Connector Damage/Failure

[jball1]

I have a model that includes several bolted joints. In this model, I have modeled the fasteners with connector elements which have the three translational stiffnesses defined (two shear, one axial).

This is a transient analysis. I am predicting failure in a minority of the bolts in two of the connections (~20% of one connection and ~10% of another). I have never done this before, but I would like to include in my simulation failure of the connectors when the fasteners reach a particular von Mises stress, so that I can see how the force gets redistributed after fasteners fail (does the whole connection fail, or can it handle the failure of a few bolts?).

I have never done this before, and so I am reading the documentation. It seems like including connector failure based upon a simple force or moment limit is straightforward. It also seems like coupling the load in two or more direction is possible. However, I’ll be honest, I am not understanding all of the menus, and the documentation seems lacking to me (or maybe I’m just too dense!).

It seems like I have to use Behavior = Damage, not Behavior = Failure if I want to include coupling. Then I need to add “Initiation Potential”, and then “Specify Derived Component”, but then this is where it doesn’t seem like it is going to work. I need to be able to do the square root of the sum of the squares to calculate the overall shear force, and then I need to calculate the von Mises stress. It doesn’t seem like I can do one operation inside of another like that.

Anyways… if anyone has done this before, or has any thoughts on whether or not this is even feasible, please let me know. From sitting thru Simulia trainings, I was aware that connector failure was a capability that Abaqus has, and that it can account for fairly complex behavior (e.g. plasticity). I would have thought a von Mises failure criterion would be fairly easy to account for, but maybe not?

 

[FEA way]

Damage behavior for connectors is simplified and you can define the initiation only in terms of forces/moments, equivalent plastic displacement/rotation or relative constitutive displacement/rotation in the connector. Maybe in your case it would be better to replace the connectors with beam elements and use general-purpose damage and failure criteria. This thread can be helpful in such a case too:

https://www.eng-tips.com/viewthread.cfm?qid=462351

 

[SWComposites]

How do you get VM stress for a connector element?
Why would you use VM stress to predict fastener failure? Fasteners typically fail in shear (in properly designed joints).

 

[jball1]

FEA way, thank you, I was thinking the same thing (beam elements). I would have to make some different modeling choices elsewhere in my model to support beam elements, but it looks like I may have to do so. Without going into too much detail, this is a fluid structure interaction problem, and so modeling the flanges with shells is preferable to solids (complicates the wetted surface). But it can be done.

SWComposites, my connector elements have the three translational stiffnesses defined (one axial and two shear). I am using the resulting forces to calculate a von Mises stress with a Matlab script. I was hoping that I could create a failure metric in Abaqus using a combination of the loads. It sounds like this isn't possible.

Regarding failure of fasteners, I am not trying to start an argument - I'd be happy to learn something here - but are you saying that axial load on a fastener never contributes to failure of the fastener? The joints that I analyze are almost always subjected to a combination of tensile loads, prying loads, and shear loads. It is true that the shear load is often what contributes most significantly to the predicted failure, but we don't ignore the contribution of axial loads.

 

[jball1]

This thread relates to the consideration of combined loading on fasteners:

https://www.eng-tips.com/viewthread.cfm?qid=265633

It'd also probably be helpful to add that this is a severe load case where preload is overcome. The tensile force that is included is due to externally applied loads, not due to preload.

 

[SWComposites]

Tensile failures in fasteners are usually at the shank/head intersection (root radius stress concentration) or at the nut/thread interface, neither of which are at the location of maximum shear stress. And I'm not in favor of predicting fastener failures by analysis; there should be test data if fasteners are the critical part of a bolted joint design (at least in aerospace we typically do not want fasteners being the critical failure mode, as that results in a joint fracture rather than redistribution of loads within the joint).

 

[jball1]

You probably know more about this than me, but it also might be the case that our industries differ in our approach to assessing fasteners, simply because we are focused on different loading scenarios.

In aerospace my assumption is that you are much more focused on fastener failure due to fatigue, whereas in my industry, we consider fatigue, but find that our designs are more often controlled by significantly more severe, short duration load cases that we are required to consider.

When predicting fastener stresses due to fatigue, stress concentrations matter. Bending stresses should be included as well. When predicting fastener stresses due to one-time, severe loads that produce stresses that overcome preload, I think average stresses across the fastener cross-section are the stresses that are of interest. Localized, outer fiber stresses may cause damage, but won't result in failure. The axial stress can be calculated as the predicted axial load divided by the minimum cross-sectional area (section at minor diameter), and the shear stress can be calculated based upon the shear area at the shear plane.

I agree that it is good, whenever possible, to avoid fasteners being the critical failure mode. Sometimes in my industry that is simply unavoidable. We do a significant amount of testing, but depending on the component, tests can be 10's of millions of dollars. The particular component I am analyzing would probably cost closer to 50 million to test. It is much preferable to convince ourselves via analysis that the design is conservative. My understanding is that our approach to assessing fasteners has been demonstrated by pretty extensive in-house testing to be conservative, if not strictly accurate.