Parametric vision cell
A fully parametric cell. Any bar from Ø20 to 60 mm and 250 to 640 mm long, found by 3D vision, loaded, turned and marked.
2018–20 · GreenBau, Romania · Industrial robotics
The problem
A maker of high-quality cast parts, working with leading European suppliers, faced rising demand for fast, precise, consistent production of complex parts, and a shortage of skilled people who could hold that precision shift after shift.
Requirements
- The full range of cylindrical parts, Ø20 to 60 mm and 250 to 640 mm long
- Work with the plant’s existing wooden crates, which vary a lot
- Integration with an Okuma lathe
- Integration with a part-marking machine, with automatic marking
- Safety without fences
- Manual work still possible
- Place each part’s zero face at a defined distance from the chuck, despite uneven part lengths
- A cycle time of about two minutes
What was built
- The robot on an angled base, keeping the work envelope free in manual mode
- Parts arrive and leave on the plant’s own carts and crates, located by mechanical stoppers
- 3D vision to find the parts in the crate
- A double magnetic gripper with a 3D laser-profile camera built in
- Safety with a multi-zone safety scanner, emergency stops and status stack lights
- Full integration with the lathe and the marking system
- An HMI inside the robot’s teach pendant that parametrizes the whole system by part type and process
Challenges
- The wooden crates varied so much that the crate and its parts could not be found reliably
- Painted a small white area on each crate, wrote a flexible vision program to find crates, and added a reference post in space for a repeatable scan position
- A semi-transparent factory roof changed the light all day, so vision was unreliable
- An optical filter on the camera that passes only its working wavelength
- A rare bad pick could hit the open chuck
- A validation system. A distance sensor checks pick symmetry and the part’s position in the gripper, and a second sensor inside the lathe confirms correct insertion
- The wide range of weights and lengths shifted the part and made long parts sensitive to gripper angle
- Per-part weight computed from its parameters, a spindle-vector calibration accurate to micro-radians, and parametric calibration of every key point by weight and length
- The 3D profiler camera returned wrong positions because of how the technology works
- A spatial geometric calibration that corrects the camera’s errors, merged into the application’s overall parametrization
Technology
Kawasaki RS080N · Okuma GENOS L3000-e · SICK 3D vision · Laser profiler · Double magnetic gripper · Safety scanner · Part marking · Parametric HMI
At a glance
- ROBOT
- Kawasaki RS080N · 80 kg
- MACHINE
- Okuma GENOS L3000-e
- VISION
- SICK TriSpector1060
- GRIPPER
- double magnetic
- MODE
- parametric part family
- MARKING
- engraving
- SAFETY
- laser scanner zone
My role. Designed and commissioned the cell layout, gripper, safety, robot programs and PLC
Screens
