Hi pro’s,

For productcontrol our products are cleaned manually by an operator at specific product points/locations. We produce a lot of different products and they are all analyzed in the same installation. When a product is analyzed, a totally different product will be analyzed next. Because of this continuous product variation an universal product fixture for positioning the products on exactly the same manner is almost impossible. That is why the product is always positioned in a different way. During this project I am trying to figure out what the possibilities are to automate this proces.

Therefore my question is:
Is it possible to create a path between two or three productspecific reference points. For example: the operator positions the robotarm (with a positioningtool) against a productspecific start- and endpoint. Because of these two reference points the position of the product can be determined and a standard programmed path will performed between these points?

Thanks in advance,

Bryan

@Bryan that should be possible. What would the path between the points look like?

@Bryan our integration coach can help you with that but you will need to tell us more about the parts and the requirements on positioning accuracy. With our FT sensor it is perfectly possible to create path and have them relative to a different starting point. But to really help you, we need to know more about the process. Could you film it show us ? Along with some tech requirements? 

If you don't want it to be public, you can send your request to support@robotiq.com

Thank you for your fast responses,

First I will try to explain more about the parts that are cleaned. We have to clean a lot of different products for example the oil pan and exhaust coolers in the pictures.

These parts are manually presented in different positions and orientations each and every time. We position them on a “tube frame”  as you can see in this picture:

Some products (with a flat product surface) are able to be placed precisely, these can positioned in the same spot each and every time. But most parts aren’t this easy (for example the exhaust coolers).

When a product is positioned it is cleaned manually by an operator with a control cleaning nozzle. The part needs to be cleaned at specific points, like all the holes and pipes. (See picture)

For an automatic process the robotarm will position the cleaning nozzle relative to the part.

With the teach pendant we want to learn the robot the “cleaning path”. During this path the robot will move the nozzle in front of all holes and pipes (under an angle of approx. 20 degrees). The position accuracy of this process needs to be +/- 1 mm. Because of the different positioning of the part I was thinking should it be possible to manually position the robotarm in front of the first hole and the last hole (for example) to determine the product orientation and then start the cleaning path between these reference points?

I will try to upload a video of the manual process.

Bryan

Hi Bryan, 

This is something that could be done using the plane feature. The way it work is you define a plane using 3 points, Point1 is the origin  Point2 is the Y-direction and Point3 give the orientation. All waypoints or path can then be taught relative to this plane feature. So if the plane move, the path will move accordingly. 
But i don't see this as a viable solution 
1-It is time consuming as you will have to reteach the plane feature for each parts (3 points). 
2-You said you need a +-1mm precision, the precision will be relative to the 3 points that will need to be redefined for each part. 
3-Because the part may be presented in a different orientation, some points might be out of reach for the robot as the relative plane move or rotate. 

I think a better option would be to design custom fixtures. You could have a universal base for your fixture and a 3D printed fixture for each different parts.



 

 

I agree with @David_Gariepy that the plane is an option but probably very tedious and subject to imprecision.  Are the holes that you are cleaning of similar size?  If so, have you thought about using a vision system to locate the hole positions?

@Bryan you just need to teach one plane, you make the waypoints that actually make up the plane variable so that they can be changed based on the actual robot position at the new points.  You would need to teach them in the same sequence each time as that is what determines the x/y/z orientation of the plane.  We use this a lot for vision applications so that I can move the plane around but keep the same relative coordinates from the camera.

Depending on how you set the robot program up the teaching of the three points should be a couple of minutes at worst case.  You would basically put the robot in free drive mode programatically, tell the operator to place the robot at the first point, use a script fucntion to get the current coordiates and assign them to the first plane point, then repeat for the next two points then you are ready to run. 

@Bryan, 
The example show how to use 2 different planes and switch between them for the same moveL.
For you application, you will have to use a plane for each product. You should be able to reteach the plane under 5 minutes, but i'm still concerned about the precision.  If you don't want to create custom fixture for each of your product, you might need to use vision as @matthewd92  suggest or a combination of both, plane + vision. 

@david_Gariepy,
Thank you for your answer.
I will take a closer look at vision systems!

What is the precision you expect, without a vision system?

That's going to come down to how precisely they teach the three waypoints. If they aren't taught well then everything else will be off. Can you use a tool changer on the robot and put on the end of arm a teaching fixture, something that can go into the hole and center up the robot?  If so that would be best as you will remove some variation from the process of teaching. 

If you use use an automatic tool changer the robot could then switch hands back to the cleaning nozzle.

I can't say for sure but think it will be hard to get below 1 mm

+1 for @matthewd92 idea. This would definitely improve the precision.

Thanks for your reactions again,

@matthewd92 i think that is a really good idea to remove the main variation.