Strain Life fatigue method for welds 1

[cpjcoe]

I am a FE analyst and have been researching fatigue software packages to add to our capabilities. It seems that when evaluating welds, the Stress Life method is always used, even if it is a low cycle life situation. I have also come across an article that states that the Strain Life method is not appropriate for welds.

Can anyone explain why Strain Life method is not used for welds?

 

[ncodeuk]

The strain life method assumes a homogeneous material and a weld is not homogeneous.

Fatigue of welds is generally dominated by crack growth - starting from small cracks, or possibly visible ones.
SN curves of welds (such as BS7608, or component specific ones) are essentially component SN curves, where the stress is a nominal or reference stress and the local response is not considered in detail. It assumes that the fatigue response is the same as that of the original SN specimens, when subject to the same loads etc.

The SN method assumes that the problem is generally elastic, although it may embody a degree of plasticity from the original tests. It is also worth remembering that a welded joint will probably already include significant stresses anyway - perhaps plastic ones at that.

 

[cpjcoe]

ncodeuk,

Thanks for the reply. It helped verify a response I received from a fatigue software engineer.

I'm guessing by your name that you might work for nCode. One of the packages I am looking at is Design Life with the Seam Weld module.

The problem I am seeing down the road is that I am supposed to correlate with our physical tests. Although they use nCode software, they lay strain gages directly on the weld toe and use the EN engine to predict life. I am not sure where they get the material specifications.

So correlating using fatigue software for welds will be difficult.

 

[ncodeuk]

You guessed right.

If you want to correlate test and CAE it is still useful to compare the local stress response (in FE) with that from some strain gauges in front of the weld. (static is easier to start with).

If you are getting the right stress/strains, then you stand a good chance of agreeing on fatigue lives. If you do not agree on the stress levels you need to look at your model and in particular compare your boundary conditions with those of the test rig.

As to fatigue properties, some data is available, but it pays to measure specimens from your production source. The weld geometry and welding process will have far more influence on the fatigue performance than the original parent metal.

 

[cpjcoe]

ncodeuk,

Our physical testing method has been in place for 25 years and is unconventional by today's standards. A few years ago there was an attempt to transition to BS7608 and I believe we were able to correlate strain pretty well with physical test measurements. This would be with the gage measured away from the weld. But this was abandoned and we are still using the practice of laying strain gages on the weld toe. We do not correlate well in this case.

This is briefly the physical test method I need to correlate with:

1) Strain gage located in a known radius ground at the weld toe. Radius is then assigned a notch factor.

2) Using time history data from a test course, fatigue lives are predicted using:

a. EN fatigue engine (nCode)
B. EN material properties for the weld (not parent material and I don't know the origin of this data)

After using Seam Weld and fe-safe Verity it is difficult to imagine successfully correlating with our physical test methods. This is because:

1. The CAE software uses SN curves for calculating fatigue life.
2. They calculate a consistant mesh insensitive structural stress at the weld (a good thing). But we cannot correlate with strain gages in the weld toe (not good).

My original question is trying to address number 1 directly above. Should our physical test procedure be using SN curves for welds? Or can the SN CAE/EN physical methods coexist?

Thanks. I really appreciate your input.

 

[ncodeuk]

cpjcoe,

I am happy to continue an impartial technical discussion on the subject of the basic principles of doing fatigue analysis of welds.

As we respect the forum's protocol I dont feel it is appropriate to discuss the relative merits or approaches of one fatigue analysis software product with another on this forum.

If you dont mind I've contacted you directly to discuss (it seems you are a registered contact of ours so feel this is appropriate)

I hope you dont mind this.