3D point cloud measurements of the surface of HFMI treated and untreated linear butt welds

The data consists of 3D point clouds of linear butt welds untreated or high-frequency mechanical impact (HFMI) treated. The scans have been acquired using a Wenglor MLWL131 profile sensor mounted onto a KUKA KR 60-3 industrial manipulator.

The purpose of the data was to develop and validate an automated system for robotic weld toe treatment and quality assurance. A 3D surface representation of the weld allows for identifying the location of the weld toe, which can be used for automatically generating an adaptive robotic trajectory for treating the weld. Subsequently, the treated weld can be rescanned, and the quantitative quality measures (e.g., indentation depth and groove width) can automatically be determined for quality assurance.

The welds have therefore been treated using three different approaches to compare the treatment quality:
- Manual: The weld is treated manually by a human operator.
- Robotic manual: The weld is treated using a robot that follows a manually programmed straight trajectory.
- Robotic adaptive: The weld is treated using a robot that follows an automatically generated trajectory that is adapted to the inconsistency of the weld toe. The treatment is generated based on a point cloud of the weld.

The data set consists of 36 scans, representing 50 mm sections of butt weld with two weld toes, all acquired using the same straight scanning trajectory and normal to the surface of the sample. 18 of these scans are of untreated weld, whereas the remaining 18 are HFMI treated welds. 6 scans of each treatment type: manual, robotic manual, and robotic adaptive.