Software version downgrade

Hello Jan, 

Could you please elaborate about excessive floods seen in results of update 3? Do you mean that you ran the model which was last opened and computed in update 2 with no changes made in update 3 and you still you see difference in the results? 

Are there any other user notifications , high continuity error in the model after computing?

You can be provided with download link of previous version, could you please let us know exact version of Connect edition update 2 of CivilStorm? 

We can then provide you download link of old version via service request. 

However please note that once the project / model is saved in higher version, the same model cannot be directly opened in the lower version,e.g. from update 2 into update 2. You would need to export it as SWMM file or shapefiles, provided you have backup of the model saved elsewhere. 

Model backward / forward compatibility

Hi Sushma

Could you please elaborate about excessive floods seen in results of update 3? Do you mean that you ran the model which was last opened and computed in update 2 with no changes made in update 3 and you still you see difference in the results? - Yes, this is what happened. With the original model (saved using Update 2) there were no flood in 5YR storm. However, when the same model is opened and ran using Update 3 software, there were 50+ floods occurred.

In the original model, the continuity error is about 10%.

How can I get the download link for the old version? I saved the model (created using Update 2 version) before opening and saving it using Update 3 version therefore I have back-up copy.

Version was 10.00.17.10 when I used Update 2 Civilstorm. Thanks

Hi Larry,

Thank you for this. I will create a service request for this. 

Regards,

Jan

Jan, I believe Sushma may have already opened a Service Request on your behalf to send you the link. We'll check on that.

We'll also take a closer look at the model to see if we can determine why the model is sensitive and perhaps a set of calculation options that work better to achieve more stable (and more accurate) results in the latest version, so that you do not need to downgrade.

Hi Jan,

I wanted to share some brief preliminary notes regarding the models you sent. I attempted to open and compute the model in the "Update 2" folder in both 10.02.04.04 (the version that the model was last saved in) and in 10.03.01.08. In 10.02.04.04, the model ran but had a continuity error of 9.2% and 53% non-convergence. I also noticed a large amount of surcharging and near-flooding. 

CivilStorm 10.02.04.04 uses SWMM solver version 5.1.012, but CivilStorm 10.03.01.08 uses version 5.1.014 of the SWMM solver, as noted here and here. This means that results can be different between these versions when using EPA's solver. However, in 10.03.01.08, I am not able to compute the model at all, as it fails with "The SWMM engine failed during the calculation.".

A refinement of the advanced calculation options (such as the Routing Step or engine version) may be necessary, or you could try using the Implicit solver. Once the model results are stable, you should expect to see comparable results. But, when the model is sensitive and unstable, you will tend to see different results. This is especially the case if many pipes are close to being flooded, as a sensitive model could experience a change that is just large enough to make the difference between flooding and not flooding. 

We'll take a closer look and get back to you. Thank you for your patience.

As a follow-up to my previous reply - I have taken a closer look at the model today. Again, using the copy of the model from the "update 2" folder, this time I can confirm that the model computes for me in both 10.02.02.04 (the version that the model was last saved in), and in 10.03.01.08 (the current version). The convergence error and continuity error are about the same in both versions but as mentioned in my previous reply, the newer CivilStorm version uses a newer version of the SWMM solver, so differences in results are to be expected, especially in sensitive/unstable models (and the continuity error and non-convergence percentage indicate that this model falls in that category). 

I do see that no elements experience overflow in the older version and about 50 catchbasins experience overflow in the newer version, but if you graph overflow for all such catchbasins (with "is ever overflowing?" = True), you will see that the overflow volume is small and the overflow does not last long. Also, you have selected to use a ponded area for the surface storage, so the "overflow" that leaves the catchbasin rim will enter this storage area and then eventually drain back into the system, so no flow is actually lost in this case. Taking a closer look at some of the overflowing elements, the ones I checked were surcharged in the old version so they were close to overflowing anyways. The instability combined with the change in solver version is likely what caused enough of a change in results that they started to experience overflow in the new version.

As I had also mentioned, it is best to not look closely at model results until the numerical stability is brought under control. To that end, I have spent some time trying to determine the root cause of the instability and/or a set of calculation options that help with stability. I have not had success yet, but I suspect that the model is having a hard time converging because of a number of challenging situations in this model. Some of these factors include:

• Many elements experiencing surcharge and near-overflow conditions
• The ponded area above all catchbasins
• Many conduit control structures in the model and many parts where several pipes converge.
• Ponds with outlets

I will continue to attempt to stabilize your model, but in the meantime you can also find information on troubleshooting unstable SWMM solver results here. Again, once the model results are stable, you should see a more accurate picture of the system (and it may or may not indicate a flooding condition).

Hi Jesse

Thank you for your response.

It is true that the model is close to flooding for 1 in 5 years. The project should also be analysed for 1 in 10 year storm scenario where 2 hours maximum flood duration should be optimised. The project has pipe diameter constraint; the client does not want pipe diameter greater than 800mm. This is a challenge. And aside from that, there are also constraints like a.) the overflow volume at the connection point to the existing network should be 25% on the total project volume during peak; b.) pond areas can no longer be increased due to landscape requirements; c.) the connection point can no longer be moved since the current connection point is the only possible connection; no other existing network adjacent to the project that can be drained via gravity.

It would be helpful if some alterations in the calculation option can be advised since the pipe network can no longer be changed even the pond sizes...

I tried changing to implicit solver but the flooding for 1 in 5 years scenario gets worse.

1.) Any suggestion to the calculation option that should be adopted?

2.) How to determine the flooded duration per node? This can be extracted via Executive summary when using explicit solver.

Regards,

Jan

Hi Jesse

Thank you for your response.

It is true that the model is close to flooding for 1 in 5 years. The project should also be analysed for 1 in 10 year storm scenario where 2 hours maximum flood duration should be optimised. The project has pipe diameter constraint; the client does not want pipe diameter greater than 800mm. This is a challenge. And aside from that, there are also constraints like a.) the overflow volume at the connection point to the existing network should be 25% on the total project volume during peak; b.) pond areas can no longer be increased due to landscape requirements; c.) the connection point can no longer be moved since the current connection point is the only possible connection; no other existing network adjacent to the project that can be drained via gravity.

It would be helpful if some alterations in the calculation option can be advised since the pipe network can no longer be changed even the pond sizes...

I tried changing to implicit solver but the flooding for 1 in 5 years scenario gets worse.

1.) Any suggestion to the calculation option that should be adopted?

2.) How to determine the flooded duration per node? This can be extracted via Executive summary when using explicit solver.

Regards,

Jan

Hi Jan,

In a very challenging model like this one, calculation option adjustment can be a very time-consuming task. I have spent quite a bit of time on this already but I do not have any conclusive recommendations yet. Here are some of the options that the model was sensitive to for the Explicit solver (the last three listed here tend to have an impact of surcharged/flooded models):

• Routing Step (tried between 0.1 and 10 seconds, 1 sec is typical)
• Solver compatibility (EPA vs. Bentley-enhanced)
• Max trials per timestep (tried between 8 and 200)
• Minimum surface area (tried between 0 and 10 m)
• Surcharge Method (Extran vs. Slot)
• Use Bentley transition equation (true/false)

The SWMM solver does tend to be more stable in situations with ponding and control structures, but I did some testing today with the Implicit solver. Results are not yet conclusive but I started to get better results with the following settings:

• Timestep and output increment: 0.025 hours
• NR Iterations: 20
• Computation Distance: 10 m

Note that because you have ponded area set for all nodes, you may not see overflow volume reported (since flow out of the node rim enters the ponded area and can recede back into the system).