Online or Offline, Rail Frames Use One Measuring Setup
A rail car frame is a large welded structure whose dimensional tolerances and form relationships affect bogie mounting and running safety. This project pairs an industrial robot with optical tracking, switches stereo vision and laser sensing by feature, and works either inline or offline to complete dimensional measurement, form evaluation and surface defect detection, subject to site acceptance results.
The hard part is the datum, not the dimension
The frame goes through several rounds of welding and machining, so datums are passed along a complex chain. The final dimensional and form state has to be confirmed fully before delivery, and confirmation requires every region's data to sit in one coordinate network. The frame is long, and once datum error accumulates along its length, the two ends stop agreeing with each other.
Occlusion and deep pockets decide the stations
A welded structure has plenty of occlusion and deep pockets, and no single station covers every element. The answer is multi-station capture with optical tracking: most elements are covered by standard stations, and the few that cannot be imaged directly get a dedicated supplementary path. More stations is not the objective; every station should own a clear area of responsibility.
Laser for planes, stereo for holes and edges
Laser scanning covers large planes and long profiles with fast, continuous capture; stereo vision measures hole positions and edges with more stable local accuracy. The sensors switch by feature, and the scheduling logic comes from the measurement plan, whose quality decides capture efficiency and cloud completeness. That plan is where a project of this type succeeds or fails.
One system, two production rhythms
The same equipment runs inline during production and offline with the frame standing in the shop, with one measurement process, one data standard and one report format. Switching is a matter of work organisation rather than equipment change: idle production windows are used for offline tasks, so equipment utilisation and line cycle improve together. That flexibility is often worth more than a single headline specification.
Form and position get their own evaluation
The frame is the base structure connecting bogie and body, and form deviations such as straightness, flatness, parallelism and position affect assembly relationships and running safety. A part can meet a dimensional tolerance and still fail on form, so the two are evaluated separately and then merged into one traceable report.
Model change by program, and where it stops
Frame dimensions and structure differ widely between vehicle types, and the measurement program switches by type: path and evaluation datum load with the program. That avoids dedicated gauges per model and means a model change does not wait for equipment rebuilding.
It suits long, large and complex structures such as frames, bogie frames and large weldments. Where space is tight or online and offline both matter, layout flexibility counts for more than a single index. Where the requirement is only feature presence, a compound measurement system is over-investment.
Questions at project review
Why stereo plus laser for a frame?
It needs both wide, fast point capture and precise local measurement: laser for large-area clouds, stereo for holes and edges. Switching by feature completes dimensions, form and defects in one pass.
How does online to offline switching work?
Same system, same process, same report format; the switch follows production organisation, not hardware. Online it is embedded in the line, offline it works with the frame standing in the shop.
How is multi-station error removed?
Optical tracking measures the robot pose in real time and corrects long-travel error so stations stitch in one frame, without accumulation along the length.
What about heavily occluded areas?
Multi-station path planning avoids occlusion and the sensors adapt by feature. Elements that cannot be imaged directly get a dedicated supplementary pass rather than a larger camera count.