I'll admit this thought isn't fully fleshed out, but it might be worth getting dialog going. CAESAR runs into convergence issues with cross supports (two trunnions attached 180 degrees apart from each other, attached to vertical pipe, may or may not have spring cans).

Let's take the case without spring cans. let's assume our pipe is in the Y axis and our trunnions are in the X. Let's assume our supports are merely +Y, with CAESAR's adamantine level of stiffness.

Let's consider the dead weight case. If we eliminated the trunnions, but wanted to replicate the movement of the vertical pipe, this would be the same as saying we have a +Y acting axial to the main pipe, with the stiffness value equivalent to the weight of the construct divided by the distance travelled due to sag.

Let's consider a case where load acts in the Z direction at the top of the pipe. If the trunnions' contact surface with the support is round, there is no resistance offered at the support to rotation (though friction effects still exist). If the contact surface is flat on flat, then there will be a rotational resistance, up to a point, and then it'll act as a flat on round as the load permits the flat surface to rotate off. In this effect, the trunnions act as an RZ2. Ideally, though, you should not have trunnions and shoes rotating off their supports, and thus the transition point for the RZ2 "should" remain out of reach for a good piping model, and ideally, can be simplified to an RZ with resistance equal to the axial rotational resistance of the two trunnions.

Let's consider a case where load acts in the X direction at the top of the pipe. The trunnions contact surface should be irrelevant; this will always be an RX2. However, like the successful design described above, the transition point should ideally be higher than that of the final product, and therefore ideally, should be an RX with resistance equal to the bending resistance of one of the trunnions if they are +Y supports, and double if they are Y supports.

If the trunnions are guided, this is the same as applying an RY restraint on the main line.

It looks like we could add a new support type that, for purposes of estimating pipe movements, acts as a single-point restraint. For calculating stresses, CAESAR could create a 4 piece model on the side that applies displacements and loads from the main model. I would be an advocate for the stress analyst still providing SIFs for such a connection, though stiffness values is fully within CAESAR's present calculation capabilities, and might be handled in a manner similar to CAESAR's obligatory HGR cases.

I note that this really only works for symmetric trunnion models, though. There is presently no option in CAESAR to have direction-specific resistance (i.e. X lb/in in the positive direction and Y lb/in in the negative direction).

I also note that having 1 non-linear restraints at 2 locations (total of 2) is not necessarily better than 2 non-linear restraints at 1 location (total of 2). My gut feeling feels as though it'll be more stable, and may be worth testing.

Thoughts?