Hi All,
I'm running a pump+network model and got a werid result. I built and calibrated the pump model and network model seperately, all both ran well. And the weird result happened after I combined the pump and network together.
For the pump, it has five pumps with its operation data from SCADA system. Since the pump has "ON & OFF" status depanding on the time, I used the VSP + pump pattern to present different pump operation status (as below) and the result in the second pic. Shown in the result, the calculatoin fails somehow, but will need some guidance on how to check the model.
Thanks,
Meng
Hello Meng,
Do you see any user notification for this? From the second image I can make out that the pressure is zero after the pump. Is my assumption correct?
Is there flow in the pumps? Have you checked if the pipes are active / open, or, if the pumps are on?
We would need to take a look at the model files to see what is happening here.
Please share the model files, following the steps provided in the below article;
Sharing Hydraulic Model Files on the OpenFlows | Hydraulics & Hydrology Forum
Let me know when you share the model files.
Regards,
Yashodhan Joshi
Thanks for the reply, Yashodhan.
There are no flow in the pumps during the fail calculations. And yes, all necessary pipes and components are active. The pump ON/OFF status shows in the pump pattern.
Also, I've attached the model thro the confidential link. The file name is "weird model" and the address here https://communities.bentley.com/p/secure_file_upload.
Eager to learn what went wrong.
Hi Meng,
I've taken a look at the model. The main reason why the calculated results do not match the assigned patterns is because your calculation options are set to use a constant, one hour reporting time step. So, although the calculation timestep is five minutes, the results were only showing datapoints at every hour. Change the reporting time step to "all" (see more here).
The other reason for mismatch is that the big negative pressure results during unbalanced timesteps throws off the graph scale. The unbalanced times are caused by a few periods of time where multiple pumps are switching on and off (1.0 and 0.0 multipliers) at the same time. Pumps 2 and 5 flip-flop at 305 minutes, pumps 1 and 2 at 410 minutes and the change at pump 1 at 1060 minutes coincides with other changes) Try making the following timing changes to the pump patterns:
"#1" - For row 212 (1060.02 minutes) change the multiplier to 0.98
"#2变频"
"#5变频" - For row 61 (304.98 minutes) change the time from start to 305.00 minutes
With this, the model balances at all timesteps for a 24 hour simulation and the results appear reasonable. More guidance on model stability can be found here: Troubleshooting the Network Unbalanced or Cannot solve network hydraulic equations user notification
Jesse DringoliTechnical Support Manager, OpenFlowsBentley Communities Site AdministratorBentley Systems, Inc.
Answer Verified By: MENG ZHANG
Hi Jesse,
Happy New Year!
Thanks for the quick and detailed answer. I'm having trouble to follow when you suggest "For row 61 (304.98 minutes) change the time from start to 305.00 minutes". Can you be more specific what to change?
Best,
MENG ZHANG said:Thanks for the quick and detailed answer. I'm having trouble to follow when you suggest "For row 61 (304.98 minutes) change the time from start to 305.00 minutes". Can you be more specific what to change?
Here, Jesse is suggesting to make the changes in the patterns used for your VSP pumps. See the screenshots below of the said times where changes are proposed;
Here instead of time 304.98 try using 305.00 mins
Here for 405 mins try using a multiplier or 0.89
As suggested the model balances with the above inputs at all timesteps.
Thanks Yashodhan, now I'm able to get some results. Still, I'm not pretty clear why these changes can affact the result so hard, and have the following two questions.
1) I understand the inconverge was caused due to pumps switch at certain step. But I saw another few time step when pump flip flaps like in 50min, 780min and other time step (as below, the highlight rows). So how to determine which is the wrong one, and what is your strategy to find the wrong switch needs to change?
2) Regarding the time in Row 61 (305min), I have a huge confusion. It is true when I changed from 304.98 to 305.00min, the model was convenge. But why this specific time step matters (not the others)? And in a general way, will this error be avoid by using integer minute (not demical) for the pump pattern time?
Thanks for your answer in advance!
Jesse Dringoli said:The unbalanced times are caused by a few periods of time where multiple pumps are switching on and off (1.0 and 0.0 multipliers) at the same time. Pumps 2 and 5 flip-flop at 305 minutes, pumps 1 and 2 at 410 minutes and the change at pump 1 at 1060 minutes coincides with other changes)
These times were identified when there are multiple changes happening at the same time-step. As Jesse mentioned, both pumps 2 and 5 flip-flop, at the same time i.e. 305 mins. Such instances are more critical when multiple elements are changing status at the same time-step leading to non-convergence
You can take a look at the Calculation Summary for the time-steps where the network is unbalanced or not converging and take a look at the intra trial messages to understand what is happening.
Thanks Yashodhan, the question has been resolved.