maintaining high detection accuracy. The proposed system facilitates the early detection of abnormal chipping or chip entanglement, providing actionable insights for preventive maintenance and adaptive machining scheduling. Consequently, this approach enhances intelligent tool condition monitoring and predictive maintenance capabilities, thus offering essential technological support for smart manufacturing and automated production lines.