<p>The retention of frozen sand chips generated during CNC machining of frozen sand molds can accelerate tool wear and compromise casting quality. To address the challenges of chip accumulation and incomplete removal during pneumatic-assisted CNC machining of these molds, this study investigates a detection method based on 3D point cloud data. The proposed approach enables efficient and accurate quantification of sand removal rates in molds with complex geometries, offering a practical solution for enhancing machining performance and casting reliability. The grid projection method was employed to measure the volume of various known structural models. An optimal grid size of 2.6 mm resulted in a volume measurement error of only 0.58%, meeting the requirements for sand removal detection and validating the feasibility and reliability of the proposed method. Based on this detection method, auxiliary sand removal experiments were conducted to examine the effects of blowing distance and airflow rate on molds with different cavity geometries, with numerical simulations simultaneously providing auxiliary analysis. The results show that a blowing distance of 110 mm and an airflow rate of 160 L/min yield an optimal flow field distribution and superior sand removal performance, achieving a removal rate exceeding 97.02% for all mold cavities. Furthermore, the analysis of airflow requirements revealed that the curved-bottom groove attained satisfactory performance at an airflow rate of 40 L/min. In contrast, the flat-bottom square groove and the flat-bottom circular groove required an airflow rate of 120 L/min.</p>