Fatigue and crack propagation life 4

[sushi75]

Hello!

I'm a bit puzzled about fatigue and crack propagation.

Given a SN curve, this one will predict for a specific level of stress what number of cycles are required to FAILURE, let's say 10e6 cycles.

Assuming an initial flaw of 1mm, and using Paris' law, a crack propagation analysis by integration would give 10,000 cycles before the crack becomes unstable and failure is obtained.

So the problem is, what is the benefit of the SN curve if it over predicts failure? does that mean that the SN "includes" a crack phase? I don't think so, but I cannot see where the initiation of the crack ends and when the crack propagation from da/dN starts.

Hope the question makes sense, thanks a lot for your help, hope I will see clear soon!

Cheers

 

[israelkk]

As you said yourself, to use fracture toughness you have to assume an initial crack size. In your case you assumed 1 mm crack. In the SN curve calculations there is no known crack in the metal. Many times, to prepare and test for fracture toughness an initial crack is initiated by rotating bending test specimen which is the standard test to produce the SN curve for specific material. The SN curve method is not over predict. It is based on statistical experience with "good" materials manufactured to specific standards such as: SAE, ASTM, AMS, ASME, DIN, etc. If a known crack is already in the material the SN curve method is no longer valid. You may claim that the SN curve life is the life to reach an initial crack in a perfect material. When you have a process of welding for example where you need to test for cracks after welding and your equipment can test for cracks over 1 mm only, then usually you assume a crack size of 1 mm because your equipment will not find it. If your equipment can find 0.01mm cracks then your initial crack for calculations will be 0.01mm.

 

[sushi75]

Thanks for your reply, it helps!

So from your explanation, the SN curve provides the life from "a 0mm crack" to failure, found from test.

However, what we do in crack prop analysis is to assume an existing flaw from NDI capability, and then find how ,any cycles it will take to failure. This means extracting a portion of life from the one given by the SN curve.

Is it the logic behind that?

Cheers!

 

[EdStainless]

This is why we run KIC, KIIC and KIIIC, so that we understand propagation is various modes. All of these tests use pre-cracked samples.
You can work backward to determine critical flaw size, below this size the SN curve predicts life, above and you mover to crack growth mechanisms.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[sushi75]

Ok so we have also da/dN curves which predict crack growth rate, for different samples, and mode I II III.

But, if the SN curve predicts life up to critical flaw size, then the crack becomes unstable and leads immediately to failure, right?

I'm still confused sorry!

I believe the best answer we can give for life prediction is a fatigue life to initiation+crack propagation to failure, right?

Cheers!!

 

[mp87]

The SN curve and Paris curve represent two different design philosophies.

The classical approach assumes a flawless material and bases the calculation on the SN curve, while according to the fracture mechanics approach the material is already cracked before being put into service and its lifetime is given by the number of cycles that are needed for the crack to grow to a critical size, theoretically corresponding to the fracture toughness.

So, to answer your question, the SN curve contains both the initiation and propagation stages. In some research works, an additional curve is added to mark the points at which the crack reaches a certain size of engineering significance (e.g. 0.1 mm) for a certain stress level, therefore corresponding to the initiation stages; but in most of the cases this information is not reported. Here is a schematic representation of what I mean:

Another comment, please take care to the terms you use when dealing with crack propagation. Fracture toughness (KIC, KIIC, or KIIIC) strictly applies only when talking about the final fracture, while during propagation it is more appropriate to refer to the Stress Intensity Factor (which is not a material property). Moreover, always according to fracture mechanics, when a crack already present into the material starts growing, it means that the fatigue crack growth threshold (DthI, DthII, or DthIII), which is a material property, is exceeded.

 

[rb1957]

the S/N curve does not assume the presence of a flaw. It's life includes the short crack growth life ... eg 100,000 cycles fatigue life = 90,000 cycle (crack initiation) + 10,000 cycle (crack growth).

remember you would not operate for the fatigue life, but rather the safe life = fatigue life / 5

another day in paradise, or is paradise one day closer ?

 

[metengr]

rb1957 is correct.

 

[racookpe1978]

And, further to the discussion - ALL are based on the statistically probability of such a length crack at such a location based on ans assumed number of cycles of equal magnitude and timing and duration.

In real life? It could be anything within 2 standard deviations of the assumed cycle number. 10,000 cycles to "failure'?
Could be 6,800. Not a great probability. But possible.
Could be 10,002.
Could be 16,812 cycles.

If you get past 10,000 cycles, the probability of failure gets larger with succeeding cycle. But failure cannot be guaranteed either. Unless you are bending paper clips. 8<)

 

[sushi75]

Hello,

Looks like it becomes a hot topic!

So just to make I got the point, let me recap your answers:

-SN will predict life to failure at a given stress, without an initial flaw. As it's based on tests, a scatter factor shall be applied at the end to account for variations. If required, I guess the design safety factor should be applied.

-If we know the detectable crack size, we can run NASGRO for instance to get the crack propagation life. This is actually a portion of life included in the one given by the SN curve.

Is it correct so far?

So at the end, if we only provide a fatigue life from SN curve, there is a risk of overpredicting it no?

 

[rb1957]

something like !

if you want to tie the two concepts together, the fatigue safe life (factored life) should be close to the crack growth life from a 0.005" initial flaw.

another day in paradise, or is paradise one day closer ?

 

[EdStainless]

Statistics is the only way to handle these issues.

We make some hydraulic tubing, it is tested in impulse fatigue.
The test pressure is 1.5x the max service pressure (43% of min UTS) and the sample is bent to create a more severe condition than is allowed in service. Minimum test life (mean-3S.D.) is 400,000 cycles, and samples come off of test at 3,200,000. So the designer feels that this part under normal service will have an infinite life (>10e10).

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube