View Full Version : Problems with units and fireflow

February 20, 2017, 06:41 PM
I have a network that was calibrated.
Demand is in cmd (m3/d) and output pressure is in psi

I was getting weird results in fireflow analysis. So I change units
Operation->Simulation Option->BASE -> change Flow unit to gpm

I also go DB Editor ->Junction Demand-> Block edit-> and I convert the demand from cmd to gpm (multiply by 0.183451)

The run the simulation, but the results (output pressure stays in psi) are very different

What could be the reason?

Patrick Moore
February 21, 2017, 01:10 AM

Can you elaborate more on your question? Such as, what do you consider is "weird" in your fireflow results?

If your flow units were originally in cmd in the simulation options, the model will expect all input flow data to be in these units. You had adjusted demand but did you also adjust the other model input data such as pump curves, the fireflow demand, and valve settings that are flow related? This is one of the reasons you generally want to maintain a single set of flow units for a model as changing them can lead to difficulties. Additionally the units used for flow will also set the pipe diameters and pipe length units. This means that the units set in the simulation options will impact all of the units used by the model for input data and should not be changed unless you are prepared to adjust nearly all of the input data accordingly.

I suspect you set FF demand in gpm and were not getting the results expected as the model thought the values were originally in cmd not gpm. But when you changed the flow units you changed a lot more and probably got very crazy results as well.

Please provide more information and we will try to assist you in resolving your issue if this did not help you resolve it.

Patrick Moore

February 21, 2017, 03:46 PM
I am comparing with field tests. The tests report: static pressure, residual pressure and hydrant flow (estimated based on pitot measurements)
When I simulate the static pressure (before hydrant flow) I have good results; difference about 5 psi. (static pressure about 75 psi)
But when I run the fire flow, residual pressure goes negative (field test have residual pressure about 70-74 psi)

Patrick Moore
February 21, 2017, 03:52 PM

You are probably doing this already, but just to be clear it is best to do this type of comparison in the standard tab rather than the fireflow tab as you know what flows were used in the field. You just enter the hydrant flow as a point demand. If the model Pressure at flow << than field pressure at flow, this is usually indicative of a connectivity issue in the model which limits water flowing through key pipes. Try using map display to highlight pipes by total Flow and look at pipes in the are with no flow as an indication of where you may have a connectivity issue. These are most often caused by either 1) a diameter discrepancy of large pipe to small pipe to large pipe or 2) Pipe split candidates (lateral pipe looks like it splits the main but does not) or (nodes in close proximity (nodes close to one another that look connected but have no pipe connecting them. These can often be found using the InfoWater- Utilities - Network review/Fix tools.


February 22, 2017, 01:22 AM
I think maybe it is the roughness.
It is an old system (60 years aged pipes). I applied theoretical corrections for the roughness for aging pipes (Walski et al.,), don't remember the year, but is the same table as in the Innovyze book. Thus, I had C roughness about 55-65. Under normal demand the C has low influence; assuming new pipe C 130, the pressures are very similar. However, when I apply one big fire demand (5450 cmd), then I get negative pressures (-22). If I change the roughness to C=100, then the pressure increases (+40).

Patrick Moore
February 22, 2017, 07:35 AM
That is possible, when doing any calibration for flow testing it is best to make sure the following match first.

1) System demands match field demands as closely as possible
2) boundary conditions (tank levels and what pumps are running closely match flow rates in the model as did the field

You also before changing anything in the model want to make sure you have no major diameter discrepancies, pipe split candidates, or nodes in close proximity either as these can cause major errors in the fireflow results.

Once these match you are at the best possible point of comparing results both in the non-flowing and flowing conditions. Most modelers would consider having the model vs field results for both within 5 psi as a reasonable calibration goal. However, the comparison of pressure drop between the static and flowing conditions for both the model and the field should also be compared and should ideally also fall within 5 psi. The pressure drop comparison (assuming all other changes made above are completed first) is generally a great comparison of pipe roughness and connectivity. As most pipes don't corrode, in many cases you can often eliminate major c-factor adjustments for many systems as there is not justification for them.

Classically, you generally hear people discuss adjusting roughness to match flow conditions and pressure results, but unless the pipes have materials that are known to have internal corrosion like unlined cast iron and very old unlined steel pie that can corrode internally, changing c-factors with age is usually not justified as only pipes with significant internal corrosion would be expected to have major changes in the roughness over time. This means that all pipes that are cement mortar lines or plastic which is most pipes since around 1970 generally don't corrode over time and should have c-factors that are similar to average new pipe book published values.

Even if you do have pipes known to corrode it takes a very, very bad pipe to ever have c-factors below 60 and generally this would only be applied to small pipes of 4-6 inches and not larger pipes. So using values in the 55-65 range for all pipes would have likely been a poor assumption which seems to be what you were seeing.

I found in the systems I calibrated with Cast iron that this method seemed to work well and you may wish to consider it in your modeling. I would start with Cast Iron pipes 12 inch and larger set in the region of a test at 110 and then for each drop in size I dropped the c-factor 5 points.
So first cut I had the following c-factors:
12 and larger - 110
10 - 105
8 - 100
6 - 95
4 and smaller 90

Then after running the model I would adjust all values together up or down 5 to 10 points at a time to better match field results. This is a reasonable middle of the road starting point for Cast Iron pipes that can be adjusted as needed to increase or decrease the headloss. But I would generally avoid using c-factors lower than about 60-65 on the low end unless the pipes are verified to be very very corroded. Only adjust c-factors when you have eliminated all possible other causes as for most systems there is not a good justification for major c-factor adjustments unless you know the pipes corrode.

Even though with your C-factor changes your results still seem to be 30-40 pounds low though. (if what you noted before you had negative pressures in the model and around 70-75 in the field and only gained about 40 psi with the psi change) You may very well still have connectivity issues as well as only something major could cause that much difference. Make sure you verify boundary conditions and what pumps are running and check you connectivity in the model and for major diameter discrepancies. I would also look for high headloss pipes and also look for major pipes with no flow in them as a check for something unusual as well to help explain the remaining large discrepancy in results.

Hopefully this may give you additional considerations to consider.


February 23, 2017, 01:45 AM

Important to remember that C is empirical one.
I was thinking, Is it possible maybe similar to the Manning roughness. for low discharges the roughness is higher, but during peak floods the roughness is lower even in the same river reach; also for low flow the C is lower (more resistance)

Patrick Moore
February 23, 2017, 08:06 AM

You are correct, C is an empirically developed value, but tends to work well for pressurized water pipe flow. That being said, the Hazen Williams C for any pipe would be expected to be constant under all conditions. That means, you would not have one set of C-values for low flows and another set for high flows. Any pipe would be considered to have only one c-factor associated with it that would need to work for all conditions. The good thing is that serves as a good validation and check for any c-factor changes made.

However, in a hydraulic model as most systems are sized for fireflow during typical "normal" system demand conditions (no fire flow occurring) there is generally very little headloss in a system because the pipe velocities are very low. Because of the low velocities under these lower demand conditions you can dramatically change your c-factors and see very little difference in system pressures as the hydraulic grade across a pressure zone is very flat (i.e. very little headloss occurs). However, when the velocities increase that is when you will notice the impact of the c-factors much more clearly. This is one of the reasons in a good flow test you generally want to see at least 10 - 20 psi pressure drop when flowing a hydrant which in certain circumstances can only be accomplished by flowing two or more hydrants. If the pressure drop during the flow test is less that 10 psi, it can be difficult to truly calibrate/validate the system as it has been only lightly stressed during the test. When the pressure drop is 10-20 psi or greater, this will usually sufficiently have enough headloss to stress the system which makes the modeler more confident when making changes during calibration/validation. Having multiple residual hydrants around the actual flow test also allows independent confirmation of results and is very helpful in identifying the possibility of closed or partially closed valves in the field. It is also helpful to get exact system data at the time of the test so that you can as close as possible match boundary conditions from the field to the model.