Vision-based inspection system for cladding height measurement in Direct Energy Deposition (DED)

Abstract

A vision-based inspection system based on three digital cameras is proposed for measuring the cladding height in the Direct Energy Deposition (DED) process. To improve the accuracy of the cladding height measurements, an image processing technique is applied to remove the undesirable zone from the binary image. Furthermore, since the unit length in the captured images is different to that in the world coordinate framework, a calibration bar method is designed to transform the pixel value to the real size. In the proposed approach, a calibration bar method is employed to compensate for the Field-of-View (FOV) and perspective effects in the trinocular system. An image-processing technique is then employed to isolate the laser nozzle and melt pool in the captured images. Finally, the cladding height is estimated based on the distance between the tip of the laser nozzle and the centroid of the melt pool. The validity of the proposed approach is demonstrated by comparing the inspection results for the cladding height of a horseshoe component with the measurements obtained using a 3-D scanner. The maximum estimation error is found to be just 4.2% Overall, the results confirm that the proposed trinocular vision-based system provides a rapid, convenient and accurate means of determining the cladding height in the DED process.