Static weld pull strength correlation to fatigue failure 1

[dgallup]

We have a line of products that have been in production for a quarter century. There is a laser weld between a tubular body of low carbon steel and a 304 stainless steel nozzle. There is a pull out strength requirement on the weld. This has worked well for production setup because they can take a part off the welder and pull test it with immediate results. We used to have weld penetration requirement but that requires skilled personnel to section, polish, etch and evaluate. There has never been a weld field failure in millions of units produced.

I'm working on developing a new application with a higher system pressure and some what higher alternating loads than our current production. In the static case the weld pull strength requirement is 24 times my maximum load. The alternating loads are are no more than +/-25% of the static load. I really did not expect a fatigue failure but have had one part that developed a circumferential weld crack about 60 degrees around after 500 million operation cycles. Parts produced at the same time pass the pull test.

I have been asked to develop a new weld specification for this new product. I can increase the pull out strength requirement but I need to be able to correlate it to improved fatigue strength. I need to have a test method that can produce a fatigue failure relatively quickly (500 million cycles of endurance testing takes nearly 60 days at 100 Hz with time out for inspections). I'm having no luck finding someone local that can do an ASTM E466 fatigue test. How safe an assumption is it that if I increase the static strength I have improved the fatigue life?

If I

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[dgallup]

Here is a typical weld.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[Tmoose]

Was a detailed analysis done of the failed part? It would be a shame to redesign an acceptable part because of a weld process deficiency or glitch.
Is the service load pure tensile loading? Does every operation cycle create full load?
Is every part pull tested?
Does the pull test approach the yield strength of the material in the failed area? If the pull test exceeded the yield a little bit, upon release the part might start with compressive stresses in the area, for a nice boost in fatigue strength similar to shot peening.

 

[dgallup]

We sectioned the weld on the failed part and it is not uniform around the circumference. The crack is in the area with the least amount of the weld pool in the low carbon steel outer body. There will always be an inherent weld pool variation because of part run out and height variation.

The axial static load is caused by internal pressure which is nearly constant. There is an axial cyclical component as the device runs. All the stress in the weld is tensile.

Only setup parts and periodic samples are pull tested. We consider it a destructive test. The current pull tester maxes out at 125% of our requirement. Most parts survive the pull test but we discard them non the less. One of the things I'm going to have to do is get a pull tester with a higher capacity so I can quantify process improvements.

My aim is to improve the process with hopefully little or no change to the components.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[CoryPad]

For our laser welded components, the relationship between static strength and fatigue strength at 1 million cycles has not been as uniform as desired for calculation/estimation purposes. For your very high cycle fatigue life, the relationship likely will be imprecise.

You can invest in faster cyclic testing equipment (ultrasonic, look up Very High Cycle Fatigue or gigacycle fatigue). Or, you can use basic force/geometry/stress relationships to improve life (enlarge weld diameter, increase penetration depth).

 

[Guest102023]

Go from basic principles. A little larger weld, eliminate stress raisers. But if you want to do all that testing I recommend you first do what a previous poster recommended, a root cause failure analysis.
Was the loading consistent? Because 500M cycles approximates infinite life.

 

[inline6]

i dont understand what the picture is of or specifically what is involved but in general remove flaws or lower nominal stresses is the way to improve fatigue life.

A bigger weld doesnt always do it especially if the crack starts at the weld toe. So making the weld stronger statically sometimes won't help

 

[israelkk]

No one mentioned fracture mechanics analysis of crack propagation based on a typical initial crack size as the result of a laser weld procedure. The minimum crack size is the minimum crack size the test equipment (X-ray, etc) can detect. I suspect a crack was already there after welding and due to higher local loads it was large enough to propagate.

 

[racookpe1978]

Tough question:

"In the static case the weld pull strength requirement is 24 times my maximum load. The alternating loads are are no more than +/-25% of the static load. I really did not expect a fatigue failure but have had one part that developed a circumferential weld crack about 60 degrees around after 500 million operation cycles. Parts produced at the same time pass the pull test."

Later, you quote: "
Only setup parts and periodic samples are pull tested. We consider it a destructive test. The current pull tester maxes out at 125% of our requirement. Most parts survive the pull test but we discard them non the less. "

So, do you face a problem that literally - happened only one time in 500 million - or "most parts survived" ?

One in 500 million? Frankly, statistics alone will show a failure every now and then at that number of parts. Don't worry about it.

But !!!!!! If you are only seeing "most parts" survive the test, then you need to use a larger weld.

So, test 1000 parts to 100%. Or, test your 1000 parts to 125% expected max stress - for a reasonable margin above nominal. (Check with your boss for whatever $$$ costs he will accept for the test run.) Then, if any more than .1% fail the test, you need to increase your weld size.

 

[dgallup]

Racookpe - We do not 100% test production parts on the pull tester, only at initial weld setup and 1 part per hour of production. These parts are all discarded. If we tested 100% of production we would have no production.

The failure was not one part in 500 million parts. It was one part that had 500 million cycles of normal operation in the field. Normal operating stresses are far below the static pull test but obviously not far enough below.

Israelkk - I can't say for sure there was not a micro crack there before but the only place I have seen a crack right off the welder is at the end of the laser path (very rarely) and this crack was not there. We are making the laser ramp out cycle more gradual just in case.

We have changed the weld aiming to make the weld throat significantly bigger and have located a tester at a local university which should have a high enough capacity to take everything to failure. I'll let you know how the test comes out.


----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.