Water topic 12 of 18 — free theory
Pipe Networks
A distribution system is a maze of loops and branches, but the whole maze obeys two rules: what flows into a junction flows out, and the head lost around any loop sums to zero. Everything else — equivalent pipes, flow splits, Hardy Cross — is bookkeeping on those two rules.
Take the free 7-question mini-quiz ↓
The two governing rules
Write every pipe's loss as hL = r Q|Q|n−1 (i.e. hL = rQn with the sign of Q). Then a network is solved when both of these hold everywhere:
Node continuity: ΣQin = ΣQout + qdemand
| qdemand | water withdrawn at the junction (0 for a plain pipe junction) |
| Sign convention | inflows positive, outflows negative — the algebraic sum is zero |
Loop energy: ΣhL = 0 around every closed loop
| hL | signed head loss: positive when the assumed flow follows the loop direction |
| n | head-loss exponent: ≈ 2.0 for Darcy-Weisbach (f treated as constant), 1.852 for Hazen-Williams |
The exponent matters. Double the flow and a Darcy-Weisbach loss roughly quadruples (22 = 4); a Hazen-Williams loss grows by 21.852 ≈ 3.61. An exam option showing ×4 is the Darcy answer; ×3.6 is the Hazen-Williams one.
Series and parallel — equivalent pipes
Two pipes in series carry the same Q and their losses add, so their resistances add. Two pipes in parallel share the same head loss and their flows add, so it is the square-roots of the resistances that combine:
Series: req = r1 + r2 + …
Parallel: 1√req = 1√r1 + 1√r2 + … (for hL = rQ²)
| r | pipe resistance from hL = rQn (Darcy: r = 8fL/(π²gD5)) |
| Flow split (parallel) | Q1/Q2 = (r2/r1)1/n — the lower-resistance branch takes the larger share |
Sanity check a split before trusting it: the branch with the bigger diameter (or shorter length) should carry the bigger flow. If your numbers say otherwise, you flipped the ratio.
The Hardy Cross loop method
For a looped network you guess the flows, compute the head-loss imbalance around each loop, then correct every pipe by ΔQ. One iteration usually gets close; exam problems ask for exactly one.
ΔQ = −ΣhLn · Σ|hL/Q| applied to every pipe in the loop
| ΣhL | algebraic sum of signed head losses around the loop (drives the correction) |
| Σ|hL/Q| | sum of absolute values — always positive, always in the denominator |
| n | the same exponent as the head-loss law you used (2 for Darcy, 1.852 for Hazen-Williams) |
| Corrected flow | Qnew = Qold + ΔQ for pipes in the loop direction; a negative result just means the real flow opposes your arrow |
Pipes shared between two loops get two corrections — one from each loop — with the sign depending on whether the loop directions agree on that pipe. Forgetting the second correction is the classic multi-loop error.
PE depth: reading a network
PE questions reward qualitative network sense as much as calculation. Three patterns cover most of them:
Closing a valve redistributes flow. Shut a branch and the remaining parallel paths carry more — their head losses rise roughly as Q², so the system head at the pump climbs and the total flow drops. Sketch the new system curve mentally before touching numbers.
Negative corrected flow is information, not an error. If Hardy Cross flips a pipe's sign, that branch genuinely flows opposite to your guess — common in loops fed from two sides. Keep the negative value and continue; do not re-guess from scratch.
Equivalent length for minor losses. Valves and fittings in a network are often rolled into an equivalent pipe length (Le = KD/f) so the whole system collapses to series/parallel resistances. Watch whether the exam gives you K values or expects you to look them up.
PE trap: Hardy Cross balances loops; node continuity must already be satisfied by your initial guess. If inflows do not match outflows at a junction before you start, no number of loop corrections will fix the imbalance.
Worked example One Hardy Cross iteration on a loop
Given: a single loop with hL = rQ|Q| (n = 2). Assumed flows (clockwise positive): pipe 1, r = 200, Q = +0.40 m³/s; pipe 2, r = 100, Q = +0.30 m³/s; pipe 3, r = 150, Q = −0.20 m³/s.
Solution:
- Signed losses: h1 = 200(0.40)(0.40) = 32.0 m; h2 = 100(0.30)(0.30) = 9.0 m; h3 = 150(−0.20)(0.20) = −6.0 m. ΣhL = 32.0 + 9.0 − 6.0 = 35.0 m.
- Denominator: Σ|hL/Q| = 200(0.40) + 100(0.30) + 150(0.20) = 80 + 30 + 30 = 140.
- ΔQ = −35.0/(2 × 140) = −0.125 m³/s.
- Corrected flows: Q1 = 0.275, Q2 = 0.175, Q3 = −0.325 m³/s. The new loop imbalance is only 2.34 m — down from 35 m in one step.
Answer: ΔQ = −0.125 m³/s; corrected flows 0.275, 0.175 and −0.325 m³/s.