[BUG] Is the default SWMM 6 the SWMM 5.3 Engine?

[dickinsonre]

Not really a bug but a question. I have compared almost a thousand SWMM 5.2.4 inp files to SWMM6, and I always get the same answer. Is this because I have to change the SWMM inp file for SWMM6 so that it uses new engine features? I am guessing this is the case.

[cbuahin]

Not really a bug but a question. I have compared almost a thousand SWMM 5.2.4 inp files to SWMM6, and I always get the same answer. Is this because I have to change the SWMM inp file for SWMM6 so that it uses new engine features? I am guessing this is the case.

If you are using the GUI, you can select the engine to be executed. If you are running using the cli, there should be separate executables/shared libraries for Swmm 5.3.0 and 6.0.0. While they should largely have numerical parity, there should be some minor differences in the results for certain configurations.

By the way, if you want to preview the alpha 3, you can download it from the url below.

https://github.com/HydroCouple/openswmm.engine/releases/download/v6.0.0-alpha.2/SWMMVis-6.0.0-alpha.3-win64.exe

[cbuahin]

We have strived to make the hydraulic routing qa problems achieve parity to the legacy engine. You should check the report file, it should have the version number of the engine used to generate it.

[dickinsonre]

Legacy answers are confirmed.

[dickinsonre]

thanks, i see differences now. Going to need an AI just to understand all of the various output. I am seeing differences,

VERIFIED RUN TABLE (3-conduit surcharged chain, 30 cfs, DW, 5 s, 2 h — alpha.3)
run routing-cont% C2 peak cfs
base (EXTRAN) -0.205 14.38
DYNAMIC_SLOT -0.478 16.32 (+13.5%)
SEMI_IMPLICIT +0.705 13.82 (-3.9%)
ANDERSON_ACCEL -0.203 14.38 (peak identical to base)
VIRTUAL_JUNCTION J2 -0.178 14.39 (VJ momentum residual max 791.64,
mean 5.82 cfs*ft/s in new rpt table)
ALL FOUR -0.478 19.29 (+34.1%)
Cross-check: the same base file in EPA 5.2.4 -> routing continuity is -0.181%.

[cbuahin]

Please keep me in the loop on any gaps discovered and what the remaining issues might be. Is this for all the models or just a particular model. I want to close these gaps and ensure parity for the same configurations as much as possible.

[dickinsonre]

This is my main app that can compare thousands of SWMM5 models, convert the swmm5 model into swmm6 with new featues and then compare the results. I have other apps that will look at the results in more detail https://batch-swmm-runner-robertdickinson.replit.app/ I amd adding an option to make a SWMM5 model bad so swmm6 will show up better, It is the opposite of the Auburn University RESWMM which makes SWMM5 models better. "UnSWWMM" Here is a typical result

[cbuahin]

Beautiful! By the way @dickinsonre , I have an explicit finite volume implementation for the 1d routing that is performing very well that we plan to release with alpha 3. It using the local time stepping that is proving to allow for more scalable simulations. I am hoping that it will be considered for evaluation in your app as well. On the 1d routing, that is the final thing on our roadblock before exploring GPU acceleration.

[dickinsonre]

fantastic, but is this not the alpha 3 release? Not that it matters, the key is finite volume being added.

[dickinsonre]

Making some progress on finding models that are different between swmm5 and swmm6 - i made a section that looks like the Lew Rossman qa reoport and seem some differences

[cbuahin]

Love it!

[cbuahin]

Because of the nature of the timestepping in SWMM6 an apples to apples comparison will require a constant timestep instead of the dynamic time stepping @dickinsonre. I want to make sure that is being accounted for. Thanks again! This is an awesome effort.

[dickinsonre]

Interesting, SWMM6 now runs all thousand models but swmm5 fails on about 5 percent with this error. some of the models came from infoswmm and infosewer so maybe i just found an issue with native swmm5 - will check on dummy links in swmm6 now. Good news is swmm6 never fails

EPA STORM WATER MANAGEMENT MODEL - VERSION 5.2 (Build 5.2.4)

user5 - DUMMY cross-section test

ERROR 134: Node 99 has illegal DUMMY link connections.