Sometimes when I run my model in SewerGEMS, I will get large spikes of flow that seemingly come out of nowhere. This spike is not in the upstream pipe, and it attenuates slowly downstream. The spikes are usually near the outskirts of the model where flows are lower and there are some steeper pipes. Sometimes when I run the model the spikes will go away, without changing anything. My model has a less than 1% continuity error. These spikes are negatively affecting my model calibration. Any advice on how to make them to go away?
Spikes can show that your model is somewhat unstable.
The spikes could be temporary instability which can still happen with low continuity error. You can utilize the time browser to examine the events at that specific moment, potentially linking it to an element either upstream or downstream.
Using the Time Browser
Another suggestion is to analyze your advanced calculation settings and adjust them correctly. I recommend that you read and use the following article to do this:
Troubleshooting unstable SewerGEMS and CivilStorm results using the implicit solver
Another good solution is to try using the other types of solvers available in the software.
Differences between solvers: GVF-Convex vs. GVF-Rational vs. Implicit vs. Explicit (SWMM)
If this does not help, please provide a copy of the model: Sharing Hydraulic Model Files on the OpenFlows
Regards,
Luca BillisTechnical Support Engineer, OpenFlowsBentley Systems, Inc.
Above is a screenshot of that pipe with the spike in flow at the specific timestep. Any tips on troubleshooting the error happening here?
I have already done a lot of troubleshooting to get my advanced calculation settings where my model is most stable. Unfortunately, I cannot use a different solver as I am trying to keep this model consistent with several other models.
here is the flow happening at that timestep as well
Erin, most likely you will need to further adjust the calculation options, or find the source of your instability. If you have already spent a lot of time refining the options to get stable results, most likely there is a deeper root cause that should be fixed instead of adjusting the options. There is some guidance on that in the article Luca provided. It involves a thorough review of the model for data entry issues, which you can do in part by using tools such as Validate, pinpointing areas of instability with the Hydraulic Reviewer, and thoroughly checking all input especially in the area near the instability. It isn't possible to tell for sure just by looking at the graph and small profile.
I did notice your flows are generally pretty small, which can make it difficult for the model to converge at times. 10 GPM is below the 0.03 CFS threshold where the implicit solver tends to struggle. The SWMM solver tends to work better in such cases unless you also have an extensive pressure pipe network/pump stations.
A deeper investigation of the model would be necessary to provide more specific advice.
I also recommend upgrading to the latest version of SewerGEMS (currently 10.04.00.158) if you have not done so already, as we improve the solver stability in each version. See: Downloading OpenFlows Software
Jesse DringoliTechnical Support Manager, OpenFlowsBentley Communities Site AdministratorBentley Systems, Inc.
I have shared my model file through the Bentley Secure File Upload. Thank you!
Hello Emily,
We are still looking at the model. Based on the Implicit solver sometimes having trouble with low flow situations, I have been testing this in the Explicit solver. I set the Routing Time Step to 1 second and SWMM Output Increment to 5 minutes, and the results in the pipe from the screenshot above is much smoother (though the "Not Converging" percentage is relatively high). I am doing some further testing, but it may be worth exploring the Explicit solver for this system, given the low flows.
If you can't use the Explicit solver, you can use the Hydraulic Reviewer to review the volume differential at elements in the model. This is similar to the continuity error in that it is comparing the volume in and out of elements. There are a number of nodes with a large differential. Reviewing these may help identify possible issues with the setup or places you may be able to simplify the system.
Scott
Just an update on this. The Explicit solver appears to have more stable results, though the continuity error is higher. Try using the Explicit solver with a Routing Time Step of 1 second and an output increment of 0.1 hours to see if the results appear more stable. I also tried setting the Surcharge Method to "Slot." This lowered the "Not Converging" percentage, but increased the continuity error. The results still looked relatively stable, at least for the conduits I looked at.
The link provided earlier on troubleshooting the Explicit solver has additional settings you can consider in troubleshooting the model.