What is a good tolerance range?

Date 07-21-2012 23:20

Hi OBEWAN!

10% is a norm for range variance for most manual and even Auto welding processes! However, I would ask myself if this amount of tolerance is acceptable on a Laser welding process? I have little to no experience on Laser welding processes so I would not suggest a 'Range' for a process I was unfamiliar with! Likewise, You cannot transpose a 'Range' from a normal welding experience to such a specialised one with 'X' amount of variables (I would imagine)!
Only you can answer your question, is a 10% variance still producing acceptable welds?
Regards

By OBEWAN

Date 07-22-2012 06:38 Edited 07-22-2012 13:34

It is not a matter that will be resolved without testing. They want to know the total process range around the already qualified weld procedures, which means I might have to do hundreds of weld macros at $50 each.

I am hesitant to even give them +/- 10% because our process is so touchy. Every new setting must be proven, and a simple DOE is complicated by the fact that we need a large population just to prove He leak test. Setting "A" could have good macros and 70% yields at He leak test while setting "B" could have good macros and 98% yields at He leak test.

At this point I am interested in what approach I should use instead of what range should be allowed. We don't have any allowance for a different range at present. It is what it is. Locked down in a password protected computer program that never changes unless a process engineer changes it.

And as for our variance producing acceptable welds, I don't know enough about our control yet to know what range it will accept or how tight the control is. We don't even have a beam quality analyzer yet but are looking at buying one. And the prices range from a few thousand dollars to several tens of thousands of dollars.

By MMyers

Date 07-23-2012 16:43

I wouldn't say there's a standard. I've been known to spec +/- 5 amps and +/- 0.1v on highly critical mechanized GTAW welds - and I'm not talking tiny fusion tube welds, that tolerance range is usually 1 to 2 % of the target nominal. We usually do a min/max test and work our way into a window. Once you have a feel for exactly how touchy the procedure is, and what variations in fit up, materials, gas quality, etc to expect, you can probably pin a pretty acceptable window with little effort. I don't see the need to do a full factorial matrix, getting the corners nailed it usually enough.

By Metarinka (****)

Date 07-25-2012 06:22

I would go with DOE, 3x3 and start with min, median, and max. That can give you a reasonable way to put up reasonable tolerances. However I think you're jumping ahead.

If defect rates are an issue, you're right it would take hundreds to thousands of mockups to prove you went from say 98%-99.5% pass rate. That means you can try to do a lot of upfront engineering and science to understand what your process controls should be, Or you can keep very good inspection and production metrics and try to slowly hammer it out in production.

Generally we do the former approach, but neither tends to be cheap or fast admittedly I work in a lab so we are willing to commit the resources to get good base understanding, usually the latter approach is what is done in production shops. I would start looking at all your process controls and trying to get understanding of what's causing your process defects. It's way to early to be trying to hammer down qualification ranges if you are having yield issues. The cost of requalification is pittance in compared to a mountain of scrap or rework. Or alternatively you can leave your WPS with wide open parameter ranges with the understanding that only the WE will be making parameter changes, and then slowly modify and bolt down your spec over time.

Best of luck.

By Metarinka (****)

Date 07-25-2012 06:12

10% is acceptable but it's also very much an arbitrary number. As well all know it's perfectly possible to start min-maxing parameters within 10% and get a bad weld.

Generally what I do is a targeted parameter range based off of some goal and science behind the joint. I usually throw 10% out the window and instead bound it based upon min/max heat input per inch. From there I start to design my secondary parameters. I don't know your product or weld enough to know what sort of tolerance you have 10% can be too loose or too tight.

I run quite a bit of DOE's, A good starting point would be a 3x3 and use Min/Max and Middle. you can probably remove 1 or 2 welds if you know they won't work. $50 isn't that bad for a macro a 3x3 would cost $450 bucks. Perhaps my idea of reasonable prices has been skewed, the mockups I'm working on are in the 6 digits... Please keep paying your taxes :smile:

By OBEWAN

Date 07-25-2012 15:10

The difficult issue with yields is leak rate. We have to pass 10^-7 cc on a helium mass spectrometer.

We can get a 100% yield on ten samples or a 70% yield on 10 samples just based on luck of the draw and all with a good macro. We can have excellent macros and still fail at He leak test.

We are in the process of updating all of our documentation to include every detail - like cross jet flow, exhaust flow, spot focus dia, focal length, etc...And we are creating startup checklists and PM procedures.

But in my Greenbelt training we were only taught to take a few samples at every DOE setting. Like say 2-4 samples. For leak test samples we need to have 10 samples at each setting.

By Tmoney Date 07-25-2012 14:15

What type of laser are you using? Typically I like to use +/- 5% on power and 2% on speed, since all laser processes are automated speed is normally not an issue. These numbers come from ASME sec IX and experience. How are you measuring the power? We use calorimeters, a.k.a. power pucks, and they themselves have an accuracy of +/-5%. So if you give the process +/-10% and the measurement tool is +/-5% you could be as far off as +/- 15%. Also your technique when measuring power could affect your reading, for some lasers. How much power are you using? And are their any other variables that could cause the process to be touchy? (Gap, material, cleanliness, motion system etc).

By OBEWAN

Date 07-25-2012 15:11 Edited 07-26-2012 00:00

We are using a 4000W Trumpf YAG laser. We have the PFO hardware and software that spins the beam with mirrors. One weld is in a rotation with a servo drive motor control from a different vendor. We do not yet have a beam quality analyzer so we don't know our accuracy.

By Tyrone

Date 07-30-2012 11:25

Obewan,
I would play around with the settings before setting up a 2 level DOE. The mid points are most likely "safe" and you are trying to find the High/Low settings.

I have no experience with Lasers, but I would play with the Watts and Travel Speeds to see how far you can push the boundaries (High Watts + Low TS, Low Watts + High TS) and produce good looking welds visually (without macros).

Once you feel comfortable with the high/low settings, you can do the full factorial with macros and leak testing as your outputs. You can calculate your sample size using the standard deviation on historial leak tests for whatever confience level you want. I would also do a confirmation run at the mid point just to see if it is linear. If the result is not what you expect, you may have to do a 3 level DOE.

Sounds like an interesting project!

Tyrone