Tank Filling over time even though controls should prevent

Hello:

I have a Watercad (Connect Edition) model where one of my tanks will fill up to about 95% full, and then continue to slowly fill up over time even though i have a FCV before the tank that should prevent any flow from entering the tank when it reaches the 95% full mark. 

I will submit the model files for your reference after this post is created.

The tank in question is TANK-B. Run the EPS Run scenario, and view a graph of water level of the tank to see results I mentioned.

Thanks,

Jose

Parents
  • Jose,

    We looked at your model again and found that not only does it run for around 2500 hours, the hydraulic grade for Tank B continues to climb (slightly) but the Flow into Tank B remains 0. 

     

    We attributed this to “numerical noise” (or “ghost flow”). This Wiki article explains this further. And to make it even more noticeable, the Accuracy in Calc Options had been changed from .001 to .01 because with the Default settings, the model was Unbalanced.

     

    We also found that using the latest version of the solver (Engine Compatibility Mode) – 2.2.0 rather than 2.00.12– and the Accuracy setting put back to the Default .001, the model converged.

    Please see if this is what you see and if you get good results using the latest version of the solver.

    Thank-you,
    Larry

  • Thanks, Larry. Adjusting these settings does help that issue. 

    However, now I am noticing another odd result. Graphing TANK-C, the level jumps to full as the simulation starts, then gradually decreases over time, but stays above 85% full, even after long time of 2500 hours. It seems to me that this tank should "float" with TANK-A, since they have the same overflow elevation, and thus should have similar fluctuations in water level. TANK-A predictably has a fluctuating water level based on the controls set up. Shouldn't TANK-C also fluctuate similarly? 

  • I'm curious as to what kind of system this is that requires 104 days of model running to understand what is going on in the system?

  • The FCV controlling inflow to TANK-C's is set to close when it is above 6,737.5 ft (98% full), but the FCV is initially open, so a large inflow occurs in the first timestep which is why you see it jump up to 98% full quickly. Once the FCV closes (this FCV could have been modeled as a TCV or even a control on the pipe to open or close), the check valve in parallel with it opens occasionally (graph flow through P1225) as the hydraulic grade on the other side dips slightly below 6,737.5 ft. This is due to the way that it is connected to the larger network south-east of it, and is why TANK-C slightly drains in the first few hours.

    On the other hand, TANK-B's control closes off it's FCV when its HGL is around 6700. It stays around this elevation and does not "float" with TANK-C because these two tanks are not directly connected and because it is closed off from the rest of the system when the pump turns off. Pipe P1080 is set to be closed and pump PMP-TALIMAN-BOOST isolates the tank tanks from each other. This pump is connecting between the two tanks and lifting the HGL of 6700 at TANK-B up to the elevation of 6737.5 at TANK-C, and then turns off, isolating the two networks. After this happens, TANK-C is disconnected from the rest of the system and therefore does not see the occasional "fluctuations" from its check valve opening and closing.

    I agree with Tom that it would also help to have a better understanding of what you are actually trying to model here and why you are running the simulation for so long (or maybe you just did that to exaggerate/demonstrate the numerical noise issue with the 0.01 accuracy setting and the 2.0.12 solver version that Larry mentioned?)


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

  • thanks for the input.

    Although, my question about floating tanks was between Tank-C and Tank-A. I see that Tank-B and Tank-C are isolated from each other when the pump is off. 

    Typically, I would run this scenario for just 72-hours, to see general operation of the tanks and ensure they are cycling as expected. But yes, I set the simulation time much longer to demonstrate the numerical noise issue to try and understand better what was happening. 

    I do want to mention that running the simulations in excess of 100-days is common for me when evaluating Water Age. We typically like to try and run the simulation time as long as needed for water age in a tank to reach its max (when the graph of calculated water age levels out).

  • Apologies, I missed that you were referring to TANK-A.

    For TANK-A, it fills and drains based on the controls as pumps PMP-TALIMAN-BOOST and PMP_E35 turn on and off. 

    For TANK-C, once the control closes it's FCV, it can only drain through the parallel check valve, which remains closed for most of the simulation because the downstream hydraulic grade is almost always higher. The downstream HGL is higher because PMP-TALIMAN-BOOST turns on at times, and when it is not on, TANK-A is at a higher HGL. In both cases they cause the check valve to remain closed.


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

  • Ok, thank you for the input. This answers my question.

    Regards,

    Jose

Reply Children
No Data