Walking a 3D Scanner Along a 1 m Cylindrical Part
Industrial pipes and cylindrical parts more than 200 mm in diameter and over 1 m long are awkward to inspect by hand: the part has to be moved and someone has to climb. Here a robot carries a 3D profile camera and light along the workpiece, and laser scanning measures both 2D defects such as scratches and dents and 3D parameters such as weld height and flatness in one pass. The cycle is about 600 seconds, reports upload to the cloud and archive automatically, subject to site acceptance results.
Why the machine travels and the part does not
These parts usually sit where they are installed or stored, and moving them is slow and costly. Equipment that moves around the workpiece removes the crane, the transfer and the re-alignment, and for the operator it removes two frequent high-risk activities: climbing and lifting heavy objects. The part goes on to the next step from where it was inspected, so production scheduling is not disturbed.
One scan, two kinds of verdict
Scratches and dents appear as local changes in surface form, which is defect detection. Weld height and flatness are geometric parameters obtained by fitting reference surfaces to the point cloud. The data is shared but the algorithms differ, and integrating them avoids a second scan and a second setup, along with the datum drift that would come between two separate measurements.
How weld height is computed
Fit the base surfaces on both sides of the weld, find the weld crest, and the difference is the weld height. The algorithm sounds simple; the difficulty is that the base surface is never ideal, and a weak fit distorts the height. Robust fitting against mill scale, spatter and a slightly curved base is where the engineering sits, not in the subtraction.
Flatness is measured against an ideal cylinder
Flatness here is not whether a surface looks even; it is the deviation between the surface point cloud and an ideal cylinder. How accurately the centre and axis are determined decides how trustworthy the deviation is. On a long part, segment stitching error accumulates along the length, so the stitching strategy has to match the purpose of the measurement.
360 degrees comes from the path, not the camera count
Cylindrical curvature is sensitive to laser incidence angle: too oblique and the point cloud thins out or drops away. The scheme plans the scan path around the circumference within a controlled incidence range, and the robot scans in segments along the length before stitching a complete surface model. Whether coverage is complete and stitching error is controlled is the implementation crux of this class of scheme.
What 600 seconds fits
The cycle suits non-continuous sites and offline inspection rather than a high-speed line. The part stays in place during measurement and moves on afterwards. For companies with several sites and scattered inspection points, a cloud report can be viewed in another location, which cuts paper records and the transcription errors that come with them.
The limits are worth stating. It suits large cylindrical parts that sit in place, not inline scenarios demanding a cycle in seconds; very poor surface condition or micron-level form evaluation needs separate assessment.
Questions about walking scans
Why must the robot travel along the part?
At over 200 mm diameter and often over 1 m length, a fixed field of view cannot cover the surface. Segmented scanning plus stitching is the practical route to 360-degree coverage without moving the part.
Can one data set really serve both defect and geometry checks?
Yes. The 3D scan carries form and geometry together: local change feeds defect detection, and height and deviation after fitting feed the parameters, with both tasks sharing the data.
Is 600 seconds too slow?
On a continuous line, yes; on a site where parts sit in place and flow is not continuous, it is a reasonable trade for removing handling, climbing and waiting, subject to site acceptance results.