Negative pressure in system due to D2A element (Hammer/WaterGems)

Question

I have a system (shown below in Option 1) which consists of two pumps (1 working + 1 standby) which is supplying water to three tanks. The pump is considered to be VSP type, with a target flow of 575 m3/hr. The demand at D2A-1, D2A-2 & D2A-3 are 101 m3/hr, 373 m3/hr and 101 m3/hr respectively. I have to carry out surge analysis for the system. To make things simpler, I have considered three D2A elements (orifice type) in place of the tanks. However while computing the intitial conditions I get a notification regarding negative pressure in the system stating "Negative pressures in system at one or more time steps." After referring to the profile of HGL and elevations, I found that the HGL at D2A-2 and D2A-3 goes below the physical elevation at the respective D2A. Moreover, the flow supplied is coming out to be 1265 m3/hr inspite of setting the target flow to 575 m3/hr. I tried increasing the head of the pumps keeping the pump discharge constant, but that didnt help as well. 
 I created another model (shown below in Option-2) where I replaced the D2A with tanks and then assigned a D2A at the outlet of each tank. I assigned the respective demands at each D2A from the outlet of tanks. After computing the model, the model did not render any warnings. 
 So why were there negative pressures in the system while I was using D2A in place of the tanks?

Jesse Dringoli · Accepted Answer

Hello Phanindra, 
 The D2A element is a good option for modeling a top fill tank where the inlet pipe is spilling into the tank (there is an air gap) and the flow acts like an opening to the atmosphere. However it is important to understand that the D2A represents a pressure dependent outflow - as the pressure drops below the "typical" pressure, the flow will be less than the "typical" flow and when the pressure is higher, the flow will be higher. 
 When you place a tank, you are forcing the hydraulic grade to be what you enter for the tank's initial HGL. Demands placed on the tank are essentially extracted downstream of it, so the flow into the tank can vary, but the head will stay the same. 
 In your "All D2A" scenario, because of the energy balance in the network, the pressure at the D2A locations are significantly different than the entered "typical pressure" so the outflow is significantly different than the "typical flow". Basically the pair of "typical" flow and pressure that you have entered, defines the size of the orifice, and the flows that you see a the D2As are what would happen if you had an orifice of that size. 
 If these D2As represent top-fill tank outflow, then a good approach would be to determine what the pressure and flow would be when you model them as tanks (with the top fill option set), and use those values in the equivalent D2A properties. With the D2As set up to accurately represent the tank outlets, you should get the same initial head/flow. 
 However when I look at the "D2A replaced by tanks" scenario, I see that the "Level (inlet invert)" are not set correctly. For example "Neboda tower" has its inlet invert level set to 61.9 m, which is an elevation of 117.675 m (base elevation of 55.775 plus inlet level of 61.9). I'm assuming the value you entered represents the inlet *elevation*, so it should be set to = 61.9 - 55.775 = 6.125 m 
 Likewise, "Batamulla Kanda reservoir" should have an inlet invert of 5.428 and "Agalawatta tower" should have an inlet invert of 6.125. Here is a related article about this: 
 Error or bad results when using top filling tank option 
 Even with these fixes, only tank "Batamulla Kanda reservoir" fills and the other two don't accept any inflow, because the HGL is lower than their inlet opening elevation. Basically with one pump at 575 m^3/hr, it can only supply the tank at the lower elevation (Batamulla Kanda reservoir). With the desired pump flow and the pipe diameters and tank elevations, it does not appear possible for all three tanks to be filling. 
 With these widely varying tank elevations and significant pipe headlosses, you may need to decide what needs to be done to ensure that all three tanks actually fill at the same time. Perhaps in the real system, only some would fill at a given time, and controls would open and close tanks upstream of the tanks, based on their level. I would imagine you have some control scheme in place, as it is unlikely that these tanks will be kept exactly in balance with the downstream demands. For transient purposes you would then need to decide what timestep/conditions to model as your initial conditions. It could be that you simulate what happens when a pump shutdown occurs with one tank open and the other two closed, for example. 
 One quick thing I could find to ensure that all three tanks fill, was to place a PSV on p-73 with an HGL setting of 62.0 m. This enables the pump to add more head to lift the water to the other two, higher tanks, while still supplying the tank at the lower elevation. However, the first tank (Neboda tower) receives most of the flow (481), due to the higher available pressure. You could use my suggested approach of taking the tank inlet elevation and this observed tank inflow to convert to equivalent D2As, but to get the exact outflows you want you would have to do something else to the network. 
 For example you could place a FCV in front of each tank, with the desired flow setting, so that the valves throttle to try to meet the desired tank inflow. This could cause an "ill conditioned" network though if the upstream pump is discharging exactly the sum of the desired FCV flows. However, I was able to accomplish this in your model and set up an example: 
 1) "Neboda tower" - replaced with FCV @ 101 m^3/hr plus D2A at an elevation of 61.9 m, a typical flow of 101 m^3/hr and a negligible Pressure Drop (typical) of 0.01 m 2) "Batamulla Kanda reservoir" - replaced with FCV @ 373 m^3/hr plus D2A at an elevation of 19 m, a typical flow of 373 m^3/hr and a negligible Pressure Drop (typical) of 0.01 m 3) "Agalawatta tower" - replaced with FCV @ 101 m^3/hr plus D2A at an elevation of 60.9 m, a typical flow of 101 m^3/hr and a negligible Pressure Drop (typical) of 0.01 m 
 I have saved a copy of the model with these changes which you can download from the below link. Note that this model was saved in the latest version 10.01.01.04. I see that you may be using an older version (check under File > help > About) so you may need to upgrade first before being able to look at this. 
 Kethenna neboda agalawatta batamulla - Jesse.zip

Jesse Dringoli · Answer

Phanindra, 
 You're quite welcome. 
 Regarding how to model valve closure - if the valves you're referring to are valves at the tank inlet, then you could either model them with the D2A set as a valve, or with a TCV. With the D2A set as a valve, only a simple linear closure can be modeled. So, if you need finer control over the closure rates, the TCV option would be best. 
 The FCVs that I had previously mentioned are just an example to illustrate one thing that could regulate flow into the tanks. It is up to you to determine how this will actually be accomplished. If a real flow control valve needs to be used in the system, then you would need to determine if the valve closure you are trying to model is a closure of the FCV or a closure of a separate valve at the tank inlet. If the tank inlet valve is what you're closing, you would need a separate TCV element, in addition to the FCV. 
 If the FCV is actually the only valve in play here (and represents the tank inlet valve set to the right position to cause the desired inflow), and the one you need to close, you could simply keep it as a FCV, and use the transient operating rule in the FCV properties to model the valve closure just as you would with the TCV. The difference is that the initial headloss through the FCV is determined for you (the headloss necessary to cause the desired flow). In fact, the equivalent minor loss coefficient and discharge coefficient are shown to you - see the "Results" section in the FCV properties. If you enter a value for the "minor loss coefficient" in the FCV properties, to represent the loss through the valve in the fully open position, it will also display the calculated relative closure in the Results Section. You can then use this as the first position in your transient operating rule.

Sushma Choure · Answer

Hello Phanindra, 
 When you are adding tanks in system in place of D2A elements, that tank is hydraulic boundary element for the pump so pump need to reach that hydraulic grade at the tank by adding required head. So you do not see negative pressure in case of tanks. 
 However in case of D2A element pump do not see that as hydraulic boundary and HGL at D2A element is not reached, so negative pressure in seen in system. Is this D2A element at the high point in system? 
 This behavior is explained in below technote, please go through it. 
 Why do I get a negative pressure at a high point in my system? Shouldn't the pump add enough head to push the water over the hill? 
 If required you can upload model files for our review. 
 Sharing model files

Sushma Choure · Answer

Hello Phanindra, 
 You should model this as D2A element itself for surge analysis, to check for the reason behind negative pressures at D2A elements (and not at upstream junctions of D2A), it could be that flow through D2A element is high so negative pressure is seen. 
 Please check the 'typical flow' and 'typical headloss' values for the D2A.

Negative pressure in system due to D2A element (Hammer/WaterGems)

Top Replies