Fatigue verifications about real component: what is the true results?

[PIZZEGHELLO.P]

Hi everyone, 

i need to perform a fatigue analysis about a component that was broken by fatigue stress. ( real photo figure 1).

figure 1

Material : Aluminium alloy 7075 T6

Fatigue curve: from MIL-HDBK-5J

 The component is connect to other component by four M6.

I have the real data about force, rotational velocity and acceleration about this component, take them from motor and sensors. I have the 3d ipt model about  all real components.

IMPORTANT: all fatigue analisys have the same external force applicated and same materials. MESH 5 mm to 0,5 mm . BOlT is not in my consideration results.

1)  I evaluated my model and force environment with two fatigue analysis:

1.1) one with only the component in figure 1. To simulated the screw i insert structural constraints on the face under head of bolt. For M6 screw is circle about 10mm diameter.

figure 2

1.2) one with the component connect to another. To simulated better, i insert a bolt connection about M6 screws.

figure 3

Results: the results in the two analysis is in agree with each other and is in agree with the real case like figure one. GOOD!!!

I design another type of component to and  with the same force environment and same material i do two new fatigue analysis: 

1) one with only the NEW component like in figure 1. To simulated the screw i insert structural constraints on the face under head of bolt. For M6 screw is circle about 10mm diameter.

figure 3

note: the red colour is 10^5 seconds, se the colour legend.

i  think yes, i'm good designer 🙂 but i do another fatigue analisys:

2.2) one with the component connect to another. To simulated better, i insert a bolt connection about M6 screws.

figure 4

Results:

1. The 2.1 ( figure 3) analysis isn't agree with the 2.2 ( figure 4). 
2. THe first, with a bigger rigidity by structural constraint and more bigger idealization , see that all is OK and fatigue life is infinite. 
3.  The second , with a "real like" bolt connections, see that nothings is good. I expected that there are bigger stress  and minor fatigue life only under bolts head. If you look carefully i edit a fitting to "sweet passage" near bolt's head space, but there aren't good results that like i expected.
4. The force environment is correct, the real case is like the simulated case.

MY BIG QUESTION IS: what is the true simulation, if 2.1 and 2.2 isn't agree with each other?

thanks for the reading and answer.

Pietro from Italy.

[John_Holtz]

Hi Pietro,

What are the stress results? They should be similar in 1.1 and 1.2 (except maybe at the bolt holes, depending on if you included a bolt preload or not). They should be similar in 2.1 and 2.2, but the different in the fatigue life implies that the stress results are not similar. If the stress results are not similar in 2.1 and 2.2, this would imply that the loads are setup differently. (Again, you may need to ignore what happens right at the bolt hole due to the difference between constraints and bolts.)

What version of Inventor Nastran are you using? What are all of the load in the model?

In 1.2, I am a little surprised that the solid fatigue life is 0. That indicates the part breaks one cycle. Is the probe really showing the value on the solid, or is the probe snapping to the beam/bolt? How many cycles did the real part take before it failed?

In 2.2, all we can see is that the solid part fails somewhere between 0 and 4E9 seconds (maybe as high as 8E9 depending on the color of red). Both of those values are "close" to infinite life of 10E10. What is the minimum calculated life for the solid? The value of 0 is the bolt. It has 0 solid life because the beam element used to model the bolt is not a solid. (Now that I think of it, it may not be easy to find the minimum life in the solid since Inventor will not hide the bolts.)

[PIZZEGHELLO.P]

The force setup is the same, applicated at the same surface/distance for all the analisys. I think isn't a variable in this problem.

i'm using NASTRAN 2020. the force is: rotational velocity and rotational acceleration, 1 concentrate mass, 1 variable bearing force. ( i insert 3 tables time-scale force).

yes, the probe is  showing the value on the solid. How many cycles did the real part take before it failed? i don't have the right value, i think and estimate 1,2e+7 seconds.

What is the minimum calculated life for the solid?  i remember it's very low value, in some node give 0, in next to other give 2-3e+7. ( tommorow i attached a new image)

But the questions is why there is so much discrepancy from 2.1 and 2.2.

In 1.1 and 1.2 with the same force as 2.1 and 2.2, the results matches.

[PIZZEGHELLO.P]

Here's the image about 2.2 analysis with indicate solid fatigue lifefigure 5:

[John_Holtz]

Hi @PIZZEGHELLO.P 

You need to look at the stress results for 2.1 and 2.2 to understand what is different in the analyses. (Maybe the displacement results, too.) If the stress results are different, you should be able to determine why those results are different.

If you want to provide the model, please use the video by Roelof Feijen to create a pack and go file.

[PIZZEGHELLO.P]

Here there is a stress results:

Von mises solid stress results of analysis 1.1:

figure 7

Von mises solid stress results of analysis 1.2

figure 7

the two Von mises stress are agree with each other. Not very similar value, but there is a maximum in the same place .

Von mises solid stress results of analysis 2.1

figure 8

These Von mises solid stress position and value is in agree with fatigue analysis 2.1. ( first post).

Von mises solid stress results of analysis 2.2

figure 10

VERY BIG MISTERY: the von mises solid stress results show there is 0 MPA of stress, and in the fatigue results ( faigue analysis 2.2 first post) show me 0 seconds of life. 

[John_Holtz]

Hi.

Sorry, I should have mentioned that the fatigue analysis should output different "subcases" of stress results. It should output one subcase for each applied load. I have not tried rotational force in an analysis, but I would expect there to be at least one set of results for the rotational force, and one set of results for the variable/bearing load. (I do not remember which you said.) I do not know how Nastran decides which subcase is for which load, so it is entirely possible that the first subcase in the first model is for the rotational force, and the first subcase in the last model is for the bearing load.

In other words, if there are multiple subcases of results, you need to determine which subcase belongs to which load. Then compare the correct subcases in each model.

Or, duplicate the analysis and run it as a linear static stress analysis.

But it looks like you are on your way to figuring out why the last analysis (2.2) is giving different results. 🙂

[PIZZEGHELLO.P]

I know there is different "subcase" for each force, they are called DYNAMIC AREA.
In all photos, I print the "total von mises solid stress" just under the subcase point , I think there is the higher stress from the higher summation about external force.