SewerGems - simulate flooding after pump failure

Thanks but while that information is helpful, it doesn't solve our issue. Specifically, how do we simulate a pump failure with the SewerGems implicit engine AND the second part is then how do we calculate available storage in the mains downstream of the first overflow point. Obviously we could just look at length, volume, and size of pipes and work out how many hours storage we would have, but that only works for zero flow. What we are really trying to discover is, if the pumps fail at ADWF flows, how much available storage do we have until the first overflow point and what is the HGL of that outfall?

Hello Stephen,

I would set the pump on and off elevations such that the pump does not turn on. Compute the model and animate some profiles of the area in question to observe what happens. You should be able to see from this when flooding first occurs and what happens afterward. Maximum HGL figures are typically also found in the properties of most elements.

Additionally, you could graph the Overflow result field for your low manholes and/or add and sort on the overflow or "is overflowing ever?" results in a flextable.

Thanks Tom, that is exactly what we have been doing. We have identified the overflow point and from this noted several manholes at low elevations that could have their lids "bolted" and provide additional emergency storage.

Thanks that seems to have worked a treat. I wasn't aware the model would run without an outfall.

After further inspection, it seems our results don't give us what we were looking for and don't make sense.

We have over 4hrs of storage at ADWF in our trunk mains from first principles, but the model shows it is full at less than two hours. This appears due to higher than expected flows because some of the pump stations in the network are not variable speed and are in fact pumping to Peak Wet Weather. What this exercise has identified are deficiencies in our network design and recommendation for a storage ballast main upstream of the WWTP  would be dependent on pump station upgrades to variable speed.

EDIT: Additionally, the network flows are below ADWF when we are overflowing and the mains are full when they shouldn't be? So the model is showing inconsistent results that just don't add up. Does the sewergems implicit engine "initialize" the network at the start of a run? 

Hello Stephen,

Regarding "Does the sewergems implicit engine "initialize" the network at the start of a run?" - By default the solver assumes a dry condition to start the simulation, but gives two options for the "start type" - warm up or transition. See the help topic "Calculation Option Attributes", section "start type".

Can you clarify if you need further assistance? I was a bit unclear on your previous response if SewerGEMS is showing you results that you will need to take action on, or if you're having a problem with the model.

Thanks Jesse, apologies for the late response (Christmas holidays and all that :))

When we try model our 1ML ballast main (ie. 2.6m diam, 200.8m length) we get an inflow of 15 l/s. The model shows the main filling up in 2 hours, however, at 15 l/s it should of only filled up by 0.1 ML in that time. We put in a 1ML wet well upstream of the ballast main which filled at rates correlated with the inflows. This identified that the ballast main was not operating as expected in our model, ie. glitching. Does anyone have any suggestion as to the cause of this?

Also, it seems when using the SewerGems (implicit) engine there is only the option for "Start Type" of "transition start" or "warm start", how do we select a dry start (default)?

" In a warm-up start, the model estimates initial depths based on non-zero inflows and uses a user-specified warm-up time (entered in the Warm Up Time field) to approach steady state conditions prior to starting the actual simulation. In a transition start, the model initiates the simulation from zero flows and immediately reaches the actual none-zero flows in the first time step and continued the simulation."

Thanks for your help,

Hello Stephen,

I set up a small test model and believe I see what you're seeing. Using a pond element set as Pipe storage with the dimensions you mention, it fills up in about 20 hours, which is consistent with my math on the volume of the pipe and the time it would take for 15 l/s to fill it (assuming no flow leaves).

With a similar conduit setup (manhole > conduit > outfall with flap gate and high tailwater), it filled in about 3 hours. I also tried a setup with a manhole downstream of the conduit in question with outgoing pipe at a higher start invert and the pipe in question filled immediately. It is very possible that the start type you mentioned is coming into play.

What is the slope of the pipe in question and how does it connect to other elements? (is the downstream pipe inside a structure with a high start invert, such that the pipe in question must fill first before flow can leave?

In my testing, the Explicit (SWMM) solver appeared to provide an accurate picture of the filling of the pipe - it filled in about 20 hours. I used a routing step of 1 second and output increment of 0.01 hours.

If you need to model the filling of large storage in a pipe, I would suggest either modeling it as a pipe pond, or try using the Explicit solver.

Hello Stephen,

I set up a small test model and believe I see what you're seeing. Using a pond element set as Pipe storage with the dimensions you mention, it fills up in about 20 hours, which is consistent with my math on the volume of the pipe and the time it would take for 15 l/s to fill it (assuming no flow leaves).

With a similar conduit setup (manhole > conduit > outfall with flap gate and high tailwater), it filled in about 3 hours. I also tried a setup with a manhole downstream of the conduit in question with outgoing pipe at a higher start invert and the pipe in question filled immediately. It is very possible that the start type you mentioned is coming into play.

What is the slope of the pipe in question and how does it connect to other elements? (is the downstream pipe inside a structure with a high start invert, such that the pipe in question must fill first before flow can leave?

In my testing, the Explicit (SWMM) solver appeared to provide an accurate picture of the filling of the pipe - it filled in about 20 hours. I used a routing step of 1 second and output increment of 0.01 hours.

If you need to model the filling of large storage in a pipe, I would suggest either modeling it as a pipe pond, or try using the Explicit solver.