This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

HGL spike at manhole

I got a no sense spike for the HGL starting at manhole (MH-2-8-1-2-R6) as shown in the screenshot below and the shared Model, that due to the flooding which no expectation to have any flooding there at scenario (FEH-5YR 120MIN), because the other scenario (FEH-10YR 120MIN) with with higher rain event does not have any issue, the physical and the alternative checked and compared without any error.

Regards,

Mohamad Azzam

Parents
  • Hello Mohamad,

    Try graphing the hydraulic grade at these elements to see if there is a momentary spike that can be attributed to hydraulic instability. If so, check the model for input error (especially around this area), simplify areas that the numerical solver may be struggling with, and adjust the advanced calculation options based on the guidance in these articles (depending on which solver you are using):

    If this does not help, we will need to take a look at your model. You mention "the shared model" but I do not see any new model shared by you today. If the above troubleshooting does not help, please go to File > Save to Package and upload the resulting ZIP file using the method described here: Sharing Hydraulic Model Files on the OpenFlows Forum Then, reply back here with the file name and steps to reproduce.


    Regards,

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

  • Hello Jesse

    The model took time to be uploaded because I upload it with these two scenario running (with the results). kindly now you can have a look in the model. I do not think there is an error of input since the same in 10 years has no issue.

    Regards,

    Mohamad.

  • Did you try the standard troubleshooting steps including calculation option adjustments? Just because results are stable with one storm event does not necessarily mean they will be stable with any other storm. If the system is especially sensitive, a small change in flow can sometimes produce a large difference. With dynamic modeling, refinement of the calculation option parameters is necessary to compensate for hydraulically challenging situations.

    I see the model now and will plan to take a look shortly.


    Regards,

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

  • I have taken a look at the model and have found that the problem is due to initial instability, which is common with dynamic modeling.

    In the statistics section of the results, manhole "MH-2-8-1-2-R4" shows a maximum hydraulic grade of 212 m, time to max HGL is 0.018 hours, and manhole "MH-2-8-1-2-R5" shows a max HGL of 90 m and time to max of 0.021 hour. These "spikes" do not show up on a graph of HGL in/out for these manholes because they only occur for a very short period of time, between the reporting timesteps. (whereas the red "high water mark" in profile considers all calculation timesteps)

    In your calculation options, you have the routing step set to 5 seconds, but the reporting timestep is set to 60 seconds. If you set the reporting timestep to 5 seconds to match the calculation increment (just temporarily for illustrative purposes) you can see the spike in graphs:

    Instability like this can be common in hydraulically challenging models and can be removed by using the guidance in the article previously provided. You can also see evidence of model instability in the Calculation Summary, where the "Non Converging (%)" shows that 66% of the simulation time does not converge. The "report" tab has a section called "Highest Flow Instability Indexes" which can provide further insight

    If for example you reduce the Routing Step from 5 seconds to 1 second, the %non-converged will drop to 10% and the HGL spikes are no longer seen. You may need to further refine the calculation options based on other observations after a careful review of model results. For example changing the "head convergence tolerance" to 0.015 m will further lower the not converged to 1% and the flow continuity error to 4%.


    Regards,

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

    Answer Verified By: Mohamad Azzam 

  • Dear Jesse - Thank you very much for the intensive investigation and clarification provided, that really very helpful.

    The reason we did not go with less than 5 sec for the routing time step because in the article mentioned if the issue is related to the flooding or surcharging it not recommended below 5 sec and better to use higher value, so that why we do not go below 5 sec.

    Jus would like to ask why that is happened in this scenario only since the model has many scenarios and the main difference between all scenarios is the rainfall, all of them are using same routing time steps and reporting time time step, so why the instability did not occur there? On other hand reducing the routing time step to 1 sec will make the model run longer specially in such large model.

    Sincerely,

    Mohamad

  • Every modeling situation is different and the article presents suggestions that may work in some cases while not in others.

    In this particular model my observation is that a smaller routing step seems to be needed to improve stability. With the 5 second Routing Step that you were using, you can see from the Calculation Summary that the model was not stable, and thus unexpected results can occur because the model becomes sensitive. In my experience most complex models will produce better results with a 1 second routing step.

    A thorough review of the model may also reveal if there are any data entry issues or particularly challenging situations that could be addressed or simplified to prevent the model from being sensitive and unstable, thereby potentially enabling you to use the larger Routing Step (and faster performance).


    Regards,

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

    Answer Verified By: Mohamad Azzam 

Reply
  • Every modeling situation is different and the article presents suggestions that may work in some cases while not in others.

    In this particular model my observation is that a smaller routing step seems to be needed to improve stability. With the 5 second Routing Step that you were using, you can see from the Calculation Summary that the model was not stable, and thus unexpected results can occur because the model becomes sensitive. In my experience most complex models will produce better results with a 1 second routing step.

    A thorough review of the model may also reveal if there are any data entry issues or particularly challenging situations that could be addressed or simplified to prevent the model from being sensitive and unstable, thereby potentially enabling you to use the larger Routing Step (and faster performance).


    Regards,

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

    Answer Verified By: Mohamad Azzam 

Children
No Data