This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Unexplainable Results of Modeling AVs in Hammer

Hello.

I'm modeling a sewerage rising main of ID 409.2mm (HDPE - Wave speed: 327.22 m/sec. The system consists of:

1- Wet well.

2- Pump station (including GV, NRV,...)

3- Nos 3 AV (AV1 @ ST 0+000, AV2@ St. 0+300 and AV3 @ St. 1+000).

4- Outlet (discharge to atmosphere).

The pump station consists of 3 pumps (2 w + 1 stdby) to deliver 200 lps. The analysis was made for 2 scenarios as follows:

1- Scenario A: 3 pumps working - worst case scenario- (to size the surge tank and AV's).

2- Scenario B: 1 pump working (to check the negative pressure along the pipe route). The steady state HGL was showing a very minor negative pressure at AV2, and therefore, the "Treat Air Valve as Junction" was set to "yes". The negative head disappeared, and a small portion of the pipe downstream this AV was not running full (see HGL figure below).

Scenario A went successfully and a :Dip Tube Hydro-pneumatic tank" of 5000L is used to mitigate the negative pressure resulting from sudden power failure.

Scenario B was checked, and no severe negative pressure will take place. However, a non explainable velocity increase happens at AV locations (as seen below). This velocity increase is of coarse associated with discharge increase downstream of the AV. I extended the simulation period to a very large extent, but the discharge (and the velocity) never dies.

I can't understand the behavior of the line at the AV. I presume, this is an error in the software, that is separated the line into 3 separate segments and each segment behaves as if it's isolated. 

Can anyone help understanding this?

Appreciated.

Parents
  • The velocity profile shows the maximum (red line) and minimum (green line) velocity over the length of the profile. If you have the calculation option set to generate animation data, you can navigate through time and see the "current" HGL (black line) change. A large difference in these velocity lines on either side of an air valve indicates that the flow and/or the diameter is different.

    Assuming that all pipes are the same diameter, the difference you see is likely due to some assumptions behind the air valve element. After the transient conditions settle down after your pumps shut down, the hydraulic grade will likely remain lower than the air valve locations (the air valves will remain open). In reality, the pipes would drain our partially, settling the flow (and velocity) to a new steady value.

    However, by default, HAMMER cannot track the liquid/air interface along the length of the pipe and will instead assume that the pocket is concentrated at the air valve location. Therefore when the HGL drops to the air valve elevation and the air pocket starts forming, the hydraulic grade will remain roughly equal to the air valve elevation. So, with the pumps off, the hydraulic essentially act as if the air valves are reservoirs, so the flow between them is based on the flow necessary to induce a headloss equal to the difference in hydraulic grade. So, the bigger the difference in elevation, the bigger the flow (and velocity). So, the increase you highlighted on the left side (AV-1) is likely caused by the flow being zero (or near zero) on the left side of the air valve (flat HGL extending to the left of AV-1) and positive/higher flow on the right side of the air valve (the flow needed to bring the HGL from ~6.5 m (AV-1) to ~5.9 m (AV-2)

    I encourage you to enable the animation option ("generate animation data" in the calculation options) and animate both profiles to get a good visual of how this is happening.

    Here are some other references related to this:

    Modeling Reference: Air Valves

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/postive-flow-in-pipe-downstream-of-element-that-is-closed


    With that said, I have another concern with how the initial conditions are set up. You mentioned you used the "treat air valve as junction" setting for AV2; I assume you mean you set this to "false". As you noticed, this causes the upstream pump to "see" the air valve and add enough head to lift the HGL up to the air valve elevation. The small portion of the downstream pipe not running full is interpreted as part-full flow. In other words, the air valve is open during the initial conditions, with some air in the system and open channel flow in the downstream pipe for some distance. When the HGL goes back above the pipe elevation, this indicates pressure flow has resumed.

    This works well for a steady state/EPS run, but the transient solver in HAMMER can have issues with this. The reason is because HAMMER assumes that pipes are flowing full in the initial conditions. So, it interprets the drop in HGL as a headloss and thinks that the pipes are flowing full. You might see an "initial surge" (movement at the beginning of the transient simulation before the pumps shut down) from this, or an air pocket may immediately start forming. In this case, it may be best to end your transient model at AV2 (put the discharge to atmosphere there). For more on this, see the section called "What if my air valve is open during the Initial Conditions?" in the above air valve modeling reference technote.

    If this does not help, please include a copy of the model files (zipped) either in this public thread (use the advanced reply editor to attach) or confidentially using the below method:

    http://communities.bentley.com/p/bentleysecurefilesupload


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

  • Thanks Jesse, 

    Your reply helped a lot explaining the pipeline behavior shown in my post. Let me itimise the conversation as follows:

    1-  Yes, the pipe diameter is the same along the entire length of the pipes, and the increase in the flow velocity is resulting - as per your reply - from the static head differences between (AV-1 and AV-2)  and (AV-2 and St 1650+00).

    2- If the flow calculated between the AVs are based on the differential level between them -when the HGL drops below pipe level during surge event- and this is unavoidable, then, may I consider this velocity as a software related assumption, and simply ignore the velocity diagram as long as the pressure envelope along the pipeline is acceptable?

    3- Although the initial negative pressure value at AV 2 is -0.35m (which is considered acceptable), I'll modify the rising main profile to make sure that no negative pressure will take place when computing the initial conditions.

    Attached is a ZIP file containing the model.

    Regards,

    RM_Surge.zip

  • Hi Ahmed,

    Re: "may I consider this velocity as a software related assumption, and simply ignore the velocity diagram as long as the pressure envelope along the pipeline is acceptable?"

    This is something that you will need to decide based on the assumptions and your own judgment. It's reasonable to assume that the velocity (and flow) would eventually reach zero in the real system, despite what you see in the model (due to the assumption discussed), but that doesn't necessarily mean that no action should be taken. The pressures and other results could be impacted by these assumptions too, not just the velocity. You'll need to carefully consider the effect that the assumptions could make on the results you need and decide how to proceed.

    Also remember that a transient could occur when the pumps start back up and expel air out of the air valves. You may want to set up a pump restart to check for this. Here are some resources on that subject:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/9076.pump-shut-down-speed-vs-torque-and-pump-start-up

    http://communities.bentley.com/products/hydraulics___hydrology/f/5925/t/79302

    With that said, I took a look at the model and have the following observations:

    1) It appears that you're using an older version of HAMMER, 08.11.03.19 (SELECTseries 3). You may consider upgrading if possible. The latest version is SELECTseries 5, 08.11.05.61. In my below analysis, I used your version, with the latest corresponding cumulative patch set applied. If your company has a SELECT subscription, you can refer to the below Support Solution for how to download:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/8175.how-do-i-download-watergems-watercad-hammer-sewergems-sewercad-civilstorm-stormcad-pondpack-flowmaster-culvertmaster

    2) The discharge-to-atmopshere node D2A-3 has some questionable values in the rating curve. I would recommend considering the Orifice option, or use a Reservoir at the same elevation. You can read more about the D2A element here:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/3443.modeling-reference-discharge-to-atmosphere-tn

    3) The copy of the model you sent has the situation where AV-2 is open during the initial conditions, with part-full flow downstream. As discussed, this will cause issues as HAMMER assumes pipes are flowing full. In case you hav

    4) Tank T-3's initial elevation is set equal to the maximum. This is generally not recommended as it can sometimes engage the built-in altitude valve, closing the adjacent pipe. It's best to set the initial elevation somewhere between the minimum and maximum elevation. See this support solution for more:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/13135.what-happens-when-a-tank-becomes-empty-or-full

    5) Consider reducing your calculation timestep, which is currently set to 1.0 seconds. Maybe you had done this as part of the test to extend the duration far enough to allow things to settle down, but you should be aware of how pipe lengths (or wave speed) can be adjusted when the timestep is too large. You can read more about this in the below Support Solution:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/9401

    If you look at the "Length Adjustment" field in the Pipe flextable, you'll see that some pipes are adjusted by over 1000 meters. This can potentially skew your results, so it is something to watch out for.

    6) Hydropneumatic tank HT-3 doesn't appear to be set up properly. First, the minor loss coefficient is set to zero. Second, the initial HGL and gas volume are at odds with the dimensions of the dipping tube. The initial HGL of ~7.6 m (HGL in the "results" section, since "treat as junction" is set to "true") is below the bottom of the dipping tube (7.89 m) yet the initial volume of gas (1943 L) is less than the full volume of the compression chamber (2099 L). Also typically the initial HGL would be well above the bottom of the dipping tube (compressing the gas during normal operating conditions where the pump is on). Here is a reference for modeling hydropneumatic tanks in case you have not seen it already:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/3104

    7) Your transient calculation options are set to store results for all points and all times. Consider reducing the amount of results (and calculation duration) by limiting report points to include only the elements you need to graph.

    8) check valve node CV-1 has an elevation of zero and may be redundant, since the pump next to it is also using the check valve option. The slow open/closure features are not being used in the check valve node, so you may want to consider removing it. if you want to model a slow closing check valve, use the open time, closure time, etc fields, change the pump valve type to Control valve and enter a large number for the "time for valve to close" (in order to model neither a control valve nor a check valve). You can read more about check valves here:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/4558

     

    Now, to illustrate how addressing the above items effects the model results, see attached modified version of your model (RM_Surge_Bentley_SS3.zip). Here are the changes I made:

    a) Changed D2A-3 to be an orifice instead of a rating curve
    b) Used a "pressure drop" of half a meter on the D2A, simulating extra headloss, causing the upstream pump to add more head, avoiding the issue with the air valve opening in the initial conditions.
    c) Set both air valves "treat as junction" to "true" to avoid unnecessary complication in the calculations since we know the pressure will be positive in the initial conditions.
    d) Set tank T-3's initial elevation slightly below the maximum.
    e) Set the duration to 100 seconds and the calculation timestep to 0.01 seconds (a typical value)
    f) Selected only a few key points for the Report Points in the transient calculation options
    g) Adjusted hydropneumatic tank T-3 such that the values make more sense. Keeping the initial HGL of ~7.6 m the same, I set:  
    - dipping tube bottom to 7.0 m  
    - dipping tube top and variable elevation curve maximum elevation to 8.0 m  
    - equivalent diameters on variable elevation curve to 1.95 m (so the new top of 8.0 m still results in a full volume of 5000 L.)  
    - Initial volume of gas to 1000 L  
    - initial liquid volume to 4000 L  
    - minor loss coefficient to 0.5.


    I've also attached another model (RM_Surge_Bentley_SS5.zip), saved in the latest version. This version uses the "Extended CAV" calculation option (improved in the latest version) to show some pipe draining effects over a 20 minute duration. You can read more about Extended CAV in the Air Valve technote previously provided.

    I used the Skelebrator tool to merge pipes in series for the two paths (which had a constant slope) downstream of each air valve, then set the calculation options to use the "Extended CAV" option, which allows HAMMER to track the air/liquid interface. This tracking is limited to the extent of the two adjacent pipes only, which is why I used Skelebrator on the pipes that would drain out. As part of this, I also had to adjust AV-13's elevation slightly, as the nodes adjacent to all air valves must be lower than the air valve in order for the Extended CAV option to work. This adds some burden to the numerical solver (and it works better in the latest version) but I wanted to illustrate how you *could* see the pipes drain out and thus the velocity approach zero. It also may help understand the impact (or lack thereof) of the assumptions in the SS3 model with air pocket tracking to the overall envelope. Here's an animation:

     


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

Reply
  • Hi Ahmed,

    Re: "may I consider this velocity as a software related assumption, and simply ignore the velocity diagram as long as the pressure envelope along the pipeline is acceptable?"

    This is something that you will need to decide based on the assumptions and your own judgment. It's reasonable to assume that the velocity (and flow) would eventually reach zero in the real system, despite what you see in the model (due to the assumption discussed), but that doesn't necessarily mean that no action should be taken. The pressures and other results could be impacted by these assumptions too, not just the velocity. You'll need to carefully consider the effect that the assumptions could make on the results you need and decide how to proceed.

    Also remember that a transient could occur when the pumps start back up and expel air out of the air valves. You may want to set up a pump restart to check for this. Here are some resources on that subject:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/9076.pump-shut-down-speed-vs-torque-and-pump-start-up

    http://communities.bentley.com/products/hydraulics___hydrology/f/5925/t/79302

    With that said, I took a look at the model and have the following observations:

    1) It appears that you're using an older version of HAMMER, 08.11.03.19 (SELECTseries 3). You may consider upgrading if possible. The latest version is SELECTseries 5, 08.11.05.61. In my below analysis, I used your version, with the latest corresponding cumulative patch set applied. If your company has a SELECT subscription, you can refer to the below Support Solution for how to download:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/8175.how-do-i-download-watergems-watercad-hammer-sewergems-sewercad-civilstorm-stormcad-pondpack-flowmaster-culvertmaster

    2) The discharge-to-atmopshere node D2A-3 has some questionable values in the rating curve. I would recommend considering the Orifice option, or use a Reservoir at the same elevation. You can read more about the D2A element here:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/3443.modeling-reference-discharge-to-atmosphere-tn

    3) The copy of the model you sent has the situation where AV-2 is open during the initial conditions, with part-full flow downstream. As discussed, this will cause issues as HAMMER assumes pipes are flowing full. In case you hav

    4) Tank T-3's initial elevation is set equal to the maximum. This is generally not recommended as it can sometimes engage the built-in altitude valve, closing the adjacent pipe. It's best to set the initial elevation somewhere between the minimum and maximum elevation. See this support solution for more:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/13135.what-happens-when-a-tank-becomes-empty-or-full

    5) Consider reducing your calculation timestep, which is currently set to 1.0 seconds. Maybe you had done this as part of the test to extend the duration far enough to allow things to settle down, but you should be aware of how pipe lengths (or wave speed) can be adjusted when the timestep is too large. You can read more about this in the below Support Solution:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/9401

    If you look at the "Length Adjustment" field in the Pipe flextable, you'll see that some pipes are adjusted by over 1000 meters. This can potentially skew your results, so it is something to watch out for.

    6) Hydropneumatic tank HT-3 doesn't appear to be set up properly. First, the minor loss coefficient is set to zero. Second, the initial HGL and gas volume are at odds with the dimensions of the dipping tube. The initial HGL of ~7.6 m (HGL in the "results" section, since "treat as junction" is set to "true") is below the bottom of the dipping tube (7.89 m) yet the initial volume of gas (1943 L) is less than the full volume of the compression chamber (2099 L). Also typically the initial HGL would be well above the bottom of the dipping tube (compressing the gas during normal operating conditions where the pump is on). Here is a reference for modeling hydropneumatic tanks in case you have not seen it already:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/3104

    7) Your transient calculation options are set to store results for all points and all times. Consider reducing the amount of results (and calculation duration) by limiting report points to include only the elements you need to graph.

    8) check valve node CV-1 has an elevation of zero and may be redundant, since the pump next to it is also using the check valve option. The slow open/closure features are not being used in the check valve node, so you may want to consider removing it. if you want to model a slow closing check valve, use the open time, closure time, etc fields, change the pump valve type to Control valve and enter a large number for the "time for valve to close" (in order to model neither a control valve nor a check valve). You can read more about check valves here:

    http://communities.bentley.com/products/hydraulics___hydrology/w/hydraulics_and_hydrology__wiki/4558

     

    Now, to illustrate how addressing the above items effects the model results, see attached modified version of your model (RM_Surge_Bentley_SS3.zip). Here are the changes I made:

    a) Changed D2A-3 to be an orifice instead of a rating curve
    b) Used a "pressure drop" of half a meter on the D2A, simulating extra headloss, causing the upstream pump to add more head, avoiding the issue with the air valve opening in the initial conditions.
    c) Set both air valves "treat as junction" to "true" to avoid unnecessary complication in the calculations since we know the pressure will be positive in the initial conditions.
    d) Set tank T-3's initial elevation slightly below the maximum.
    e) Set the duration to 100 seconds and the calculation timestep to 0.01 seconds (a typical value)
    f) Selected only a few key points for the Report Points in the transient calculation options
    g) Adjusted hydropneumatic tank T-3 such that the values make more sense. Keeping the initial HGL of ~7.6 m the same, I set:  
    - dipping tube bottom to 7.0 m  
    - dipping tube top and variable elevation curve maximum elevation to 8.0 m  
    - equivalent diameters on variable elevation curve to 1.95 m (so the new top of 8.0 m still results in a full volume of 5000 L.)  
    - Initial volume of gas to 1000 L  
    - initial liquid volume to 4000 L  
    - minor loss coefficient to 0.5.


    I've also attached another model (RM_Surge_Bentley_SS5.zip), saved in the latest version. This version uses the "Extended CAV" calculation option (improved in the latest version) to show some pipe draining effects over a 20 minute duration. You can read more about Extended CAV in the Air Valve technote previously provided.

    I used the Skelebrator tool to merge pipes in series for the two paths (which had a constant slope) downstream of each air valve, then set the calculation options to use the "Extended CAV" option, which allows HAMMER to track the air/liquid interface. This tracking is limited to the extent of the two adjacent pipes only, which is why I used Skelebrator on the pipes that would drain out. As part of this, I also had to adjust AV-13's elevation slightly, as the nodes adjacent to all air valves must be lower than the air valve in order for the Extended CAV option to work. This adds some burden to the numerical solver (and it works better in the latest version) but I wanted to illustrate how you *could* see the pipes drain out and thus the velocity approach zero. It also may help understand the impact (or lack thereof) of the assumptions in the SS3 model with air pocket tracking to the overall envelope. Here's an animation:

     


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

Children
No Data