Computer Vision Method for In Situ Measuring Forming Accuracy of 3D Sand Mold Printing

Abstract

The three-dimensional sand mold printing technology (3DSP) for casting sand molds via a binding jet is a breakthrough in the casting mold-making process. It is a favorable combination of digital forming technology and sand casting, which is a significantly interesting research area in the foundry industry. This study has proposed an edge extraction approach for the forming region in the sand bed image. With the edge information, the study measures the forming accuracy of the printed molds, which offers a basis for assessing the forming quality after 3DSP. The extracted edges essentially match the original image through the validation of cube printing. The error between the measured and actual size is below 0.6 mm, and the standard deviation of the straight line edge is below 0.170 mm, which fulfills the accuracy requirements for 3D sand mold printing.

Keywords

• three-dimensional sand mold printing
• computer vision
• forming accuracy
• convolutional neural network
• non-maximum suppression