When pump or component data is entered, PIPE-FLO generates a "best fit" equation to model the device. If the curve has abrupt changes in its slope, this method may result in an inaccurate model. To more accurately model these types of curves, the data can be adjusted to give a better fit over the operating range. Or, the data can be broken up into separate pumps or components in order to model the performance over more than one operating range.

In modeling pumps and components in PIPE-FLO, users occasionally run into a problem where the model does not accurately reproduce the curve data that it is based on. This usually occurs because the curve has abrupt changes in its slope. For example, a component curve may start out relatively flat and then become very steep.

Shown below is a set of data that a PIPE-FLO user had for his component:

   Flow (lb/hr)        dP (psi)
       0                       0
    30000                   1
    75000                  50
   125000                100
   175000                150

The following curve is obtained when this data is plotted by hand:


When this data is entered in PIPE-FLO, the program generates an equation to model the component:

         dP(comp) = CWⁿ

The values of n and C are determined using geometric regression. Using this method, the program comes up with a "best fit" for the curve, which is shown below:



Below is a comparison of the actual component data and the data produced using the PIPE-FLO equation:

    Flow        Input dP         PIPE-FLO dP          %Difference
        0                0                        0                         0
    30000              1                   1.523                    52.3
    75000            50                     21.4                   -57.2
   125000          100                   93.38                    -6.6
   175000          150                   246.4                    64.3

As can be seen in the above comparison, there is quite a bit of difference between the actual input data and the values produced by the equation.

A good way to model components such as these is to eliminate the abrupt changes in the curve by adjusting the input data that the curve is generated from. Making these adjustments can sometimes be a trial and error process. However, in PIPE-FLO's Pump and Component dialog boxes the data can be graphed as soon as it is entered, giving users instant feedback on their adjustments.

For this particular example, there are essentially two regions to the curve: Region 1 is from 0 to 30000 lb/hr and Region 2 is from 30000 to 175000 lb/hr. By eliminating the 30000 lb/hr data point from the component model (only the 0, 75000, 125000, and 175000 flow points are entered), the resulting curve gives a much more accurate model of the component from the 30000 to 175000 lb/hr region (Region 2):

    Flow        Input dP         PIPE-FLO dP          %Difference
        0                0                        0                         0
    30000              1                  15.29                    1429
    75000            50                    50.4                      0.8
   125000          100                      98                     -2.0
   175000          150                  151.8                      1.2

NOTE: The 30000 data point was not entered for this model, it is listed here for comparison purposes.)

With this revised component model, the only data point that is significantly off is the 30000 lb/hr data point. This model should be acceptable if the flows in the component pipeline are around or above 75000 lb/hr (Region 2). If the flow rate is in Region 1, a separate component can be installed in parallel to model this region (with data ranging from 0 to 30000 lb/hr) and a separate lineup created for the lower flow rate. To evaluate the system using the high flow rate lineup, isolate the low flow rate component. To evaluate the system using the low flow rate lineup, isolate the high flow rate component. 

The above example demonstrated how to adjust a component's data to get more accurate representation of the device in PIPE-FLO. This same methodology can be used when modeling a pump that has abrupt slope changes in its performance curve.