Here's a half-baked idea:

Say you have a short line you're tying into a really long existing header that has expansion loops and the like.

If you wanted to model that header with accuracy, you'd include every support.

If you wanted to subtract some accuracy and add some speed, you might count supports between important features (anchors, loops, etc) and then split into even increments.

If you wanted to speed things up further, you might replace the previous step with a uniform load equal to the weight of the pipe, effectively making it weightless, and requiring far fewer nodes - just at anchors and expansion joints.

What issues would you speculate you would encounter with either of these methods?

Do you have alternative methods for estimating displacements of header tie-in locations?