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

Pressure Subnetwork Accuracy

How does changing the pressure subnetwork accuracy affect overall model results?   I know it is the model convergence criteria set point.  

Specifically,  I am in the process of evaluating piping diameters for a force main replacement project.  Currently my calculation options for pressure hydraulics are as follows.  Outside of the replacement pipe limits, all other model parameters are the same.  

Pressure Subnetwork Accuracy: 0.005  (i know the default is 0.001 but changed to 0.005 based on previous help received from Bentley.)

Pressure Subnetwork Trials: 100

Option 1 (10" DIP - ID of 10.2 after cement liner) -  convergence was achieved with a relative flow change of 0.00498 (just under the limit)

Option 2 (10" DR 17 IPS - ID of 9.41) - convergence was not achieved.  Pressure summary has a relative flow change of 0.00887.  

Does this mean option 2 is not a valid option.  If i upped the accuracy  limit to 0.009 it would show convergence.  But this seems like it is bad modeling practice.  but i do not know how it actually impacts the quality of the model?  Is there an upper accuracy limit to stay under?

Ignoring the scenario not converging, the pump run times and pipe flows and velocities are comparable.  

Thanks, 

Scott

  • Hello Scott,

    There aren't really any set rules for determining what a good convergence accuracy is, but the default should be well within acceptable tolerances. You can find a bit more information on Accuracy in the Help and in our wiki such as this article:

    communities.bentley.com/.../12475.what-is-the-purpose-of-the-accuracy-calculation-option-solution-500000079151

    In your example where you needed to change the accuracy to 0.009, convergence was achieved because the relative flow change between trials was within the tolerance you set, so the program considers it balanced. Since the convergence relative flow change of 0.00887 is higher than the default accuracy of 0.001, it means that the converged results are slightly less accurate than the default level of accuracy. It doesn't necessarily mean that the results are wrong or invalid, but I generally agree that it's not the best practice to change the tolerance (and settle with a less accurate result) unless under challenging circumstances. The preferred approach would be to investigate why the model could not balance within the default accuracy in the default max trials. There might be a problem with the data input somewhere that could be fixed, rather than compensating with calculation option tweaks. Though, your analysis of the results where you found that the run times, flows etc were not much different between the two cases was a good sensitivity check that might indicate that the accuracy difference is still small enough that you may not need to worry about it.

    That said, if you need further help, it would be best to see a copy of the model. There are two options for sharing your model on Communities. Either way, be sure to zip your files first. The first option is to attach to your reply on the forum using the Advanced Reply Editor (see link below and to the right of the reply box). If your data is confidential, use the instructions in the link below to send it via Bentley Sharefile. Files uploaded to Sharefile can only be viewed by Bentley.

    communities.bentley.com/.../7079.be-communities-secure-file-upload


    Regards,

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

    Answer Verified By: Scott Kampa 

  • Jesse - Thank you for this information. I have made several adjustments to my model. I was able to lower the Pressure Subnetwork Accuracy back to 0.001 and still obtain convergence.