Laser scanning is often used for reverse engineering or manufacturing part verification. In this application, we scan an electronic package to generate a CAD model to later verify the dimensions with the original physical package.
The dimensional accuracy of laser scanning is critical to disambiguating features.
CloudCompare (www.cloudcompare.org ) has been used to carry out the analysis together with Autodesk Fusion 360 and AutoCAD (www.autodesk.es) for the conversion and inspection of models.
The CAD model was purchased from SparkFun Electronics (www.sparkfun.es). It is the encapsulation of the Big Red Box (Box – PRT-11366 – SparkFun Electronics 1) which comes with a prototyping board used for electronic projects. A 3D CAD model of the package is freely available for download in Sketchup, STL, IDES and Solidworks formats. Once the Solidwoks model was imported into Autodesk Fusion 360, only the base was exported to STL format to be used in subsequent analysis. There are some differences between the CAD model and the actual object. The rounded corners of the mounting flange and the lip around the top of the box are not shown. The CAD model provided by Sparkfun is probably not a manufacturing model, but rather a simplified design model.
The laser scan was performed in the high-precision scan mode. The object was scanned on the turntable while the scanner was stationary to capture the exterior. Then, the scanner was held in hand to scan the interior. The surface was a bit reflective, however good results were obtained as a layer of Aesub Blue was applied to dull the surface and compare the quality of the scans. Overlay scanning performed better in the final results. Scanning and processing only took a few minutes.
Scan to model registration
The coordinate system of the point cloud data originally did not match that of the CAD model, so cloud-to-grid registration had to be performed to align the models. This alignment was performed in CloudCompare by first selecting three common points between the two models to calculate a rough alignment that was then refined by an ICP algorithm. The resulting RMS of the final record was 0.36 mm and was calculated using 50,000 points.
Point cloud analysis
The laser-scanned point cloud was then compared to the CAD model in CloudCompare by analyzing the distances of the points to the mesh surface. The point cloud was then colored with a heat map and provided with a histogram to give the average metric of the quality of fit. Points are within ±1mm, averaging 0.15mm. The mean offset of 0.15mm is likely due to differences in the CAD model, causing a population of points to be more than 1mm from the CAD model surface.
The sides of the box are 0.45mm to 0.54mm apart at the top of the flange, while the top of the box is set to the point cloud. This is because the plastic injection molded part has a draft angle that is not expressed in the CAD model to allow for removal of the mold. The laser scanner was able to capture the small border around the top of the box, it looks good on the mesh and point cloud models. The scan shows that the base of the case is 0.26mm wider than the CAD model, but this could also be due to the angle of inclination. The screw bosses on the bottom of the enclosure are shown to be in the wrong location on the CAD model.
The deviations between the mesh model created from the point cloud and the CAD model are similar and the mesh has enough detail to be used instead of the CAD model in design work.
Our 3D scanner has enough precision and resolution to capture a mechanical part with a high level of detail and quality. Point cloud noise has the tolerance needed to perform manufacturing verification and reverse engineering of mechanical parts. In the absence of CAD models, the mesh model produced by the scanner can be used in mechanical design with a high level of reliability.