Fact-checked by Grok 2 weeks ago

Hardy Cross method

The Hardy Cross method is an iterative technique for determining the steady-state flow distribution in looped pipe networks, such as those used in water distribution systems, where inflows and outflows at junctions are known but internal flows and head losses must be calculated. Developed by Hardy Cross and first published in 1936, the method applies principles from to balance head losses around closed loops through successive corrections to an initial flow estimate, enabling manual or computational solutions for indeterminate systems. The procedure begins with an arbitrary initial guess of flows that satisfies at each , followed by computation of head losses around each independent loop using an empirical relation like the Darcy-Weisbach equation, where head loss h_f = K Q^2 and K incorporates pipe length, , roughness, and fluid properties. A correction \Delta Q is then calculated for each loop as \Delta Q = -\frac{\sum h_f}{\sum 2 K Q}, adjusting flows in pipes shared across multiple loops with appropriate signs for direction, and the process iterates until corrections are negligible, typically converging in a few steps for simple networks. Cross originally described two variants—balancing heads by adjusting flows and balancing flows by distributing excesses—but the head-balancing approach became predominant due to its alignment with principles. Key assumptions include steady, incompressible flow with negligible kinetic energy changes and minor losses at junctions, as well as a power-law head loss function h = r Q^n (often n=2 for turbulent flow), which simplifies the mathematics but requires validation for varying conditions. The method applies to municipal , , and industrial piping systems, and has been extended to gas distribution, , and ventilation networks by adapting the head loss formulation. Despite its historical significance as the first practical engineering solution for looped networks predating widespread computer use, the Hardy Cross method excels in small systems with 2–3 loops due to its simplicity for hand calculations, but it suffers from slow or in large, with many loops or varying factors. Modern implementations often integrate it into software for dynamic friction adjustments via the Colebrook-White equation, though more efficient matrix-based methods like Newton-Raphson have largely supplanted it for extensive simulations.

History and Development

Origins and Inventor

Hardy Cross (1885–1959) was an American civil engineer and educator renowned for his innovations in . Born on February 10, 1885, in , he earned degrees from institutions including Hampden-Sydney College and the before joining the University of Illinois as a professor of in 1921, where he served until 1937. Cross is best known for developing the in the early 1930s, a technique that revolutionized the analysis of indeterminate structures by simplifying complex calculations. In November 1936, Cross published the foundational work on what would become known as the Hardy Cross method in the University of Bulletin No. 286, titled "Analysis of Flow in Networks of Conduits or Conductors." This paper introduced an iterative approach applicable to various network systems, including electrical circuits, structural frameworks, and fluid conduits such as . Although Cross's primary expertise was in structures, the method extended his analytical principles to , marking a significant crossover in applications. The development of the method stemmed from the need to solve flow distribution problems in complex looped pipe networks, particularly for water distribution systems, where external inflows and outflows are known but internal flows remain indeterminate. Prior to widespread computer use, formal algebraic solutions for such systems were impractical due to their computational intensity, prompting Cross to adapt relaxation techniques from , such as those inspired by slope-deflection methods, into a successive process for networks. This innovation addressed a key challenge in manual engineering computations of the era, enabling more efficient analysis without requiring advanced machinery.

Evolution and Modern Successors

Following its introduction in 1936, the Hardy Cross method was adopted in the analysis of water distribution systems, becoming a standard tool for engineers designing municipal networks where manual calculations were feasible for simpler looped systems. This widespread use stemmed from its efficiency in balancing flows and heads without requiring complex algebraic solutions, enabling practical application to urban water supply designs that previously relied on trial-and-error approximations. By the late 1950s and into the 1960s, the advent of digital computers facilitated the method's integration into automated computations, with early adaptations such as the 1957 program by Hoag and Weinberg applying Hardy Cross iterations to the , water system, marking the transition from hand calculations to computational analysis. This era also led to matrix-based successors, including variants akin to the Gauss-Seidel iterative technique, which enhanced convergence for larger networks by solving nodal equations simultaneously. Further refinements in the 1960s, such as the Newton-Raphson-based simultaneous node method proposed by Martin and Peters in 1963, built directly on these principles to handle more complex configurations with pumps and variable demands. Key successors emerged in subsequent decades, including the global gradient algorithm developed by Todini and Pilati in 1987, which addressed nonlinear head-loss equations across entire networks for improved stability and speed over loop-based iterations. This approach gained prominence through software like , released by the U.S. Environmental Protection Agency in 1993, which automated using the global gradient method and became a benchmark tool for simulating hydraulic and water quality dynamics in municipal systems. The method's influence extended deeply into , particularly for municipal water infrastructure, where it informed design standards and was featured in prominent textbooks starting from the 1950s, such as those on urban hydrology and . As of 2025, the Hardy Cross method remains a staple in curricula at universities worldwide, taught as a foundational iterative to illustrate balancing before introducing advanced computational methods like finite element for transient flows.

Fundamentals of Pipe Network Analysis

Basic Principles of Flow in Networks

Pipe in consist of interconnected systems of pipes forming loops and junctions, where external inflows and outflows are known, but internal flows and pressures must be determined to ensure proper distribution of fluids such as in municipal systems. These typically include multiple branches connecting nodes (junctions) and sources/reservoirs, with pipes characterized by lengths, diameters, and roughness that influence flow resistance. The analysis of flow in such networks relies on two fundamental equations: the and the head loss equation. The enforces mass conservation at each junction, stating that the algebraic sum of flows into the junction equals zero, or \sum Q = 0, where Q represents the in each connected (positive for inflow, negative for outflow). This ensures no accumulation or loss of fluid at junctions under steady-state conditions. Head losses due to friction are quantified by the Darcy-Weisbach equation, given by h_f = f \frac{L}{D} \frac{V^2}{2g}, where h_f is the frictional head loss, f is the dimensionless friction factor (dependent on pipe roughness and flow regime), L is the pipe length, D is the diameter, V is the average velocity, and g is gravitational acceleration. This equation, originally formulated by Julius Weisbach in 1845 and refined by Henry Darcy in 1857, relates energy dissipation to flow velocity and pipe properties. In pipe networks, a refers to any closed path formed by interconnected , which allows circulation of without external inputs. Under steady conditions, requires that the net head drop around any is zero, meaning the algebraic sum of head losses in the forming the must balance exactly. This principle, derived from of applied to incompressible fluids, ensures that the total (head) is conserved along closed paths in the absence of pumps or other sources.

Key Assumptions and Limitations

The Hardy Cross method relies on several core assumptions to simplify the of in networks. It assumes steady-state, , where the fluid properties and flow rates remain constant over time, and the does not vary significantly, which is typical for distribution systems under normal operating conditions. Pipes are treated as rigid, with no elasticity or deformation under , allowing the focus to remain on steady hydraulic gradients without considering transient effects. Additionally, minor losses such as those at junctions, bends, or fittings are considered negligible compared to friction losses in the pipes, and friction factors, which may vary across pipes in the network, often derived from empirical formulas like the Darcy-Weisbach equation for fully turbulent flow regimes. The method is typically applied to fully turbulent flow regimes, where head loss is nonlinearly proportional to flow (n ≈ 2), but can be adapted for laminar regimes (n = 1) or other conditions by adjusting the exponent n in the head loss formulation. Key simplifications further enable the method's iterative approach. Head loss is modeled as proportional to flow raised to a power, h_f = K Q^n, where n \approx 1.85 for the Hazen-Williams equation commonly used in water pipes or n = 2 for the Darcy-Weisbach equation in turbulent flow, linearized around an operating point for small flow corrections during iterations. The original formulation does not initially account for pumps, valves, or other active elements that introduce variable heads, focusing instead on passive looped networks with fixed inflows and outflows. These assumptions stem from the principles of head loss in , where dominates the . Despite its utility, the method has inherent limitations that can affect accuracy and applicability. Convergence may be slow or fail in large, due to the accumulation of errors in iterative corrections, particularly if initial flow estimates are poor. The approach assumes small corrections per iteration to maintain the for head changes, rendering it inaccurate for highly nonlinear systems—such as those with significant variations in flow or extreme turbulence—without modifications like acceleration factors. Furthermore, the network must consist of interconnected loops that are solvable, meaning no isolated components or underdetermined branches, as the method requires a balance of continuity and energy equations across closed paths.

Derivation of the Iterative Method

Method of Successive Approximations

The Hardy Cross method employs a relaxation approach to solve for steady-state flows in pressurized pipe networks by iteratively refining initial flow estimates until the hydraulic grade line closes around each independent , satisfying the condition that the net head loss in any closed path is zero. This technique draws from principles of successive approximations, where discrepancies in head balances are progressively minimized through incremental adjustments, ensuring compatibility with nodal continuity constraints. The general derivation commences with an arbitrary assignment of flows that adheres to the at all junctions, thereby establishing a feasible starting point without initial loop imbalances. For each , the unbalanced head \Delta H is computed as the signed sum of head losses across the pipes forming the , using an empirical friction formula such as the Darcy-Weisbach or Hazen-Williams equation. A uniform flow correction \delta Q is then introduced to all pipes in the —positive or negative depending on the assumed direction—to counteract this imbalance and drive \Delta H toward zero. At its core, the method linearizes the inherently nonlinear relationship between flow and head loss via a first-order Taylor series expansion around the current flow estimate Q. The incremental head loss is approximated as \Delta h \approx \frac{dh}{dQ} \delta Q, where h(Q) represents the head loss function, often of the form h = K Q^{1.85} for turbulent , yielding \frac{dh}{dQ} = 1.85 K Q^{0.85}. Summing these approximations algebraically over the loop pipes, the optimal correction that nullifies the first-order imbalance is \delta Q = -\frac{\Delta H}{\sum \frac{dh}{dQ}}, with the summation taken over all pipes in the loop, incorporating signs based on flow direction relative to the loop traversal. This formulation transforms the nonlinear loop equations into a series of linear corrections, solvable sequentially for each loop. Convergence of the iterative process is assured for networks with positive resistance coefficients and sufficient external heads, as each correction diminishes the residual imbalance. Specifically, for small \delta Q, the truncation error from the Taylor expansion—governed by the quadratic remainder term—ensures that the method reduces the overall error quadratically when head losses are approximately linear, akin to the behavior of Newton-type methods near the solution.

Balancing Heads Formulation

The balancing heads formulation constitutes the core of the Hardy Cross method for , emphasizing iterative corrections to flow estimates in order to equalize heads around each closed , thereby enforcing the physical requirement that the algebraic sum of head losses in any must be zero. An initial arbitrary distribution of flows is assumed that satisfies nodal but may result in head imbalances; for a given m, the imbalance ΔHm is computed as the directed sum of individual pipe head losses hf, where positive values align with the assumed traversal direction (typically ) and negative values oppose it. The head loss in each pipe is modeled empirically as hf = k Qn, where k denotes the resistance coefficient (incorporating pipe length L, diameter D, and friction parameters), Q is the flow rate (positive in the direction of traversal), and n is the exponent specific to the friction formula. To derive the flow correction δQm that nullifies ΔHm, a first-order expansion approximates the head loss change for small δQ: Δhfn k Qn−1 δQ. Summing these changes around the loop and setting the corrected imbalance to zero yields the relation δQm = −ΔHm / [∑(n k Qn−1)], where the summation is over all pipes in the loop and the terms in the denominator carry directional signs via sign(Q) (since Qn−1 = |Q|n−1 sign(Q)n−1, approximated as sign(Q) for the ). This proves the correction formula, with the denominator representing the loop's aggregate sensitivity of head loss to flow changes, often denoted using rk for clarity in derivations. For practical application with the Hazen-Williams formula (prevalent in networks), n = 1.85 and k = 10.67 L / (C1.85 D4.87), where C is the Hazen-Williams roughness coefficient; the correction becomes \delta Q_m = -\frac{\Delta H_m}{\sum \left[ \operatorname{sign}(Q) \cdot 1.85 \cdot |Q|^{0.85} \cdot \left( \frac{L}{D^{4.87}} \right) \cdot 10.67 / C^{1.85} \right]}, which can be expressed incorporating the dimensional factor (L/D)0.54 in simplified computational forms for the resistance term, as 4.87 ≈ 1.85 × 2.63 and the exponent adjusts accordingly in the . Similarly, for the Darcy-Weisbach formula (common in pressure conduits), n = 2 and k = 0.0252 f L / D5 in customary units (with f the ), yielding δQm = −ΔHm / [∑(2 k |Q| sign(Q))], emphasizing quadratic sensitivity. In networks with multiple loops, corrections for non-overlapping loops are applied independently, as adjustments in one do not affect others. For overlapping loops sharing s, superposition is employed: the net flow correction in a shared is the algebraic sum of δQm from all containing loops (with signs reversed for opposing traversal directions), enabling simultaneous updates across the network before reassessing imbalances in the next iteration.

Balancing Flows Formulation

The balancing flows formulation represents an alternative to the loop-based head balancing approach within the Hardy Cross framework, applied in scenarios where head differences, such as those imposed by fixed elevations, are known a priori. This method iteratively adjusts pipe flows to achieve zero net flow at internal junctions (), while ensuring compatibility with the energy equations derived from fixed boundary heads. It is particularly useful for networks with pressure-specified boundaries, where external inflows or demands may vary, but piezometric heads at key nodes remain constant. Developed as an extension of the original technique, this variant emphasizes nodal over loop closure and was elaborated in literature for analogies to electrical circuit analysis, treating flows as currents and heads as voltages. The derivation proceeds by linearizing the nonlinear head-flow relationship around an initial flow estimate and minimizing the sum of squared flow imbalances across through successive corrections. Start with an arbitrary initial flow distribution satisfying boundary heads and external demands. At each i, compute the flow imbalance \Delta Q_i = \sum Q_{\text{in},i} - \sum Q_{\text{out},i} - Q_{\text{demand},i}. The correction to flows in connected is then determined by adjusting an equivalent head increment \delta H_i at the junction, assuming other nodal heads fixed. For a friction loss h = K Q^n (where n \approx 1.85 for Hazen-Williams or n=2 for Darcy-Weisbach), the sensitivity is \frac{dh}{dQ} = n K Q^{n-1} = n \frac{h}{Q}, so \frac{dQ}{dh} = \frac{Q}{n h}. The induced flow correction in pipe k is \delta Q_k = \frac{Q_k}{n h_k} \delta H_i. Summing over m connected yields the total imbalance correction \sum_k \delta Q_k = \delta H_i \sum_k \frac{Q_k}{n h_k} = -\Delta Q_i, giving: \delta H_i = -\frac{\Delta Q_i}{\sum_k \frac{Q_k}{n h_k}} Updated flows are Q_k^{\text{new}} = Q_k + \delta Q_k, and the process iterates across junctions until imbalances fall below a tolerance. For simplified cases, such as junctions with pipes of uniform resistance (constant \frac{dh}{dQ}), the denominator approximates m / \left( \frac{dQ}{dh} \right), leading to a uniform flow correction \delta Q = -\frac{\Delta Q_i}{m} distributed equally among the m pipes. In more complex looped networks, adjustments account for shared pipes between junctions, modifying the denominator to include overlap factors (e.g., +1 for each unique pipe segment, +0.5 for shared branches) to prevent overcorrection. This incorporates the inverse head-flow relations, ensuring energy consistency. The method converges by reducing the global \sum_i (\Delta Q_i)^2. In matrix form for simultaneous corrections across the network, the imbalances \mathbf{\Delta Q} relate to loop flow adjustments via \mathbf{\Delta Q} = \mathbf{B}^T \mathbf{S} \delta \mathbf{Q}, where \mathbf{B} is the loop-pipe incidence matrix and \mathbf{S} diagonalizes the linearized head-flow sensitivities. The optimal \delta \mathbf{Q} = -(\mathbf{B}^T \mathbf{S} \mathbf{B})^{-1} \mathbf{B}^T \mathbf{\Delta Q} uses the pseudoinverse of the loop flow matrix to minimize squared imbalances, providing a least-squares solution for interdependent loops. Post-convergence, internal heads are recovered from flows via the energy equations if required. This formulation differs from the head balancing method, which assumes fixed external flows and corrects loop circulations to nullify head discrepancies; the flow balancing variant is less common due to its sensitivity to initial guesses but excels in pressure-driven systems, often requiring fewer iterations for nodal-dominant networks.

Procedure for Application

Step-by-Step Balancing Process

The Hardy Cross method applies an iterative balancing process to achieve compatible flows and heads in a pipe network, relying on the correction formula derived from successive approximations for head losses around loops. This ensures nodal and loop head balance through repeated adjustments. The process initiates with initialization, where arbitrary initial flow rates are assigned to each pipe such that is satisfied at all junctions (inflows equal outflows, excluding external demands). Independent loops are then identified, typically as the minimum number of elementary closed circuits that cover all pipes without redundancy; for a network with J junctions and P pipes, this requires L = P - J + 1 loops. These initial flows serve as a starting guess, often estimated based on proportional distribution of total supply or demand. In the iteration , head losses are computed for each using the formula h_f = k Q^n, where k is a pipe-specific incorporating , , and roughness, Q is the , and n is the exponent (typically 2 for Darcy-Weisbach or 1.85 for Hazen-Williams). For each independent , the algebraic sum of head losses \Delta H = \sum h_f is calculated, considering the direction of (positive in the direction, negative counterclockwise). The correction \delta Q is then determined as \delta Q = -\Delta H / \sum (n k Q^{n-1}), which approximates the change needed to balance the . Flows are updated as Q_{\text{new}} = Q_{\text{old}} + \delta Q for pipes in the direction of the correction and Q_{\text{new}} = Q_{\text{old}} - \delta Q for opposing pipes. When loops overlap, corrections from multiple loops affecting the same are applied by summing the signed \delta Q values sequentially across all relevant loops in each iteration; alternatively, simultaneous application can be used by solving the system of corrections at once for better stability in . If a computed \delta Q would cause a to exceed its capacity or reverse unrealistically, the correction is scaled proportionally to stay within limits. These steps are repeated for all loops in successive iterations. Finalization occurs when the maximum absolute head imbalance |\Delta H| across all loops falls below a specified , such as 0.01 ft of head loss, indicating to a balanced state. At this point, the final flows are used to compute nodal heads by tracing from a reference point, applying cumulative head losses along paths. The process typically converges in 3–10 iterations for simple networks. For practical implementation, manual calculations historically employed tabular formats to track flows, head losses, and corrections per loop, facilitating error checking during iterations. In modern practice, spreadsheets like automate the computations through formulas and iterative solvers, enabling rapid analysis of larger networks while maintaining the method's iterative logic.

Convergence and Error Correction

The convergence of the Hardy Cross method is typically assessed through specific criteria that ensure the corrections and head losses have stabilized sufficiently for accuracy. Common thresholds include an absolute head error less than 0.1 meters around each or a relative change in rates below 1% between successive iterations, where the maximum relative correction is calculated as \max(100 \times |\Delta Q| / |Q|). For small networks, is often achieved within 5 to 10 iterations, though more complex systems may require up to 12 or more, depending on the initial estimates and network size. These criteria are monitored after each full cycle of balancing to verify that the imbalances in head or have diminished to negligible levels. The method's self-correction mechanism relies on the iterative application of small flow corrections (\delta Q), which progressively reduce the overall imbalance in the network by addressing head loss discrepancies by . Each minimizes because the corrections are derived from the current imbalance divided by a term involving pipe resistances and flows, ensuring that subsequent adjustments build on increasingly accurate estimates. This successive approach, inherent to the balancing process, inherently dampens accumulated errors from initial guesses, leading to a balanced state where at junctions and in loops are satisfied within the specified tolerance. Non-convergence issues, such as oscillations or , can arise from poor initial flow assumptions or highly nonlinear head loss relationships; in such cases, refining the initial guess—such as assigning flows inversely proportional to pipe resistances—can accelerate stabilization. If oscillations occur, applying an under-relaxation (e.g., multiplying \delta Q by less than , such as 0.8) reduces the step size to prevent overshooting and promotes smoother . For , further of the head loss or re-evaluation of boundary conditions is recommended to ensure physical consistency. To enhance convergence speed, particularly in nearly linear systems, acceleration techniques like over-relaxation can be employed by multiplying the computed \delta Q by a factor between 1.1 and 1.5, which amplifies corrections to reach faster without introducing . This approach is most effective when the network's head losses approximate linear behavior with respect to flow, as higher factors (up to 2) risk in nonlinear cases. Such modifications maintain the method's simplicity while reducing the number of required iterations in practical applications.

Advantages and Practical Considerations

Computational Simplicity and Self-Correction

The Hardy Cross method employs straightforward algebraic operations, primarily involving summations of head losses and basic for flow corrections, eschewing the need for inversion or solving large systems of simultaneous equations typically required in direct applications of Kirchhoff's laws. This reliance on successive approximations transforms complex nonlinear problems into manageable linear relations at each , making it accessible for manual computations without advanced mathematical tools. A key feature of the method is its self-correcting mechanism, where computational errors introduced in initial flow assumptions or intermediate steps do not accumulate but are progressively diminished across iterations as the approaches . This inherent error-handling allows practitioners to perform manual verifications at each balancing step, ensuring that discrepancies in head losses around loops are systematically reduced until and conditions are satisfied. Compared to traditional methods based on Kirchhoff's laws, which demand solving cumbersome sets of simultaneous equations, the Hardy Cross approach offers superior speed and practicality for hand calculations in looped pipe networks. It was particularly effective pre-computer for small networks with a limited number of loops (typically up to around 10), enabling efficient preliminary designs and educational applications through its , loop-by-loop adjustments. Each scales linearly with the number of pipes, typically requiring computations proportional to the network size, which supports its use in both teaching and initial engineering assessments.

Limitations and When to Use Alternatives

The Hardy Cross method exhibits several key limitations that restrict its applicability in certain scenarios. Primarily, it becomes computationally inefficient for large pipe networks exceeding approximately 50 loops, where the number of iterations required for can escalate dramatically, often reaching hundreds or thousands, due to the iterative correction process across multiple interdependent loops. Additionally, the method is sensitive to nonlinear head loss relationships, such as those governed by the Darcy-Weisbach equation, leading to potential under-correction, over-correction, or oscillatory counter-corrections that hinder reliable . The approach fundamentally assumes steady-state flow conditions, neglecting transients, storage effects, or time-varying demands, which limits its use to static analyses without dynamic elements like or accumulation in pipes or junctions. While the basic Hardy Cross method is designed for simple pipe networks, it can be extended to incorporate pumps and certain valves (such as , , and pressure-reducing valves); however, for systems with complex control devices or highly nonlinear characteristics, extended formulations or alternative solvers are often necessary to handle these elements accurately. Similarly, for analysis or scenarios requiring rapid iterations, such as operational monitoring, the Hardy Cross method's dependence on manual or sequential corrections proves too slow, favoring instead robust numerical solvers that offer faster global solutions. Viable alternatives include the Newton-Raphson method, which provides superior global convergence through matrix-based linearization of the full , reducing iteration counts significantly—often by factors of 3 to 12 compared to Hardy Cross—especially in with poor initial flow estimates. For comprehensive modeling, commercial software like WaterGEMS employs advanced methods to simulate large-scale efficiently, incorporating pumps, valves, and multiple scenarios without the iterative bottlenecks of manual approaches. In cases involving unsteady flow, methods are preferable, as they explicitly account for temporal variations and transients absent in the Hardy Cross framework. In the modern context as of 2025, the Hardy Cross method remains valuable primarily for educational purposes or preliminary designs of small, simple networks due to its conceptual clarity and low computational overhead. However, it is largely outdated for full-scale simulations in professional practice, where computational power enables more sophisticated, automated tools to address complexity and scale effectively.

Applications and Examples

Real-World Uses in Engineering

The Hardy Cross method finds primary application in the design and analysis of municipal systems, enabling engineers to determine rates and pressures in looped networks for efficient delivery. It is also employed in networks, particularly for assessing in systems by iteratively adjusting flows to minimize head losses in closed versus open circuits. Additionally, the method supports sizing pipes in pressurized systems, such as force mains, to ensure balanced hydraulic conditions and prevent overloads in looped configurations. Today, it remains relevant in preliminary layouts for networks in developing regions, where or spreadsheet-based iterations provide cost-effective initial designs before advanced simulations, particularly in resource-constrained settings. The method aligns closely with (AWWA) guidelines, such as those in Manual M32 for head loss calculations using the Hazen-Williams , ensuring compliance in sizing and management. It is adaptable to fire flow demands by incorporating peak loads into loop balancing, supporting requirements for minimum residual s (e.g., 20 ) during emergencies as outlined in AWWA M31. As a foundational technique, the Hardy Cross method underpins modern hydraulic modeling software like , where its iterative principles evolved into global gradient algorithms for large-scale simulations following early computer adaptations in the . It continues to be taught in ASCE-accredited programs, emphasizing practical network analysis skills for both manual and computational contexts.

Numerical Example of a Simple

To illustrate the Hardy Cross method in practice, consider a simple closed-loop pipe consisting of two loops and seven pipes connecting six junctions, with supply from a and demands at the end junctions. The pipes have uniform lengths of 2000 ft and diameters ranging from 6 to 12 inches, with a Hazen-Williams roughness C = 95 for all pipes (a value typical for older cast-iron pipes). The total inflow at the source is the sum of junction demands, approximately 18 cfs in this case. Head losses are computed using the Hazen-Williams equation: h_f = \frac{10.67 L Q^{1.85}}{C^{1.85} D^{4.87}} where h_f is head loss (ft), L is length (ft), Q is flow (cfs), C is the roughness coefficient, and D is diameter (in). The exponent n = 1.85 is used in the correction formula. Initial flows are assigned arbitrarily but satisfying nodal continuity (inflows equal outflows at each junction), such as 5.0 cfs in the main supply pipe and distributed around the loops (e.g., loop 1: pipes 1, 2, -3, -4; loop 2: pipes 5, 6, -7, with shared pipe 3). Initial head losses are calculated for each pipe based on these flows, resulting in unbalanced net head losses around the loops (e.g., positive or negative Σ h_f depending on direction conventions, clockwise positive). For instance, if initial Q1 = 5.0 cfs in pipe 1 (12 in), the head loss is approximately 64 ft using the formula above. In the first iteration, the flow correction δQ for each loop is computed as: \delta Q = -\frac{\sum h_f}{\sum (n K Q^{n-1})} where K = 10.67 L / (C^{1.85} D^{4.87}) for each , and the sum is over pipes in the loop with sign for direction. Typical δQ values are small (e.g., 0.05-0.2 cfs per loop), reflecting minor initial imbalances. Flows are then updated by adding δQ (or subtracting for opposite direction pipes). Head losses are recalculated with the new flows. In the second , δQ values decrease (e.g., <0.05 cfs), and updates are applied similarly. A third iteration further refines, with δQ approaching zero. The process converges when |δQ| < 0.01 cfs and net head loss per loop <1% of total loop head loss. The converged flows and corresponding head losses after three iterations are presented below, verifying balance (net head loss ≈0 ft per loop within 1% accuracy). Pressures at junctions are determined by subtracting cumulative head losses from the reservoir head.
PipeDiameter (in)Initial Flow (cfs)Converged Flow (cfs)Head Loss (ft)
1125.05.71252.29
282.02.82725.90
360.50.2574.90
4102.52.75773.30
5103.02.885106.32
6100.50.3855.18
7104.04.61585.59
This example demonstrates self-correction, with flows stabilizing after few iterations due to the method's relaxation nature. For networks with higher inflows like 100 cfs, the procedure scales proportionally, maintaining the same relative corrections.

References

  1. [1]
    [PDF] Analysis of flow in networks of conduits or conductors / - IDEALS
    This paper deals directly only with networks of definite conductors. ... Analysis of Flow in Networks of Conduits or Conductors, by Hardy Cross. 1936 ...
  2. [2]
    [PDF] Steady Flow Analysis of Pipe Networks: An Instructional Manual
    The oldest and most widely used method for analyzing pipe networks is the Hardy Cross method, a description of which can be found in most hydraulics or.
  3. [3]
    Chapter 4 Water flowing in pipes: energy losses
    A typical method to solve for the flow in each pipe segment in a small network uses the Hardy-Cross method. This consists of setting up an initial guess of flow ...
  4. [4]
    Hardy Cross | Civil & Environmental Engineering | Illinois
    Hardy Cross was Professor of Structural Engineering in the Department of Civil Engineering at the University of Illinois from 1921 to 1937.Missing: biography | Show results with:biography
  5. [5]
    [PDF] Water Supply Systems and Evaluation Methods
    The Hardy-Cross method was sufficient for single-loop systems, but without the aid of a computer it was impractical for a system having several loops. An ...
  6. [6]
  7. [7]
    The History of Water Distribution Network Analysis: The Computer Age
    This paper will attempt to catalog and review those methods that have been developed and applied since the dawn of the "computer age" in 1957 when the original ...
  8. [8]
    CIVILENG 3005 - Advanced Civil Engineering Hydraulics
    Hardy-Cross method for hydraulic analysis of looped networks; Steady-state theory of pumps; Water hammer; Gradually varied flow (prismatic and compound ...
  9. [9]
    Water Distribution Analysis and Design - EPIC Civil Training Course
    Steady-state equations of flow in pipe networks; Using the Hardy-Cross method to solve simple networks; Quasi-steady state equations of flow in pipe networks ...
  10. [10]
    [PDF] 4 Pipelines and Pipe Network Hydraulics – I
    Pipelines are pipes in series, networks have loops. Steady flow is described by continuity and energy equations, with head loss relating to frictional losses.
  11. [11]
    [PDF] TOPIC T2 - FLOW IN PIPES AND CHANNELS
    1.9 Simple Pipe Networks. For all pipe networks the following basic principles apply: (1) continuity at junctions (total flow in = total flow out);. (2) the ...
  12. [12]
    (PDF) The History of the Darcy-Weisbach Equation for Pipe Flow ...
    The historical development of the Darcy-Weisbach equation for pipe flow resistance is examined. A concise examination of the evolution of the equation ...
  13. [13]
    None
    ### Summary of Basic Principles of Pipe Flow from the Document
  14. [14]
    Hydraulic modelling of closed pipes in loop equations of water ...
    Jan 15, 2016 · The energy conservation equation is described for each loop of the network, where the algebraic summation of the head losses (hj) of the ...
  15. [15]
    (PDF) Hardy Cross Method for Pipe Networks - ResearchGate
    Hardy Cross originally proposed a method for analysis of flow in networks of conduits or conductors in 1936. His method was the first really useful engineering ...
  16. [16]
    [PDF] An improvement of Hardy Cross method applied on looped ... - HAL
    Sep 13, 2017 · Abstract: Hardy Cross method is common for calculation of loops-like gas distribution networks with known node gas consumptions.
  17. [17]
    [PDF] Flow in Systems with Multiple Pipes
    Apr 29, 2012 · Define a set of independent pipe loops in such a way that every pipe in the network is part of at least one loop, and no loop can be ...<|control11|><|separator|>
  18. [18]
    [PDF] 2.5: Simultaneous Path Adjustment Method - eCommons
    Oct 26, 2023 · From the previous section on the Hardy-Cross method we saw that the derivative or gradient of the energy equation for a loop/path with.
  19. [19]
    THE ANALYSIS OF FLOW IN NETWORKS OF PIPES. (INCLUDES ...
    EPUB | PDF. Citation. CORNISH RJ (1939), "THE ANALYSIS OF FLOW IN NETWORKS OF PIPES. (INCLUDES PLATES AND APPENDICES).". Journal of the Institution of Civil ...
  20. [20]
    [PDF] Publication No. 77 PIPE NETWORK ANALYSIS By Mun-Fong Lee
    The Hardy cross method which sparked off the evolution of the numerous techniques of simulating pipe networks, is suitable only for relatively small networks.
  21. [21]
    ∑UQ ∑ - Repository IHE Delft Institute for Water Education
    Pipe network ... The method of balancing heads is originated from Hardy Cross method, proposed in. 1936. ... The flow balancing method was proposed three years ...
  22. [22]
    (PDF) An Efficient Iterative Method for Looped Pipe Network ...
    Oct 16, 2025 · This is the main advantage of the new Node-loop method, as the number of iterations is the same as in the modified Hardy Cross method.Missing: concept | Show results with:concept
  23. [23]
    PIPE Networks Analysis: Design & Flow in Water Distribution Systems
    Rating 5.0 (3) 1 Quantity Balancing Method. This method, also known as the nodal method and presented by Cornish in 1939,. can be used when the heads at various points in a ...
  24. [24]
  25. [25]
    [PDF] Implementation of the Hardy-Cross method for the solution of piping ...
    This method is based on the successive addition of flow-rate corrections in each branch, in order to achieve satisfaction of energy conservation along every ...<|control11|><|separator|>
  26. [26]
    Numerical and simulation analysis comparison of hydraulic network ...
    results shows that Hardy cross method convergence after 12th iteration and ... flow change (correction) in loop 1, loop 2 , loop 3 and loop 4 of the ...
  27. [27]
    [PDF] Algorithms for Pipe Network Analysis and Their Reliability
    Of the two methods described by Hardy Cross, the method of balancing heads (single path adjustment method) became the most widely taught and used method.
  28. [28]
    [PDF] Modelling of Hardy Cross Method for Pipe Networks
    Feb 4, 2023 · So the network can only be solved iteratively. The best-known method so far is called the. Hardy-Cross method for pipe flow networks.
  29. [29]
    A simple iterative method for water distribution network analysis
    However, depending on the size and the complexity of the networks, the Hardy Cross method requires too much iteration to converge, and sometimes, it may diverge ...Missing: troubleshooting | Show results with:troubleshooting
  30. [30]
    [PDF] Steady Flow in Pipe Networks by the Simple Loop Method
    convergence acceleration techniques are investigated. A computer program ... by Hardy Cross (Ref, 1), Satisfying continuity at the nodes gives rise to.
  31. [31]
  32. [32]
    Short Overview of Early Developments of the Hardy Cross Type ...
    Moreover, the Hardy Cross method can also be used for water pipe networks (district heating [7] and cooling networks [8]) and ventilation systems [9,10] (a ...
  33. [33]
    [PDF] The Hardy Cross Method and its Successors in Water Distribution ...
    In 1936 he published a paper entitled, “Analysis of Flow in Networks of Conduits or Conductors”. ... For many years the Hardy Cross method was the primary method ...
  34. [34]
    None
    ### Summary of Limitations and Alternatives for Hardy Cross Method
  35. [35]
    [PDF] Evaluation of nonlinear Iterative methods on pipe network
    Mar 31, 2021 · Although the Hard Cross method is widely accepted and used, it has some limitations. ... The Hardy Cross method and EPANET were used to create a ...
  36. [36]
    (PDF) Applying the Hardy Cross method to assess the energy ...
    Mar 22, 2024 · This approach relies on the successive addition of flow-rate adjustments in each pipe to achieve the energy balance in each network segment, ...
  37. [37]
    Short Overview of Early Developments of the Hardy Cross Type ...
    Hardy Cross originally proposed a method for analysis of flow in networks of conduits or conductors in 1936. His method was the first really useful ...
  38. [38]
    [PDF] WRC RESEARCH REPClRT NO. 83 - Illinois Water Resources Center
    While the Chicago's water supply system serves the city of Chicago. Page 20 ... The best known solution technique to this problem is the Hardy Cross Method.<|separator|>
  39. [39]
    ANALYSIS OF HYDRAULIC NETWORK USING HARDY-CROSS ...
    This paper aims to develop a simple procedure for analysis of water distribution network using hardy cross method with the help of electronic spreadsheets and ...Missing: successors algorithms
  40. [40]
    None
    ### Summary of AWWA M32-2012 on Hardy Cross Method, Head Loss Calculations, and Fire Flow Demands
  41. [41]
    [PDF] History of Water Distribution System-Pipe Network Analysis
    the major algorithms to be examined in this paper include: • The Hardy Cross Methods (Cross, 1936; Hoag and Weinberg 1957) o. The Node Method o. The Loop ...
  42. [42]
    408-the hardy cross method and its successors in water distribution ...
    This course covers the history, basic principles, assumptions, step-by-step procedures, advantages, and disadvantages for solving pipe network problems using ...Missing: 2025 | Show results with:2025