# Effect of Spline Coupling on Rotordynamics

Ok, so we're running a compressor rig powered by an air turbine that keeps breaking. I've been asked to review the rotordynamics for the entire system. Prior to my employment, the compressor shaft itself was evaluated, however they want the effect of the turbine and the spline couplings to be included in the model.

My question is if anyone has any experience in this, or even a paper on the topic. I googled and found what seems to be a decent paper titled
Effect of coulomb spline on rotor dynamic response
Which SHOULD be available on the NASA technical report server, but the entire site is down.

I also found a paper titled
DDM Rotordynamic Design Sensitivity Analysis of an APU Turbogenerator Having a Spline Connection
but unfortunately that paper didn't indicate (probably purposely) how the actual connection was modeled.

So, I have two approaches. The first is a Transfer-Matrix method using a program such as Dyrobes (XLRotor, etc). This assumes that everything is a cone/cylinder and computes everything analytically. In that case, I will need to define a "bearing" stiffness/damping value to connect the shafts.

The second approach is finite element using ANSYS. In this case I would connect the two shafts using....? I'm guess it would have to be contact elements. I would assume that coupling equations would be "too stiff". Either way, I'm sure that a damping value would need to somehow be input.

So, I will stop rambling thanks for the help.

FredGarvin

What kind of spline? What are the L/D for the coupler to engagement length?

The spline is an involute with 12 teeth with a +0.0002" profile error.

Using the pitch diameter, the L/D for the compressor coupling is about 1.67.

One the air turbine side it's a little trickier because all I have is a 2D layout, no model, no actual drawings. From what I can tell though, it seems closer to 1.0, possibly around 1.16.

FredGarvin
I am not one who gets to do rotordynamics, so I can't help you on that end other than what I have learned from talking with the guys here who do it who take the first method you mentioned as the analytical approach.

What I can do, if desired, is share my design experience in a lot of compressor rigs and splines. I have had the displeasure of living through a catastrophic failure that was attributed to a bad spline design.

I can tell you right now that, from where I am sitting, the L/D is too big on the one end. I do my darnedest to keep it <1.

Can you go into any more detail on the failure or the design specifics of the coupler? Are they straight or crowned splines? Are the two ends approximately the same size?

I've been comparing the two rotordynamic approaches recently, and I've started definately leaning towards FEA, for two main reasons.

1 - Making changes to a Dyrobes model is a PITA. Since each little section is a cylinder, if you make one small change, then you pretty much have to redo the entire thing. With ANSYS, if I make my input file decent, then all I need to do is define my components and hit start. Solution time is a little longer, but iteration time is orders of magnitudes shorter.

2 - Uncertainty of Solution. When I use Dyrobes, I know that the program doesn't care what the shaft looks like. All it cares is that substation i, has mass: x, I: y, and substation i+1..... In other words, it doesn't care if there are "weak" sections in the shaft that occur from either holes or whatever. I recently came across an issue where we had a shaft where there was a very small piece of material holding together two shafts. Analytically the moments of inertia were simliar, so that approach didn't pick up on it. ANSYS did though, and you could see some massive strain in that section.

But I digress. Basically the failure has been in the air turbine each time. It has been an array of problems ranging from bearing issues (theirs not ours) to clearance problems, so on and so on.

We're just trying to see if there is anything going on dynamically that we didn't predict the first time. As far as the coupler..let me see if I can find a drawing.

OK, I don't think I know exactly what you're referring to with straight vs crowning, however if crowning in splines means the same thing as in gears, then yes, they're crowned, by what appears to be 30°.

And yes, the two ends 'are' the same size. The coupler has one internal and one external spline.

Thanks Fred,