A novel 3D dimension estimation approach in additive manufacturing based on virtual-real hybrid point cloud data and semantic segmentations

Abstract

The advancements in additive manufacturing (AM) technology, while empowering the manufacturing of complex structures, have also increased the demand for corresponding measurement techniques. While 3D scanning and reconstruction have been employed for quality inspection in AM, there remains a gap between scan results and specific dimensions, hindering the progress of AM processes toward greater precision and speed. Aiming to bridge the gap between AM components and dimensional features, this paper introduces a novel method to estimate pre-defined dimensions from point cloud of AM objects. Building upon the foundation of semantic segmentation and post-processing calculations, hybrid data and down sampling are applied and evaluated. Comparisons with Coordinate Measuring Machine (CMM) measurements confirm that the proposed method in this paper significantly reduces measurement time and simplifies the measurement process, cutting the computation time down to 12 % of the original while maintaining high accuracy. The segmentation accuracy can reach 89 % when using a hybrid dataset with virtual data. The measurement uncertainty of the proposed method is quantified, confirming that the dominant contributor to the measurement uncertainty comes from the scanning instrument, validating the reliability of the proposed method.