We propose 3D-CSAD, a novel method for untrained 3D anomaly detection in complex manufacturing surfaces using 3D point cloud data. Our method segments the surface into simpler components and uses Robust Principal Component Analysis (RPCA) for accurate detection.