Philosophically, you want CAESAR to calculate the displacements that will match real life, and can vary from equipment to equipment.

How you achieve that is a personal decision, and can be subject to debate.

In general, square boxes and flat plates are anything but rigid, and you have to make the decision if the box moves the pipe or if the pipe moves the box. Modeling the box as a rigid is somewhere in the middle, which leans towards the box moving the pipe, but will allows pipe loads to be transmitted through the box to the other pipe.

As anubis indicates, the equipment is mounted, and if heated, the equipment attachment locations will move upwards from the ground.

In an ideal world, you would be able to get a rough estimate of the piping loads, run the equipment connections through FEA and estimate stiffnesses for your equipment connections (rather than the 1e12 lb/in anchors in CAESAR), and cycle until you get reasonable agreement between the two.