Dear Mr. Diehl,

As usual, your answer is very instructive and, no doubt, the described procedure is correct. Probably I would prefer to avoid "If you have the time"… I mean, I consider is OK to model 2 times VAL beyond the first elbow and stop there, but I would consider as mandatory to check that 2 times VAL length run is enough.

To comment your examples, modeling just 2 times VAL and after that stopping the model, it means to have a run with one end- elbow, and other end- a "virtual" cap. This "virtual" cap is not visible in model, but the "cap effect" disturbs the axial strain distribution on a considerable part of run considered in model. So you need to experiment more by changing the length of run between the real elbow and the "virtual cap" end and see the results are sensitive to these trials. Of course, also in this case one must have some concerns (soil parameters- springs), exactly as you’ve described for a model with a run and 2 elbows\legs (first part of your example)

Some companies prefer – as an alternative- to force in a way the model to consider a zero longitudinal strain in the zone where this situation should be. The "imperfect tool" to do this is a fictitious anchor placed there, and I would admit is unaesthetic and questionable by the Contractor/ Consultant when looking in the model…but in the same time, one can keep in mind that is also true no cap is there.

IMO placing a fictitious anchor at 1.5…2 times VAL will help me to be focused on the elbow, with a good chance to be conservative in this approach. Of course, to be sure, I would prefer to consider different length and soil parameters and see the sensitivity of the results, exactly as you suggested.

I cannot say one method is good and one is bad. For computational effort, shortening in a intelligent way the run would be an advantage, but in our days we have a supercomputer on our desk….

Anyway, my opinion is to consider as worst case modeling just a run with the presumed VAL length, stop the model there and doing nothing more; I consider this approach shall lead to nonconservative results for elbow/leg in the end of the run.