In a model with local high points that need to be accounted for with air valves (see Modeling Force Mains With Air Valves [TN]) one or more of the following problems may be encountered:
1) Pressure results indicate the user notification "network unbalanced"
2) "system disconnected" user notifications are seen
3) The error "Unable to solve network hydraulics equations" occurs
4) Negative pressures
5) Generally strange or unexpected results seen (such as in profile view)
When using air valves in WaterCAD, WaterGEMS or SewerCAD (or SewerGEMS with the GVF Convex solver), and the property "treat as junction?" is set to "false", it is important to understand that additional complexity is introduced into the model. To summarize the information in this article, the presence of such air valves allow an upstream pump to "see" the high point and add enough head to overcome it, preventing otherwise negative pressures. Behind the scenes, these open air valves are essentially treated as a Pressure Sustaining Valve (PSV) with HGL setting equal to the air valve elevation. So, the model must calculate the approach head loss/drop across the element, which can be viewed in profiles as part-full flow just downstream of the air valve location. Since this loss/drop is calculated dynamically, if there are multiple open air valves, it can introduce instability as they attempt to achieve a balanced condition.
The first step in troubleshooting a challenging model with multiple air valves at high points would be to set all of the air valves' "Treat as Junction?" property to "true", then run the model to see which high points experience negative pressures. You may want to run several different model conditions that you will be studying (different steady state periods, control arrangements, peak vs average loads, etc) to ensure that you find the air valves that will be open during at least one of those modeling cases.
Equally important, you should check that all of your data input is correct. With the air valves set to be treated as junctions, if your results are stable, you should be able to check some of the basic results (even if high points experience negative pressure) to ensure that everything is entered correctly. Do this with a steady state simulation to start. For example make sure that the pumps you expect to be on are turning on. If they are not, check their control range. It should be within the minimum and maximum elevation of the upstream tank/wetwell (and not equal to). If messages are encountered about a pump not being able to deliver head or flow, look at the profile view, pump operating point, pump curve and profile of the hydraulic grade line. Think about if the pump curve is correct or sized appropriately given what you see. If an unexpected headloss is seen in profile view, check the properties of the pipe and confirm the diameter, roughness, etc are set correctly.
Also, if you are using check valves in pipes ("Has Check valve?" = "True") consider removing those for the sake of simplicity. This is because check valve also introduce complexity in the calculations and can "fight" against the air valves. The pump element already has a built in check valve, so pipe check valves can be redundant.
Furthermore, ensure that your air valve nodes are added on the main line (in series with the pipes and junctions) and not at a tee with a lateral pipe.
Once the basic input and steady state results are confirmed, try running an EPS to check if controls and varying loads/demands are working. If the model fails midway (for example "cannot solve network hydraulic equations"), try backing off the simulation duration incrementally until the model runs. This can help narrow down the time when the failure is occurring. Try graphing the "percent full" attribute of all tanks (Edit > Select by Element > Tank, right click > graph). If you see a tank about to become full or empty, this could potentially cause problems. In this case, check the tank in question and confirm that the related pump control is set correctly. Also check if the pump (or pumps) are able to keep up with the tank inflow. For example you might find that when the pump(s) are on, they cannot keep up with the upstream wetwell inflow and cause the wetwell to overflow. In this case, you may need to check the pump definition of the downstream system (which effects the operating point and thus pump outflow).
Once the basic data input has been checked, start enabling air valves one at a time. Start at the upstream end and work your way downstream for example, focusing on the high points that you anticipate would likely be open, setting the "Treat as Junction?" property to "false". Some care may be necessary here. For example if you have two high points, you may need to assess if only one of them would be open (pressure otherwise below zero). Try one at a time, compute the model in steady state, then check the results in a profile view of that area (for example between the upstream pump and downstream wetwell). You may find that due to part-full downstream flow (again, see this article for more on that), a nearby downstream air valve may also need to be set to "false" to prevent negative pressures.
Note that you may sometimes find that additional air valves may need to be inserted in the model. For example if you have a high point that consists of a few junctions all around the same, high elevation and the air valve is at the upstream side, the next-downstream junction may experience negative pressure. In this case, it may be that the air valve should be placed at the downstream side of the hill, so the HGL on the upstream side of the hill remains positive. For an illustration, see "Example 2" at the bottom of this article.
Due to known issue #353236, if a pressure subnetwork has an air valve (with "Treat air valve as junction?" set to "False") and all upstream pumps are initially off, the orientation of the PSV (which is used behind the scenes to model the air valve operation) can be reversed, causing the PSV (air valve) to close, causing the upstream pumps to operate at their shut off head when turning on, with zero flow.
To resolve this, either turn the air valve off (set "treat as junction?" to "true") if it is not needed (if pressure is always positive), or ensure that at least one upstream pump is initially on. For example make sure that the wetwell initial elevation is between the pump on/off control range, with at least one pump in a station with an initial status of On.
After your steady state run is acceptable, continue on with an EPS (if you need to model one). If problems are noticed, they can be due to different conditions from a different combinations of pumps being on or off, or from differences in loading/demand compared to the steady state. Or, there could be problems with pump controls which may need to be checked. See Example 1 further below for an illustration of an example model with air valves, with good results with the pumps on and off.
If the profile still does not look right when the pump is off, you can place an imaginary wetwell on a short branch with a tiny diameter pipe at an Elevation (Initial) equal to the air valve elevation. This wetwell (which will not contribute significant flow) can eliminate the disconnected system message and correctly represent the fluid in the upstream pipe when the pump is off. Please note that the wetwell will need to be connected to the upstream side of the air valve. When the upstream pump is off, you may still see negative pressures downstream of the high point, which may be interpreted as part-full flow (the pipe may drain out).
If this approach does not work, try adjusting the diameter and length of the fictitious pipe. For example a slightly larger diameter. Confirm after computing that there is no significant flow through the pipe. You may also need to adjust the "pressure subnetwork accuracy" and "pressure subnetwork trials" calculation options, if the network is unbalanced at some timesteps. For example try setting the max trials to 400 and/or try an Accuracy value slightly larger (such as 0.005) or slightly smaller (such as 0.0005) compared to the default.
If you've used the above suggestions and are still running into problems, consider the possibility of using conduits and manholes instead of pressure pipes/junctions and air valves, for sections of the network that always experience part-full flow. Meaning, if there is an air valve that always remains open (pressure of zero in the model) with part-full flow in the downstream pipe, consider morphing the air valve into a manhole, with conduits going downstream to the wetwell. In the manhole properties, select the option for a bolted cover to allow for surcharging if needed. Or, use the Transition element instead of the manhole (as it represents an enclosed change from one pipe to another).
Of course, this approach may not make sense in cases where an air valve is not always open (where in some conditions it is, and in others the pressure is positive.)
Model with air valves configured, pumps on:
Same model, time step when pumps are off:
Air valve on upstream side of high point, negative pressure issue just downstream:
Same network, air valve on downstream side of high point - better results:
Downstream profile more accurate by adding additional air valves:
Modeling Force Mains With Air Valves (WaterCAD and WaterGEMS
Modeling Force Mains with Air Valves (SewerCAD / GVF Convex Solver)