Hi,
We have modelled a transient scenario, where a variable speed pump is operating at a fixed head, and 12 downstream isolation valves simultaneously closing linearly over 60 seconds, reducing the flow by 95% (one branch remains open passing 5% of the original flow).
We have defined a pump curve, based on an actual pump selection, which is relatively flat at low flows (see below).
If the scenario is run with a circular gate or globe valve characteristics (TCV element using Hammer default characteristics), the results appear reasonable (transient pressure peaks as the flow is shut off). The figure below shows the pressure and flow at the pump discharge with 12 linear (globe) valves closing linearly over 60s, after an initial 3s delay.
However if we run the exactly the same model with ball valve or butterfly valve characteristics, the pressure peak is truncated and the reduction in flow becomes linear. The figure below shows the pressure and flow at the pump discharge with 12 butterfly valves closing over 60s, after an initial 3s delay.
Could you please suggest what we could try to resolve this issue, with the butterfly valves closing.
Thank you
Luke
Hello Luke,
Could you elaborate on what you mean by the pressure being "truncated" and the flow becoming "linear"? From the screenshots you sent, it appears that the closure occurs more gradually in the second case and the ending steady state condition is reached sooner and with less disturbance (transient effects). I am not necessarily seeing any problem based on those screenshots.
Can you provide a copy of the model so we can give you a more specific answer as to why the pressure and flow trends are what you see?
Sharing Hydraulic Model Files on the Haestad Forum
Regards,
Jesse DringoliTechnical Support Manager, OpenFlowsBentley Communities Site AdministratorBentley Systems, Inc.
Hi Jessie,
Thank you for the prompt reply. Both screen shots in my original post use exactly the same model; the only difference is that the TCV ‘valve type’ has been changed from a globe valve to a butterfly valve. We would have expected some difference with the change in valve type, but not the flow changing to a linear rate of reduction at time is at approximately 57.5 s (see mark-up below). The initial velocity through the valves (at the start of the run) is relatively low (~1.5 m/s) so they need to be nearly closed before having a significant impact on the system.
I will upload my model.
Thanks
Luke, the pressure may not be "spiking" in the second case because of the nature of the last bit of closure. In other words, the momentum change at the last bit of closure may not be as fast and thus the resulting change in head may not be as severe. You can see some information on the impact on the effective discharge coefficient at a given position of the valve depending on the valve type, here:
Valve Type field assumptions and use with a TCV
If you need further help, please let us know the name of the file you have uploaded and we'll take a closer look. There could be other things influencing this from elsewhere in the model, which can be observed by animating multiple profile paths in the Transient Results Viewer.
Hi, I have uploaded "Valve closure query Rev A 1Dec2017.zip". This includes a text file which lists the applicable TCV elements. Thank you
Thanks for sending the model.
Can you confirm that the correct model was sent, and that the correct scenario to look at is "Case 1: Base (as tested by MEL)"? When I attempt to compute the transient simulation after computing initial conditions, it fails with "Maximum number of iterations 100 is exceeded in constant-speed pump. Check input data.". In order to achieve a successful model run, I had to set the transient pump type of PMP-2 to "Constant Speed - no curve". This allows the pump to use the built-in four-quadrant curve, as explained here.
I reviewed the model for anything obvious and have the following comments:
1) This model has extremely short pipes. I am not sure I have seen a model with this small of a scale before. You may need to use a much smaller calculation timestep to avoid excessive length adjustments as explained here:
Understanding length/wave speed adjustments and their impact on results
However, timesteps that are extremely small (on the order of 0.001 seconds) can also tend to cause problems and instability. So, you may need to try combining pipes to ensure they are as long as possible.
2) PMP-1 has the transient pump type set to shut down after time delay, but the initial status is off. Also, the pump valve type is set to Control Valve with a zero entered for the delay until valve closure, which would result in the pump valve immediately closing. If this pump is expected to stay off, set the transient pump type to "constant speed".
3) The TCVs in question are missing their Minor Loss Coefficient setting. Because of this, HAMMER is not able to determine the initial relative closure of the valves, and the closure patterns may not work as expected. Since your operating rule pattern has the initial relative closure of 0%, the TCV's "Status (Initial)" should be set to "inactive" (representing the valve fully open), and the "Minor loss coefficient (local)" setting should be set equal to the "headloss coefficient setting (initial)", to represent the loss through the fully open valve.
4) There is some instability due to check valve NRV-PR1 oscillating from open to closed. This check valve is on a pipe with zero initial flow, due to the upstream pump being set to off.
However, even if I try addressing the above issues, I still see the difference you're referring to between butterfly and globe. Although the characteristics of these valve types are different as explained earlier, there are some oddities with the butterfly option that just do not seem right to me at first glance. There is a small "blip" at around 57 seconds that does not seem to correspond to anything else happening in the network.
My guess is that this issue is impacted by the extreme scale of this network and possibly the time step used, but further investigation is needed. I will get back to you shortly on this, but you may want to address some of the above items in the meantime.
Note that if you happen to have your own valve characteristics curve, you can enter that using the "user defined" option, to see if that works better for you.