- Joined
- Feb 12, 2013
- Messages
- 160,900
- Reaction score
- 57,844
- Gender
- Undisclosed
- Political Leaning
- Centrist
I started a new job a couple weeks ago. Basically, my company sells automation equipment, with my specialty being in Robotic Welding Systems. In short, we are exploring uncharted waters. Primarily, our focus is in CAD to Path, semi-Artificial Intelligent robotic welding.
CAD to path can be demonstrated by the simulation shown below.
What you see above was basically a CAD drawing run through simulation software. Those numbers are then crunched and fed directly into an actual robotic system.
Currently, one big headache is applying the above to what you see below.
What you see in the pic is a prefab steel stud wall, used in commercial building construction. The future is to automate this process by using a series of welding robots instead of 5 guys. The problem is that, in a given building, almost no two walls are exactly the same.
To automatically weld these walls, the robot has to be told which one of dozens of wall designs is coming, where the welds are to be located and what path to take when welding it. That's what the CAD model is good for.
The robot also has to find all the weld seams in real life, since neither the robot nor the wall construction is all that accurate, at least not accurate enough to find the weld locations to within the weld tolerance. So, we use the virtual location off the CAD model to get us in the ballpark. Then a laser, which is mounted to the robot head, finds the exact location of the weld seam (kind of like eyes). The program then sends the robot to the corrected location, so it can lay down a weld bead in the exact spot it needs to.
Needless to say, none of this works perfectly right out of the box. To say the least. :lol:
CAD to path can be demonstrated by the simulation shown below.
What you see above was basically a CAD drawing run through simulation software. Those numbers are then crunched and fed directly into an actual robotic system.
Currently, one big headache is applying the above to what you see below.
What you see in the pic is a prefab steel stud wall, used in commercial building construction. The future is to automate this process by using a series of welding robots instead of 5 guys. The problem is that, in a given building, almost no two walls are exactly the same.
To automatically weld these walls, the robot has to be told which one of dozens of wall designs is coming, where the welds are to be located and what path to take when welding it. That's what the CAD model is good for.
The robot also has to find all the weld seams in real life, since neither the robot nor the wall construction is all that accurate, at least not accurate enough to find the weld locations to within the weld tolerance. So, we use the virtual location off the CAD model to get us in the ballpark. Then a laser, which is mounted to the robot head, finds the exact location of the weld seam (kind of like eyes). The program then sends the robot to the corrected location, so it can lay down a weld bead in the exact spot it needs to.
Needless to say, none of this works perfectly right out of the box. To say the least. :lol: