Similarity evaluation of topography measurement results by different optical metrology technologies for additive manufactured parts

Abstract

Similarity evaluation of topography measurement results by different optical Similarity evaluation of topography measurement results by different optical metrology technologies for additive manufactured parts metrology technologies for additive manufactured parts

An Analysis of Polymer Gear Wear in a Spur Gear Train Made Using FDM and FFF Methods Based on Tooth Surface Topography Assessment

“…There are a number of non-contact methods of verifying the condition of the surface layer using vertical scanning. The most frequently used ones include: confocal microscopy [53][54][55], structured light projection [56,57], focal differentiation microscopy [58][59][60] or coherence scanning interferometry [61]. Although increasingly modern optical methods of surface geometrical structure measurements are widely available, contact methods are still superior in terms of measurement repeatability [41,50].…”

In situ monitoring of direct energy deposition via structured light system and its application in remanufacturing industry

“…The resolution of the SLS system could be adjusted according to various scenarios in the application. In this study, the SLS system was used to monitor the DED printing process, capture the 3D morphology [23] of additive manufactured parts [24] and evaluate measurement results among different optical systems [25][26]. Each printed layer's threedimensional characteristics were captured and stored to optimize printing parameters and repair used parts.…”

Similarity evaluation of 3D surface topography measurements

“…The similarity evaluation between STMs is practically useful for several reasons: (1) it can be leveraged to resolve the confusion or miscommunication in practice caused by system discrepancies, especially in this new era of Industry 4.0 with cyber-environment [6,7]; (2) it can be used to examine the repeatability and reproducibility of 3D optical gauges because variation in surface topography data leads to the dissimilarity between measurements. Larger variations are expected to result in lower similarity scores between measurements; (3) it can help to answer important questions such as whether the measurement system can adequately discriminate between different parts and whether a measuring process is stable over time; (4) it can be used to monitor the uniformity of additive manufactured parts both within and across layers to achieve automatic in-situ process monitoring since the uniformity of printed layers in surface topography is critical for the product quality; and (5) it provides a mechanism to examine the discriminating ability of the similarity metric by binary classification, in which we classify a pair of topography measurements as matched or non-matched.…”

“…However, these feature-based methods may neglect geometric details and thus are prone to larger error rates. The second category of research by Zhang et al [6] and Zheng et al [7] used the Pearson correlation coefficient (PCC) and image distance to compare STMs, which considered the geometric details of the entire surface. Nevertheless, the work in the second category suffers from multiple limitations: (1) it was assumed that the scanned region of one STM was completely covered by that of the other STM, which could be violated in practice because STMs were usually partially overlapped with an irregular shape; (2) the interpolation estimated depth values on manually selected meshed grids, which was subjective and might result in information loss; (3) the comparison was in a small scale; (4) the comparison only considered measurements from different systems.…”

“…Our important contributions in methodology are summarized as follows: (1) the registration strategy we established is more efficient and robust without any requirement of initial transformations, while the registration method used in Zhang et al [6,13], Zheng et al [7] requires manual initial transformation; (2) we are the first ones to use convex hulls to extract overlapped regions of irregular shape in this application context so that the framework can be used to compare partially overlapped measurements; and (3) our interpolation strategy keeps all the measured data without any manual operation or down-sampling, which makes the similarity measure more reliable. For decision making and experimental studies, (1) we provide two strategies to distinguish matched vs. non-matched pairs with clear and easy-to-use guidelines; (2) besides comparing systems, we also examine how repeated measurements and operators affect the similarity, which are critical for quality assurance; (3) we provide error rates evaluation and statistical validation for our framework in large scale by careful experimental design; (4) we create a database of STMs with both raw and preprocessed STMs across systems, operators, and repeated measurements; and (5) the thresholds we calculated and the similarity scores with binary labels (matched vs. non-matched) can be used as references for related work.…”