When we last left our story our daring hero Alex Hollis, through conversations with one of our Synergi Gas users, realized that for long pipes, Synergi Gas’s use of average values for velocity, temperature and compressibility can yield some unexpected results. Read on to see what happened next!
The obligatory example, complete with numbers and stuff
So, here’s your better basic example that borders on the absurd, used to illustrate the point.
Synergi Gas - Hither to Yon results
We have a 175 km long pipeline going from nodes Hither to Yon. Shown here are the Synergi Gas steady-state solution results. The various data are annotated, so there’s no need to reiterate them here in the text, other to say that the 590.55mm is the pipeline inside diameter, and we’ve displayed it to 4 decimal places just to prove we have a sense of humor.
The key is that we’ve modeled it as one single pipe. Note the disparity in upstream and downstream velocities – while neither is terribly concerning, they are pretty widely different. Now, we’re going to run a few more cases, looking at the results (Synergi Gas has a cool tool that lets us easily add nodes to make these next steps easy, so it wasn’t nearly as much work as it might look, in case you were starting to stress over the amount of work I had to do to create this stuff. I appreciate your concerns.) Here are the results:
Synergi Gas - Critical Pipes Results
There are three conclusions we can gather from this:
- Better granularity yields ‘better’ results, because the averages are taken over smaller ranges of pressure, temperature, and the like.
- It didn’t take a whole lot of splitting of the pipe to find that ‘better’ solution. As soon as we went from 1 pipe to 2 our averages started yielding consistent results.
- If you had infinitely short pipes you’d have infinitely more granular data and thus infinitely ‘better’ answers.
Point 3 is impossible, and it’s also not true. That’s what we call a teaser for a potential future entry in this series – why lots and lots of really short pipes can be a problem.
The changes made
It’s really not unreasonable to expect that our clients are considerably brighter that I typically am, and thus will figure out for themselves the self-evidentiary nature of this whole discussion. But if we can do something in Synergi Gas to make things better, we should do so.
Synergi Gas - Knot Insertion Controls
And thus we added a feature called automatic knot creation. A ‘knot’ is a computational increment that is imposed on an individual pipe – adding what you can think of as pseudo-nodes that are used for computations. In Synergi Gas we will base knot creation on high pressure drops, high velocities, and high temperature changes. Under Edit > Model Settings > Advanced, we have the three switches you see here.
When these are selected, Synergi Gas will automatically add knots as it determines are needed (see your Synergi Gas documentation for detailed information as to how we ‘decide’ when to add knots) at each balance. If we go back to our original one pipe model, turn on the knot-creation capability, and balance that one pipe model again, we would find that the solution for node “Yon” shows a pressure of 17.80 bars.
Synergi Gas - Knot Log
Reviewing the problem summary from the Analysis Log, we can see that Synergi Gas added pipe knots twice during the simulation, at the 8th and 13th iterations, and that the result of 17.80 bars agrees quite nicely with the results from the cases shown above.
It is entirely possible that at this point you are saying, “Gee. I wonder if that feature might have been of some benefit to me, and if using it will make a difference in any of my solutions?” Or something like that. Well, if you never knew the feature was there, you might not need to be concerned. The knot creation feature is on by default, so when it was available in Synergi Gas it was turned on automatically. Unless you turned it off, it’s there and on patrol.
Which is probably a good thing.
The curmudgeon’s summary
- Averages matter, and it’s always important to keep that in mind when modeling a steady-state system – these issues are not an issue in any kind of transient analysis.
- Just because we’ve never heard of it doesn’t mean it’s not valid. Indeed, our never having heard of it might just be a strong indication of its validity.
- Great ideas are great ideas. Period. And if you have a technique that you believe gives ‘better’ answers, let us know.
Have an idea for a modification or enhancement to any of our software? Please contact our Software Support team, and share it with us.
Author: Alex Hollis