Understanding length/wave speed adjustments and their impact on results

Applies To 
Product(s): Bentley HAMMER
Version(s): 10.00.XX.XX, 08.11.XX.XX
Area:  Modeling
Original Author: Jesse Dringoli, Bentley Technical Support Group

Problem

When computing a transient simulation, the following error is often seen:

"WARNING: The wave speed or length approximations deviate excessively from the entered values. Lengthen short pipes and/or subdivide longer pipes."

What does this mean? If adjusting the time step to reduce this adjustment, what effect can that have on the results, and why?

Understanding length or wave speed adjustment

The deviation mentioned here is based on a percentage, which you can see by the units, under Analysis > Transient Time Step Options. In short, a pipe needs to be as long as the pipe wave speed times the calculation time step.

Based on the time step, HAMMER tries to have a wave travel from one end of the pipe to the other end in even multiples of the time step. HAMMER will adjust the length or wave speed of the pipe to allow this to happen. For length, the default adjustment + or - 50% (see screen shot below) is allowed before you get the user notification. Basically, you want to have your pipes as long as reasonably possible. You can also adjust the time step to be smaller.

Or, you could possibly ignore the warning if the results make sense and look correct to you - it`s your judgment. You can certainly make some adjustments, go back to the time steps, click the Update button and see the new max/mean adjustment. You can also change the time step as previously mentioned and check the new max/mean using the same update button as well as choose between adjusting wave speed or length. Typically this is just a trade-off between accuracy and run time. The smaller you make the time step, the better the accuracy (less adjustment will need to be made), but the longer the transient run will take to complete.

To expand on this, let's say that there are two pipes in a model, 10,000 ft and 100 ft. If the wave speed was 2000 feet/second, the time step 1 sec and adjustment set to length, then it will need to adjust the 100 ft pipe by 1900 ft, because the end nodes would have to be 2000 feet apart (2000 ft/sec with 1 sec time step) If you adjust the time step to 0.1 sec, then it will only need to adjust the 100 ft pipe by 100 ft, since the wave would travel 200 ft in 0.1 seconds.

After computing the transient, you can view the max adjustments on a per-pipe basis in the pipe FlexTable by adding the fields "Length Adjustment", "Length Adjustment Percent", or "Wave Speed Adjustment" and "Wave Speed Adjustment Percent" (if adjusting wave speed instead of length.)

The effect of these adjustments is the same effect as if you manually changed the pipe lengths or wave speed. So if the program needed to increase a pipe length, then the transient waves will take longer to travel that distance of pipe than they really would in real life (in the pipe length entered by the user). If the wave speed was increased by the program, then the transient waves would be traveling faster than they really would in real life. Depending on the model/situation, these adjustments may have a negligible impact on the calculated results.

If you are not confident on the effects of this, we recommend that you run the program with the default time step and then with a small time step and examine the results to see if they are similar. We also recommend that you adjust the length and not the wave speed, since inaccuracies can occur especially when the wave speed is reduced by a lot.

Note that you can find documentation related to this by going to the Help documentation, under the topic “Bentley HAMMER V8i Theory and Practice” > “Time Step and Computational Reach Length”.

How the time step effects results due to length or wave speed adjustment

On the subject of the effect that the time step has on the transient response, from the logic explained above, a smaller time step results in a smaller adjustment to either the length or the wave speed (depending on the option you selected under Transient Time Step Options) and therefore in theory you should have more accurate results. The reason is because the more the length or wave speed is adjusted, the more the results may be skewed. The degree to which the results are skewed can be highly dependent on the system and transient event being modeled. For example, if you have a system where not much transient activity is happening around the vicinity of shorter pipes that have relatively large adjustments and the transient response is relatively slow (no vapor pocket collapses or other things that would cause a 'sharp' spike in pressure), then a change to the calculation time step will likely have a relatively small effect on the overall transient envelope (maximum and minimum HGL).

On the other hand, a system may experience a relatively 'unstable' transient response - sharp drops or spikes in pressure, vapor pockets forming, waves reflecting and combining together, etc. In this case, a change to the calculation time step could potentially have a much larger effect on the transient envelope. If you reduce the time step from 0.05 seconds to 0.002 seconds, pipe length adjustments will be reduced, resulting in a shorter overall pipeline length and effecting the way that transient waves reflect and combine together at critical points. It could be that with the 0.05 second time step, the waves are traveling at just the right speed to cause two spikes to combine and form a downsurge that forms a vapor pocket, which later collapses and causes a large upsurge. With the 0.002 second time step, the change in pipe lengths could result in those same waves not combining together in the same way, resulting in a vapor pocket not forming, and lack of subsequent collapse, and therefore a reduction in the maximum pressure. This is just one possible example that illustrates the dynamic and sometimes chaotic nature of a transient event. To check if this is what's happening in your system, consider creating profiles of critical areas, then animate them in the transient results viewer. Pay special attention to the vapor/air pocket volume at the top of the transient profile, and ensure that the "Generate Animation Data?" transient calculation option is set to "True".

If you find that the above is indeed what is happening in your model, then you might interpret this as the model telling you that there is a fair chance that the transient response of this system may result in unstable transient waves and vapor, which you may want to remediate. Or, by looking closely at what is happening in the transient profile animation, this might identify potential data entry errors that are significant to the transient response and instability, such as elevations, or perhaps an unexpected surge emanating from an element with a data entry error.

Anonymous
Recommended
Related