I have read some articles in this forum and i think it is a very good source to share knowledge.

I´m trying to do some fatigue estimation of manifolds under internal pressure.

The more i´m reading about fatigue life prediction the more question i have.

1. stresses and strains calculated linear elastically are concentrated at notches. In my model there are intersections of bores and therefore very sharp notches. I have learned that at such singularities there is no prediction of stress allowed or possible.
Do i have to reduce this stresses with something like NEUBER´s rule when plasticity occurs (above yield)and if the stresses are not beyond yield can i thrust them ? or can i just fillet the notch ?

2. the material data i use are stress-cycles curves from uniaxial tension test on smooth specimen and at a stress ratio of R=-1 (fully reversed).
The stress ratio in my model is R=0 and the state of stress is not unaxial, how to handle this ?

3. i have read that there is several material data that can be compared with the stresses / strain in the model. I mean i can compare calculated stress with stress-cycles-curve or with strain-cycles-curve or even campare it with damage parameter curve (Pswt...smith, watson, topper) and so on.
Which one should i use ?

I have been struggling with exactly the same problem for some time and am grateful to this forum for teaching me a few things. You have indicated that you have a sharp corner at the intersection of two bores. In real world there are no abosolute sharp corners. Even if there was a sharp corner to begin with, as soon as the load is applied this corner will yield and develop some form of radius. However for preliminary analysis the first model can be simple and may not have all the fillets. When you look at the results of your preliminary analysis it will tell you if you have any significant stresses near a sharp corner. For such corners I nrmally start by using a fillet radius of 1/32. If you see high stresses as a result, it would indicate that the fillet radius needs to be increased.

However after doing all of the above if you still see stresses higher than yield stress of the material, it is my impression that to get a real answer one has to go for nonlinear analysis using nonlinear material properties.

Neuber's rule is an empirical rule to estimate notch sensitivity of the material. However I have not found enough experimental evidence to know how reliable this rule is. It seems to me a little far fetched that a single formula can simulate the behaviour of whole range of materials and all types and shapes of notches. Is there any body out there who is using Neuber's rule in combination with FEA to evaluate fatigue?

I know that some of this information has been discussed in earlier threads. But some more information will be appreciated.

Before you do any fatigue analysis make sure you have a good life history of the part. This is the toughest part of fatigue analysis. If you want the part to survive 10,000 cycles but it is above yield the part will more than likely break before the 10,000 cycles. When I do fatigue I will use /ballvalve/stress/life analysis if the part is below yield. Stress life is usually used for below about 80 % yield and roughly 10,000 cycles or more. If the part is beyond yield then strain life should be used. To my knowledge of SWT is roughly that it is the same as strain life but allows for a smoother transition between elastic and plastic zones. You will probably want to correct for R=0. Most predictive softwares for fatigue have several options. When you are just starting out with fatigue, the only way to get good results and find the best approach for the corrections is to put strain gages on the part and test it yourself. Also remember that the FEA model is for an ideal situation, it may already have the notch effect in the model but it does not have other corrections such as surface finish or surface treatment. Keep in mind that Max Principle stresses are used most often for fatigue analysis.

MORE NEWS