Simulating wheel passage on railway track in abaqus

[supradude001]

Dear all,

I have been working on the simulation of a single train wheel rolling along a segment of rail. I have the contact between wheel and rail set up as surface to node with wheel having the master surface and rail having the slave nodes; and the contact property being frictional in tangential direction and "hard" contact using augmented langrange enforcement method in normal direction.

I rotate the wheel by means of applying both translational and rotational speed to a reference point located at the center of the wheel with the software defined "instantaneous" amplitude (I got the translational speed by multiplying the rotational speed with the radius of wheel). The reference point is connected to the wheel using rigid body constraint. The rolling speed I am trying to achieve is anywhere between 30-60 mph.

However, what I saw from the output database is that there are stress concentration resides on the rail at couple points even if the wheel has passed these points. Moreover, the normal contact force I saw from the history output is much larger than the vertical axle load I applied to the reference point, and there are severe fluctuations to the normal contact force.

I suppose this was because the wheel wasn't dynamically stabilized before reaching its full speed so it vibrates vertically. Then, I fixed the vertical dof to a small value so that the vibrations can hopefully be eliminated while a normal contact force can still be generated based on this vertical deflection value I defined. However, I am stilling seeing squiggly behavior in normal contact force curve which I am not sure what I could possibly attribute it to.

So my questions are:
Why am I getting severe fluctuations in normal contact force?
Am I using the right approach to accelerate the wheel? or there is a better way?

PS: I have also tried to use coupling constraint to connect to reference point to the wheel, but that wouldn't allow me to apply a translational speed at the same time which means the acceleration is limited by the coefficient of friction at the wheel-rail interface.

Any input would be extremely appreciated!!!
Thank you

 

[rickfischer51]

How coarse is your mesh? You need a very fine mesh to avoid significant faceting. Are you using symmetric or asymmetric contact? I think symmetric contact is prefered for this type of analysis. Softer contact stiffness might also help.

Are you comparing stresses to measured data? If so, you need to consider how you have constrained the rail. An actual rail is on a fairly flexible foundation. If you have simply fixed the base of the rail, this may adversely affect your results. Compressive bending stress in the top of the rail should be included in your solution, and if you just fix the rail, you may not get it. In the US, wood ties on 19" centers give a foundation stiffness of about 3000 lb/in/in. if I remember correctly, concrete ties are typically 24" on center and with their better ballast, typically produce a foundation stiffness of about 12000 lb/in/in. See Stresses in Railway Track, by Timoshenko and Langer, or vol 2 of Timoshenko's Strength of Materials book.

 

[supradude001]

Hi Rick,

Thank you for your reply.

I had a complete model with all the fastening systems, crossties, and ballast, but since I was encountering this wheel "jumping" issue, I reduced my model to only a wheel and a rail in order to investigate and resolve this issue.

I have changed the modulus of elasticity of the wheel to a large value which essentially makes it a rigid body and that worked out just fine - no more wheel "jumping"

However, the contact force in the vertical direction extracted using the history output was 500 times greater than the vertical axle load I applied (i.e. I applied 20 kips vertical load but the vertical contact force I am seeing in the dynamic step is around 950 kips). I have a static step before the dynamic one just to stabilize the wheel on rail with the 20 kip axle load applied, and the vertical load I saw from the end of the static step is exactly 20 kips. Therefore, I suppose my contact enforcement method was appropriate and it was because of other issues in the modeling? Just to mention again that I applied a rotational and translational speed to the a reference point (which is tied to the nodes of the whole wheel using rigid body constraint) simultaneously. I don't know if this issue is caused by the way I apply the speed or not, or might be other issues as well.

Thank you,

Austin

PS: I attached the snapshot of my model and the vertical contact force curve. the static step ends at 2 sec.

 

[rickfischer51]

So how is the rail restrained? Again, if you fix the rail, it cannot bend, you will not get bending stresses in the rail, and you overly stiffen the problem. You do not need to model the ties, fastenersm etc, but you need to account for thier response in you boundary conditions. Railway trackis often modeled as a continuous beam on an elastic foundation. A foundation with discrete support points is termed a Winkler foundation, and this model is valid if there are at least three contact boints within a half wave of deflection. This is discussed at length in the Timoshenko references, also in Advanced Mechanics of Materials by Cook and Young, and probably any other good solid mechanics text. Place a spring under the rail that represents each tie.

Another useful reference, if you can find a copy, is Stresses in Railway Track: The Talbot Reports.