Flow comparison measures the water entering a pipe section and the water leaving it, and the loss share is (inflow − outflow) ÷ inflow. The comparison is only as good as the two measurements: two flows each timed to within 5% give a 15% loss that could really be anywhere from 8% to 22%, straddling the 10% limit, so lengthen the timings and use bigger containers before acting on the result.
Part of Water Distribution
Flow comparison is a leak test that puts a measurement at the inlet of a pipe section and another at its outlet and attributes the difference to loss. It is more precise than walking the line for locating losses between known points.
A Difference Of Two Numbers Is Less Certain Than Either Number
Take the inflow at 1.00 L/s and the outflow at 0.85 L/s. Loss = (1.00 − 0.85) ÷ 1.00 = 15%, which is in the “needs attention” band of 10-25%.
Now suppose each flow is only known to within 5%: the inflow is 1.00 ± 0.05 and the outflow is 0.85 ± 0.0425 L/s. Independent errors combine as the square root of the sum of their squares: √(0.05² + 0.0425²) = √(0.0025 + 0.0018) = √0.0043 = 0.0656 L/s. As a share of the inflow, the loss is 15% ± 6.6%, or between 8.4% and 21.6%, roughly 8% to 22%. That straddles the 10% limit between “acceptable” and “needs attention”, so the measurement cannot tell which side the system is on.
Where The Error Comes From
With a bucket and a stopwatch, the timing error is a fixed amount, say ±1 second (an assumption for a hand-timed bucket). Filling 20 L in 20 s gives a 1 ÷ 20 = 5% error. Filling the same volume in 60 s gives 1 ÷ 60 = 1.7%. So a longer fill, which means a bigger container or a smaller flow, gives a better number. For flows above about 20 L/s a bucket is no longer practical; Flow Measurement gives other methods.
What You Need
- Two identical containers and two stopwatches, or one container and one person doing both measurements one after the other
- Valves that hold the flow steady during the test (Isolation Valves)
- A way to hold demand constant: all other taps closed, or an agreed fixed draw
- The map, with the measurement points marked
Method: Same Conditions, Long Timings, Repeat
Step 1: Choose the section and the two measurement points.
Step 2: Fix the conditions. Close every other draw on the section. A change in demand between the two measurements is a larger error than the timing.
Step 3: Time the inflow to a container large enough for a fill of a minute or more.
Step 4: Time the outflow the same way, straight after.
Step 5: Repeat three times and look at the spread. A spread bigger than the supposed loss means the result is noise.
Step 6: Compute the loss and its uncertainty as above.
Worked Example
Three inflow fills of 60 s each give 1.00, 1.02 and 0.99 L/s; outflow fills give 0.83, 0.85 and 0.84. Means: inflow 1.003, outflow 0.84. Loss = (1.003 − 0.84) ÷ 1.003 = 16%. With a ±1 s error on 60 s (1.7%) on each flow, the errors are 0.017 L/s on the inflow and 0.014 L/s on the outflow, combined √(0.017² + 0.014²) = 0.022 L/s, or about 2.2% of the inflow, so the result is 16% ± 2.2%, clearly above 10% and inside the “needs attention” band.
Common Mistakes
- Comparing two measurements made at different times of day. Demand changes.
- Trusting a single run. Repeat it.
- Measuring with a container that fills in a few seconds. The timing error dominates.
- Reading a meter that is out of calibration. Check against a bucket.
What’s Next
- Learn next: Leak Detection — the whole system’s loss
- Combine with: Visual Inspection — once the section is known, find the place