🎓 Lesson 13
D5
Measuring Network Robustness: Redundancy and Entropy
Redundancy measures how many backup paths exist in a water network if one pipe breaks, and entropy measures how unpredictable or evenly spread the flow is across the system.
🎯 Learning Objectives
- ✓ Calculate nodal and global redundancy indices for a given pipe network topology
- ✓ Compute hydraulic entropy from simulated flow distributions using EPANET or similar software outputs
- ✓ Analyze trade-offs between redundancy investment and entropy-based resilience in network retrofit design
- ✓ Explain how low-entropy networks are vulnerable to localized failures despite high pipe count
- ✓ Apply ASCE 7-22 and AWWA M32 guidelines to interpret redundancy and entropy thresholds for critical infrastructure
📖 Why This Matters
When a main water pipe bursts in a city—like the 2022 Atlanta pressure main failure—it’s not just the broken pipe that matters; it’s whether water can still reach hospitals, fire hydrants, and homes through alternate routes. Redundancy and entropy tell engineers *how well* the network bounces back—not just whether it’s built, but whether it’s *designed to endure*. These metrics transform abstract 'resilience' into quantifiable, design-ready targets.
📘 Core Principles
Redundancy originates from graph theory: in a water network modeled as a graph (nodes = junctions, edges = pipes), the *global redundancy index* R_g compares actual pipe count to the minimum needed for connectivity (i.e., number of edges minus (nodes − 1)). Nodal redundancy R_i measures local path alternatives per node. Entropy H is derived from normalized flow probabilities: H = −Σ(p_i log₂ p_i), where p_i is the fraction of total system flow carried by pipe i. High H means flow is distributed broadly—no single pipe dominates; low H signals bottleneck concentration. Critically, redundancy without entropy can mislead: a network may have many pipes (high R_g) but all flow funneled through one corridor (low H), making it fragile despite appearances.
📐 Key Calculations
Two essential formulas: (1) Global redundancy index quantifies topological surplus; (2) Hydraulic entropy quantifies flow distribution equity. Both require network topology and hydraulic simulation results (e.g., from EPANET). Entropy is unitless and bounded: 0 ≤ H ≤ log₂(N_pipes), with higher values indicating greater distributional robustness.
💡 Worked Example
Problem: A small municipal network has 42 pipes and 30 nodes. Hydraulic simulation shows flow rates across pipes: [0.8, 1.2, 0.5, ..., 0.9] L/s (42 values); total system flow = 48.6 L/s. Calculate R_g and H.
1.
Step 1: Compute R_g = (e − (n − 1)) / (n − 1), where e = 42 pipes, n = 30 nodes → R_g = (42 − 29) / 29 = 13 / 29 ≈ 0.448
2.
Step 2: Normalize each pipe’s flow: p_i = q_i / Q_total (e.g., first pipe: 0.8 / 48.6 ≈ 0.0165)
3.
Step 3: Compute H = −Σ(p_i log₂ p_i) over all 42 pipes. Using computed p_i values (excluding zero-flow pipes), H ≈ 4.12 (max possible = log₂(42) ≈ 5.39)
4.
Step 4: Interpret: R_g = 0.45 means 45% more pipes than minimally connected; H = 4.12/5.39 ≈ 76% of maximum entropy → moderately balanced flow distribution.
Answer:
R_g = 0.45; H = 4.12 (76% of theoretical max). This indicates acceptable topological redundancy but room to improve flow equity—e.g., by adding cross-connectors near high-flow pipes.
🏗️ Real-World Application
In the 2019 resilience upgrade of the City of Austin’s South Service Area, engineers used R_g and H to prioritize retrofit zones. Zone A had R_g = 0.32 and H = 3.8 (low entropy due to radial layout), while Zone B had R_g = 0.28 but H = 4.6 (looped grid with distributed flows). Despite lower pipe count, Zone B showed superior post-failure pressure maintenance during simulated valve-closure events. The project reallocated $1.2M from redundant pipe duplication in Zone A toward strategic loop closures in Zone B—raising average H by 0.5 and reducing outage duration by 41% in validation simulations (AWWA RF-2021 Case Study #44).
🔧 Interactive Calculator
🔧 Open Water Distribution Network Analysis Calculator📋 Case Connection
📋 Calibration of Lagos Metropolitan Water Network
Persistent model–field mismatch (>25% pressure error) due to undocumented pipe replacements and unaccounted demand growt...
📋 Leak Localization in Tokyo’s Historic Cast-Iron Network Using ITA
Acoustic methods ineffective due to soil attenuation and ambient noise; conventional pressure zoning lacked resolution
📋 Climate-Adaptive Reinforcement of Cape Town’s Drought-Resilient Network
System unable to maintain minimum pressure during prolonged low-storage operation and projected 20% rainfall decline
📋 Water Quality Model Validation for Singapore’s Deep Tunnel Sewerage System (DTSS) Supply Branch
Disinfectant residual dropping below 0.2 mg/L at farthest nodes despite design dosing; suspected wall reaction dominance