You can't do this with "predefined displacements", since this will "fix" the pipe at that point, preventing further movement in that direction.

You are heading down the path refered to as "load stepping", where the analysis of load case "n" begins at the position where load case "n-1" completed. This is something CAESAR II can't do.

As stated above, you first need to confirm that the pipe will still have a displacment (off of the support) when it cools back down. Then you have two modeling alternatives, neither of which will be 100% correct.

Option A) If the pipe doesn't sit back down, then that section of pipe is probably not stressed, but something someplace else yielded to result in this condition. So this section of pipe could just be initially coded "up off of" the support. The problem with this alternative is that you don't know what yielded or what the state of stress is in other areas of the model.

Option B) You could code the pipe as you have now, on the support. However, at the locations where the pipe lifted off, you could define a vertical "cold spring element" (with an anchor at the top) whose length is equal to the "lift off" amount. Then when you put CS in the load case, CAESAR II will pull the pipe up. The problem with this method is that the CS will stress the pipe, which probably isn't true at that instant.

The issue here is in knowing and defining the state of the pipe after it has been cycling.