Ideas for Weld Repair System

Hey everybody,

I had a question regarding an application my company wants to develop, all our work revolves around industrial laser processing, from technology repair and integration to process development. I handle to programming and integration for all robotic based laser applications.

My question regards a weld repair system, we setup a remote laser welding cell to weld rails, each rail has 80 welds a piece on them, they do all weld repairs manually right now. I was wondering if it’d be possible to design a cell that did all the weld repair for you, using a handheld laser welding gun attached to a robot; the idea would be to figure out which welds need to be repaired based on a cut and etch, from there the operator could punch that into the HMI and the robot goes and repairs those specific welds.

I guess my question would be on the robot side is how would the PLC effectively communicate that to the robot? Could we link the numbers in the HMI to position registers inside the robot and from there the robot could go and repair those specific welds? That would be a lot of position registers if so, I figured I could just use one PR to save the offset; as in once you get to position register 76 offset along the y-axis however many mm to succesfully repair the weld.

Any and all suggestions would be greatly appreciated, as I mentioned this isn’t a current project but one we’re hoping to build on and hopefully be able to offer to customers.

This is a Fanuc robot we’ll be testing it out on, I put it in the general thread though because we might attach it to a staubli or ABB robot as well.

Quote from SkyeFire

PRs... are you planning to use a Fanuc, then?

It sounds doable. As long as each weld is an identical set of motions, you could use a PR to offset to the start point of each weld, then run a subroutine that performs the weld, then moves on to the next point. If the rails are curved, or the welds have unique requirements per weld, it gets trickier.

Yes, it would be a fanuc robot. That’s what I was thinking, I could save each weld as a position register and from there have one other PR as my offset and then as you mentioned run a subroutine that does the actual weld.

Quote from NoBugsOnlyFeatures

I am really intereseted how you would repair a bad weld with a robot? What kind of problems would the machine be repairing? Missplaced welds, porosity, ...? You usualy need to grind the weld out and make a clean weld groove, so Im genuinely curious how you would go about something like that

So this would be a cell meant to repair autogenous laser welds on a lap joint, autogenous welding means there’s no filler metal being added to the part (my apologies if you knew that). With autogenous laser welding, you need there to essentially be zero-gap in the tooling, if you’re welding with gaps in-between the two materials, this can cause concavity— this is when the top surface of the weld tries to escape between a gap in the piece or if the base material is blown through will escape through the bottom before it solidifies. This happens when doing what’s known as “keyhole welding,” which is a specific method of welding for lasers; it’s really interesting if you have the time to read about it. Anyways, to repair autogenous laser welds according to GM standard you use GMAW welding to either fill gaps or replace the top surface of the weld, we’re experimenting with handheld laser welding to do the repairs on a robot; it’s essentially GMAW but just with a laser and we’ve attached it to a robot because as far as we can find, nobody else has done it. I hope this answered your question!

Quote from Rex Tyrran

What I have found easiest in repair programs I have made for missed spot welds is to have the robot go through each weld position and use DI's from the PLC to determine if that position needs welded. If not, it skips the spot instruction and moves on to the next point.

This would actually be used to repair welds from a remote laser welding cell, what I mean by that is based on how the welds were placed and the application we were able to hit 10 welds from one position. Remote laser welding is becoming increasingly popular for its speed, you’re able to weld a lot faster with remote laser welding; so I’ll use the piece I’m talking about in this thread as an example, there’s 80 welds on these rails that get attached to a floor pan— it takes about a minute and a half to hit all 80 welds, this is because depending on your focal length has well as the scanners FOV you can cover a lot of real estate from one spot.

That’s also what makes this repair system so difficult, because I get all 80 welds from 12 positions— so for the repair program I’d have to make a position for each individual weld, which isn’t too bad but it’ll be a big program that’s Forsure. I was thinking of just making PR’s for each position, then I could save a offset as another PR, from there I could just send the robot to whatever weld that needs to be repaired and just have it offset however many mm’s to repair it.

Quote from Rex Tyrran

So due to GM’s autogenous laser welding standard, we can’t particularly just go over a keyhole that fails— it has to specifically be repaired using GMAW or as I mentioned a handheld laser welding gun with a wire feeder. What your talking about seems like a general rework or a process that’s more suitable to fixing missed welds as opposed to welds that fail due to standards.

Do you guys have a remote laser welding application or is it a wobble head? Scanner optics are normally fairly bulky with a big optic so they can have a wide FOV; wobble heads or regular welding heads have a very small opening that the beam comes out of and they can really only weld seams or areas the robot moves the head around.

With robotic remote laser welding, there pretty much two different variations of the process: point & shoot and on-the-fly— if you have laser welding cells, I assume you’re familiar with these terms, but I’ll write a brief description for anybody that doesn’t still. So scanner heads/remote laser welding heads generally have two mirrors attached to galvanometers that sit on both the x and y axis, if you have a 3D head, there’s a mirror attached to the z-axis to help adjust the focal length for a number of reasons— this allows programmers to make different shapes from stitch welds, circles, infinity signs, etc. Due to this functionality we get our two main processes, point and shoot is where to robot comes to a complete halt and hits one weld or multiple welds based on weld location and scanner specs; on-the-fly is when the robot lays down the weld in-motion, it’s essentially an OLP depending on the software you use. The software takes your weld parameters and builds a path program for you with custom acceleration and deceleration values based on a point-of-reference set in the software relative to the TCP of the robot so that each weld is laid down at the right spot. These processes have two totally different I/O structures normally.

To answer your questions about what systems we’re using, they’re IPG YLS-6000 lasers with IPG 2D high-powered scanners and we have the extended I/O interface— while the beam delivery systems are terrible, the IPG generators are some of the best and most cost-friendly lasers on the market. This makes them one of the more common systems now in automotive, even though everything except their laser is garbage—even the way the system interfaces with everything is archaic and nowhere near as practical and user-friendly as Trumpf, RLSK, or Blackbird heads.

My apologies if you have prior knowledge and all I did was tell you things you already knew, in my experience, people either know a fair amount about these systems or nothing at all. I’m a robotic laser applications specialist so that’s all I do, I’ve been wanting to get into it more from the controls side of things.

Quote from Rex Tyrran

I have a passing knowledge of the laser system. I am a decently competent general robotics guy, but welding is not my strong suit. I know enough about the laser systems to be dangerous, but mostly I only ever need to make minor adjustments to the weld location if the need arises. I have only ever created welding programs in practice and have never tested any program I've written.

We use remote laser welding. I know that we also use the YLS 6000, But I believe we use an RLSK 3D scanner. We use the point and shoot method. I honestly didn't know that on the fly was even a thing. I found your post to be quite informative.

If you don’t mind me asking, did Valiant integrate that system? The only reason I ask is Valiant normally uses RLSK for beam-delivery and IPG for beam generation.

Since you guys are doing point & shoot, does the part allow for you to hit multiple welds from one location or do you guys hit welds one at a time? On-the-fly is really cool, especially if you have quality software to build the process on like RLSKstudio; a lot of the time though it can be a lot bigger hassle than anything, most people think because the robot is constantly in motion, that on-the-fly is the best route to go and if we’re speaking in terms of cycle time it definitely is. Customers see that and they automatically want to use that process, I always advise that if we’re still able to meet production and cycle-time with point and shoot than that is easily the best method; this is because of practicality, as I mentioned, building on-the-fly processes’ is essentially building a OLP except it can be a lot more difficult. It’s nowhere near as practical as point and shoot, as you mentioned sometimes you have to make mirror adjustments; with OTF you have to go into the software and figure out when the “Trigger” (the DO that tells the laser to start firing) sends the signal relative to where the TCP is over the fixture based on your reference point. Because of this you normally need either more people to maintain the process, or highly skilled people because not a lot of people know robotics, simulation, lasers, and welding so it can be hard to implement and keep running.