Dear sirs
Our customer want us to use shoes + roller bearing supports on some 12" to 20" lines.
Friction factor for these kind of supports are very low, approximately 0.04-0.06, in the axial direction (the direction parallel to the pipe), much higher in lateral direction About 0.33)
This is because the rolling friction is much lower than sliding friction.
Normally, for a +Y support we define a friction that works fine for the x and Z direction.
In these support we would have different friction in X and Z.
So, how can I model a friction that changes this way?

Moreover the customer asked us how would me model the guide effect of the sloping rollers, if exists.

I'm afraid I'm pushing CAESAR beyond its limit.
I would model +Y support with 0.05 friction factor but I don't know how to model the transverse sliding friction.
What is your valuable opinion?

Regards

Here's what I would do.

CAESAR doesn't have Anisotropic friction options, so you can try to fake it.

Step 1: Model all as supports as +Y
Step 2: Run model and note weights for suspended case
Step 3: Replace all +Y with two rotated +Y supports normal to the pipe (we can't tell what angle those rollers are, but let's say 30°). Use 0.025 friction factor since you have one +Y per roller.
Step 4: Augment those supports to add X2 / Z2 "guides", with super high resistance and breakaway force equal to 0.33 x the weight from the SUS case. After the breakaway force is met, set resistance to ~1/10 the breakaway force.
Step 5: Run

For the varying mu's, this might work for you...
Place two +Y restraints on your pipe node but each +Y restraint will have it's own unique CNode. For each of these CNodes, specify a complete zero displacement set (i.e., 0,0,0,0,0,0). Now, for one of these two displacement sets, remove the zero in the Z direction and the other, remove the zero defined in the X direction. The first CNode will resist X displacement and the second will resist Z displacement.
Now, for each +Y restraint, define the appropriate mu. For the first +Y (axial), set mu=(2*0.05) and for the second (lateral), set mu=(2*0.33). So now you ask, why 2 times mu. With two +Y's defined at the same node, each will carry half the load so you have to double the mu to get the correct total friction load.
I don't know what to do if the pipe axis is skewed.

To model sloping rollers, think of your restraint vector (Y, here) as defining a plane of free motion. A Y restraint provides no restraint in the horizontal plane. With the sloped rollers, each has it's own plane of free motion. Change your restraint vectors to match these new planes using a skew definition such as +Y(0,1,0.1) and +Y(0,1,-0.1).

I have not considered how your two concerns here interact.

CHECK YOUR RESULTS!

Dear sirs, thank you very much for your answers.

I made some test by modelling the same piping with different support models.
The first good new is that Dave's approach works very fine to simulate different frictions in different directions, but it's difficult to use for skewed pipes.
In a second step I modeled sloping rollers with double friction, and I had several no-convergence issues, but I eventually managed to get some results.
In my model (I'm modeling an existing line) the effect of sloping rollers is much higher than the effect of different frictions in different directions. That's why I'll model the sloped rollers, with only axial friction applied.
Thank you again for you valuable opinion.