Introduction #

When flow changes direction inside a drainage sump, access hole, or manhole, engineers may want answers to three different questions:

1. How does the structure affect the hydraulic grade line and flooding in the pipe network?
2. How much energy and momentum are associated with the turn?
3. What water-level difference, wall impact, or local pressure occurs inside the structure?

These questions require progressively more information. MiTS’s routing engine can answer the first within a one-dimensional network, and the second can largely be derived from that result and known geometry. This document is now deliberately limited to those two levels: Group 1 network results and Group 2 bulk calculations and diagnostics. The third question—local water-level rise and the limits of a geometric reference estimate—is treated in the companion document The Superelevation Equation for Flow Through Curved Conduits.

Definitions #

Let the inlet tangent point toward the sump and the outlet tangent point away from the sump:

\displaystyle \delta=\cos^{-1}\left(\operatorname{clamp} \left(\boldsymbol t_\text{in}\cdot\boldsymbol t_\text{out},-1,1\right)\right)

so \delta=0^\circ is straight-through flow and \delta=90^\circ is a right-angle turn.

Hydraulic questions and what MiTS can answer #

For a sump where flow changes direction, hydraulics engineers often want to know the answers to certain questions. One can separate the distinction between questions MiTS can answer from the network model, questions it can support with additional calculations, and questions that require a spatial hydraulic model: see below.

These rows define the capability boundary used throughout the article:

• Group 1: network consequences represented directly in the routed model.
• Group 2A: reproducible calculations that can support a check when an external acceptance basis is declared.
• Group 2B: reproducible indicators for comparison and prioritisation, without a universal pass/fail meaning.
• Companion preliminary estimate: a sump-derived geometric radius and reference water-level rise documented separately.
• Outside MiTS: resolved spatial free-surface behaviour, wall pressure, and structural load.

The result classes must remain distinct in calculations, reports, and design decisions. Linking to the companion calculation does not promote its preliminary reference rise into a Group 1 network result, a conservative screen, or an upper bound.

The equations presented below applies to gravity stormwater or sewer routing, AKA gravity network. It assumes free surface of water, which is the condition that the gravity network is mostly analyzed under.

How MiTS gets there automatically #

For a supported one-inlet/one-outlet sump, MiTS assembles the complete analysis as illustrated below:

No separate angle, K , sump-size, or drain-width input is needed for Groups 1–2 — MiTS derives all of it from the connected geometry itself. This is the main value added by MiTS over a pure 1D hydraulic analysis engine; we do consider bends and channel widths to the relevant and possible extent.

It should be noted that the derived K coefficient remains preliminary. The angle-only HEC‑22 method does not represent relative pipe sizes, benching, plunging flow, multiple inflows, or the detailed flow regime inside the chamber. MiTS therefore calculates K consistently according to the Hec-22; it does not claim that K is the exact coefficient of the constructed sump, but then, it’s good enough for practical purposes.

About K #

The coefficient K affects the coupled network state and the quantities derived from the represented energy loss:

\displaystyle h_L(t)=K\frac{V_o(t)^2}{2g} \qquad\qquad P_L(t)=\rho g|Q_o(t)|h_L(t)

However, it does not determine Froude number, bulk momentum redirection, local superelevation, wall pressure, or structural force.

Why bend loss and superelevation are different quantities #

Upon looking at the formula above, and the superelevation water formula, one might conclude that since the minor-loss head and the bend superelevation both scale with velocity squared, they are the same thing. But really, they are not.

For the bend loss, the quantity

\displaystyle h_L=K\frac{V^2}{2g}

is an irreversible decrease in total mechanical head in the streamwise direction. Here K represents dissipation caused by real-flow effects such as viscosity, turbulence, boundary-layer growth, separation, and mixing. It is a total-head loss, not automatically an equal drop in local water-surface elevation; the latter also depends on the velocity heads and elevations at the two comparison sections.

Superelevation is instead a transverse water-surface or piezometric-head difference. The outward increase in pressure supplies the inward acceleration needed to turn the flow. In the ideal inviscid limit, that pressure force acts normal to the local velocity and can redirect the flow without reducing its speed or total mechanical head. An ideal curved flow can therefore have nonzero superelevation while h_L=0 .

Worked decision examples: redesign or investigate? #

We illustrate the cases for the analysis with and without K with chosen values. We will see that in the first, the adopted bend coefficient changes a Group 1 compliance result and justifies redesign directly. In the second, Group 1 still passes, but the additional indicators show that unresolved local behaviour is material enough to justify precaution or further investigation.

Example A: the bend loss changes the Group 1 decision #

Consider a sump with a 90-degree inlet-to-outlet turn and a containment/rim elevation of 13.10 m. The original model omits the bend loss. MiTS derives the angle from the connected drain geometry, adopts the applicable preliminary angle-only K , applies it to the designated conduit without double counting other losses, and runs an otherwise identical design case:

Given 13.10 m as the accepted design limit and otherwise suitable inputs, the design case establishes a Group 1 failure within the adopted model, without needing a Group 2 or companion local-rise result. Redesign, increased capacity, a changed invert, additional containment, or another network intervention can be justified directly. Without the paired-run workflow, the omitted bend loss would have left an apparent 70 mm pass.

The conclusion is conclusive against the declared criterion within the adopted model, not proof that every field condition is represented perfectly.

Example B: Group 1 passes but the local uncertainty is material #

Consider two feasible layouts for the same catchment, pipe sizes, sump, and boundary conditions, both draining to a 400 mm square sump with containment/rim elevation 13.10 m:

Both layouts share a rim/containment elevation of 13.10 m.

The node HGL is one value for the whole sump (Group 1). Any separate local-rise reference applies only to the local turn and is not fed back into the node HGL; its derivation and interpretation are now kept in the companion superelevation document.

Group 1 shows both alternatives pass the rim check, but the alignment change increases the available modelled margin by 90 mm — a real network consequence of the adopted coefficient, to be weighed against the cost of realignment. It does not yet prove the current layout fails.

At the governing high-flow timestep, Q=0.12\ \text{m}^3\text{/s} , V=1.50\ \text{m/s} :

\displaystyle h_{L,90}=1.00\frac{1.50^2}{2(9.81)}=0.115\ \text{m} \qquad h_{L,\text{straight}}=0.15\frac{1.50^2}{2(9.81)}=0.017\ \text{m}

\displaystyle P_{L,90}=1000(9.81)(0.12)(0.115)=135\ \text{W} \qquad P_{L,\text{straight}}\approx20\ \text{W}

For approximately equal inlet/outlet discharge and speed, the 90-degree bulk momentum-change indicator is:

\displaystyle M\approx2(1000)(0.12)(1.50)\sin45^\circ=255\ \text{N}

while the angle component approaches zero for the straight layout. Group 2 therefore shows the turned layout has substantially greater energy dissipation and momentum redirection. 255 N is not a wall design force, but it is a reason to examine wall impact, benching, anchorage, erosion protection, or durability if those consequences matter.

The companion local-rise analysis may be run using the same routed state and two-dimensional sump geometry, but its numerical result is intentionally not reproduced here. Group 2 already establishes the narrower conclusion needed in this document: the turned layout dissipates more represented energy and redirects more bulk momentum, so it is the layout that deserves local hydraulic review if the available margin is consequential.

Then engineers can choose two paths: redesign the alignment if the nearly straight alternative is practical—the paired Group 1 result already shows improved margin and the Group 2 indicators reduce—or retain the turned layout and follow the companion local-rise and escalation workflow if realignment is expensive or constrained. The table below states general guidance for combinations of evidence. Only the first two rows are illustrated by Examples A and B above; the third describes a more benign combination not shown in either worked example, included for completeness:

MiTS shows whether the available evidence supports a design change, reveals when uncertainty is material, and lets engineers make an informed decision from there.

Group 1 — Network-model results #

Representing the sump loss #

FHWA HEC‑22 presents a simple preliminary method for access-hole losses, h_L=K V_o^2/(2g) , where V_o is the outlet velocity:

with linear interpolation between tabulated angles:

\displaystyle K(\delta)=K_1+\frac{\delta-\delta_1}{\delta_2-\delta_1}(K_2-K_1)

HEC‑22 explicitly cautions that access-hole losses are more complex than a simple relationship between outlet velocity head and angle, describes this as a preliminary estimate, and says it does not replace a detailed energy-grade-line calculation — benching, plunging inflow, relative pipe sizes, multiple inflows, and water depth can all matter. See FHWA HEC‑22, Fourth Edition, Sections 9.1.6.6–9.1.6.7 and Table 9.4.

At zero deflection, K is not zero — the chamber can still cause expansion, contraction, and mixing loss. This is separate from a transverse water-surface effect, which is zero for a straight path under the local-turning approximation.

MiTS’s Dynamic Wave engine supports entrance, exit, and average conduit minor-loss coefficients, evaluated as KV^2/(2g) ; MiTS supplies the selected coefficient before running the model, which then solves the network with the represented loss already coupled into the hydraulic state. The calculated h_L reported afterward is an audit of the loss already represented — it must not be added to the resulting HGL a second time.

Signed connected-link flows are needed to identify the actual inlet and outlet at every reporting timestep — node total inflow cannot replace individual link flows when reversal or multiple paths are possible.

Group 2 — Bulk hydraulic calculations and indicators #

Group 1 answers “does the network flood or lose margin?” Group 2 exists because a margin pass does not mean the bend is hydraulically mild — it answers a different, physically motivated question at the same node:

• Froude number Fr (defined formally in Group 2A below as Fr=|V|/\sqrt{gD_h} , the ratio of flow speed to the speed a surface disturbance can propagate at) tells the engineer whether the approach flow is subcritical (Fr\lt 1 , level controlled from downstream) or supercritical (Fr\gt 1 , level controlled from upstream and prone to standing waves and hydraulic jumps). It is also the gate used by the companion local-rise method: its urban-drainage reference estimate is returned only for a free-surface state with Fr\lt 1 . A Froude number near 1.0 is therefore informative because a small change in depth or grade can change both the regime and the method’s applicability.
• Specific energy locates the section relative to critical energy, the point at which a small change in geometry produces a disproportionate change in depth (choking). It is the same variable an engineer already uses to check culvert or channel controls elsewhere; Group 2 simply evaluates it at the turning node.
• Velocity head and loss power scale directly with the kinetic energy and dissipation rate at the bend — larger values correlate with a more physically severe turn (more turbulence, more potential for erosion, wear, noise, or vibration) even without a specific failure threshold, which is why Example B’s 135 W versus 20 W is a meaningful comparison before any criterion is declared.
• Bulk momentum-flux change is the closest bulk indicator to “how hard is this turn pushing on the structure.” It does not quantify a wall force, but a large value signals the same underlying momentum redirection that produces one — the natural trigger for asking whether benching, anchorage, or a wall-load check is warranted.

In short: Group 2 exists because “does it flood” and “is this bend hydraulically severe” are different questions, and a network model that only answers the first can pass a layout an engineer would still want to look at more closely. Group 2 quantities are calculated from the model’s time series and known directions. They retain dimensions and physical meaning, but describe the flow in bulk rather than the spatial distribution inside the sump, and they are not all equally actionable:

Group 2A quantities may support a design check once their acceptance basis exists; Group 2B indicators remain useful for comparison and for deciding what additional check to perform, but a large value alone does not establish failure. MiTS should label an output diagnostic only when no applicable criterion has been supplied.

The two Group 2A rows are not symmetric. Represented head loss can acquire an external pass/fail number — a project declaring “no more than 0.10 m EGL loss at this class of structure” is a real, checkable budget, as in the worked check below. Froude number and specific energy do not work that way: there is no handbook value of Fr that a sump bend must stay under, because Fr has no “severity” of its own. Fr=1 is a physical fact (critical flow), not a design allowance. Its actual use is as a gate on whether some other formula is valid — including the companion curved-channel reference method, which is withheld once Fr approaches or exceeds 1. “Declaring a criterion” for Fr means declaring which method is intended and confirming its regime assumption holds — not picking a target number to stay under.

This is how an engineer actually uses Group 2, even without a declared criterion: as a way to rank candidate layouts and decide where else to look. In Example B, 135 W against 20 W and 255 N against a near-zero momentum change do not fail the 90-degree layout (Group 1 already passed it), but they single it out as the one worth a benching, anchorage, or erosion-protection review if realignment is not adopted. That comparative ranking is the engineering value Group 2 adds beyond Group 1’s pass/fail result; it is not, by itself, a second pass/fail test.

Group 2A — Represented head loss, hydraulic depth, Froude number, specific energy #

\displaystyle h_L=K\frac{V_o^2}{2g}

This is where Group 2A earns a pass/fail conclusion Group 1 cannot give by itself. Suppose the project declares an allowable local loss of 0.10 m against the available energy-grade-line budget. Example B’s 90-degree layout computed h_L=0.115 m at the governing timestep — that fails the 0.10 m allowance by 0.015 m, even though the same layout passed Group 1’s rim- margin check (13.03 m against 13.10 m, +0.07 m spare). The straight alternative’s h_L=0.017 m comfortably passes the same 0.10 m allowance. This is the point of keeping Group 2A distinct from Group 1: a network-margin pass does not imply an EGL-budget pass, and only Group 2A can test the latter — but only once the 0.10 m figure has been declared; without it, 0.115 m is energy accounting, not a check.

This becomes actionable only when compared with a declared allowable local loss or available energy-grade-line budget; it is already included in the coupled design run and must not be added to the resulting HGL a second time. For free-surface flow:

\displaystyle D_h=\frac{A}{T} \qquad Fr=\frac{|V|}{\sqrt{gD_h}} \qquad E=y+\frac{V^2}{2g}

where A is flow area, T is free-surface top width, and y is flow depth. For the same outlet state as above (V=1.50 m/s, depth y=0.30 m, top width T=0.40 m in a part-full 400 mm pipe, so D_h\approx0.23 m), Fr=1.50/\sqrt{9.81(0.23)}\approx1.00 — flow at the classification boundary — and E=0.30+0.115=0.415 m. Froude number classifies the bulk flow regime and can establish whether a regime-dependent equation applies; specific energy can be compared with critical energy at a meaningful section. Neither supplies a universal sump acceptance limit.

Physically, Fr compares the flow velocity V with the speed c at which a small surface disturbance propagates. Picture a series of small disturbances released one after another as the flow moves downstream, each spreading outward as a circular ripple at speed c :

This is the same construction used for Mach number in gas dynamics, applied to shallow-water waves instead of sound waves: Fr is exactly the water-wave analogue of Mach number, and it carries the same physical meaning — whether a local effect can be felt upstream at all.

Group 2B — Velocity head, loss power, bulk momentum-flux change #

\displaystyle h_V=\frac{V^2}{2g}

is the kinetic-energy scale available for losses, waves, or impact — at V=1.50 m/s, h_V=0.115 m, the same 0.115 m that appeared as h_L when K=1.00 (h_L=KV^2/2g reduces to exactly h_V whenever K=1 , not a coincidence). It is an input to other relationships, not itself a prediction of local rise or a pass/fail quantity.

\displaystyle P_L=\rho g|Q_o|h_L

is the rate hydraulic energy is dissipated by the represented loss, permitting consistent comparison between events, locations, and alternatives. There is no universal wattage at which a sump erodes, vibrates, or becomes unacceptable — that requires a validated relationship to material, geometry, duration, and failure mechanism.

For one active inlet and outlet:

\displaystyle \boldsymbol M=\rho\left(Q_o\boldsymbol V_o-Q_i\boldsymbol V_i\right) \qquad M=|\boldsymbol M|

and, for approximately equal inlet/outlet discharge and velocity magnitude,

\displaystyle M\approx2\rho|Q|V\sin\left(\frac{\delta}{2}\right)

This has units of force, but the unit does not make it a wall-force check — it is one term in the control-volume momentum balance, not the structural reaction on the sump.

What Group 2 still does not provide #

The control-volume momentum equation also contains pressure forces, body forces, and unsteady storage. A complete wall or structural load additionally needs the pressure distribution over each wetted surface, the actual control-volume boundary, the water volume and weight in the structure, transient acceleration, load sharing between walls/benching/cover/outlet, and the impact location and wetted contact area. The bulk momentum change is a useful severity indicator and balance check, but must not be reported as “the force on the wall”; velocity head is an available-energy scale, not a predicted splash or jet-rise height. Two sumps can share the same Q , V , and HGL — and therefore the same Group 2 values — while having very different wall-pressure distributions: one may contain an attached broad turn, the other a narrow jet striking a wall directly.

Unless an applicable allowance, limit, or validated response relationship is declared, MiTS must present a Group 2 result as a hydraulic quantity or comparative diagnostic — not as evidence that the design passes or fails.

Companion analysis — Preliminary local water-level rise #

MiTS’s one-dimensional routing engine provides one hydraulic-head value for the sump; it does not resolve different water levels at the inner and outer walls. The separate companion document The Superelevation Equation for Flow Through Curved Conduits contains the complete local-rise treatment.

Steady design flow versus dynamic routing #

“Steady-state” is not one thing #

An external peak-flow hand calculation and MiTS’s own STEADY routing option are both, in effect, peak-flow states with the same defect for this purpose: neither represents storage, backwater, or conduit entrance/exit losses, so neither can include the adopted K — any HGL or margin computed either way is unreliable here, not merely approximate, and should not be read as a check on the sump loss at all.

Only a Dynamic Wave run with constant design inflows avoids that defect: it is the same solver as a full dynamic simulation, just with steady forcing, so it represents K , backwater, and pressurisation the same way a full event does. Everything else about “using one state instead of a full event” still applies to it, though — no storage/attenuation, no test of whether tributary peaks or downstream tailwater actually coincide — so it is not automatically conservative, just simpler; whether the single assumed state over- or under-states the real event depends entirely on how that state was assembled.

Why an independent peak-flow/Manning’s capacity check cannot substitute #

MiTS already contains a standing, routing-independent check: a per-drain Rational Method peak flow against that drain’s own Manning’s normal-depth capacity, used for preliminary pipe sizing. It cannot support this analysis, and not merely as a cheaper approximation: it compares one pipe’s imposed discharge against that same pipe’s own capacity, in isolation — no node, no backwater, no storage, no pressurisation, so there is no mechanism by which K at one structure could raise the level anywhere else. That coupling is exactly why K is written into the network model rather than reported standalone.

The same gap affects the imposed discharge itself: the Rational Method inflow hydrograph sums each contributing drain’s own peak at its own intensity and time of concentration — the standard assumption that every branch peaks simultaneously, which can overstate the true coincident peak at a junction with materially different tributary travel times, and a single-pipe check has no network state in which to test that. The Rational Method peak is still usable, though — as the imposed inflow to a constant-inflow Dynamic Wave run, the same discharge is subjected to the coupled solve with K , backwater, and storage represented, so the resulting HGL, margin, and flooding are network outputs rather than one pipe’s alone.

Which result groups need a dynamic event? #

A steady design state is nearly as useful as a dynamic event for Group 2 when inflows and downstream level vary slowly, storage/travel time are unimportant, the selected flows are hydraulically compatible and reasonably coincident, no important control operation or reversal occurs, the flow regime is stable, and the question concerns one operating state rather than duration or volume. It is less reliable when a “peak flow” is assembled by assigning every branch its individual peak at the same time — those peaks may not be coincident after routing and storage, and a single discharge peak may not coincide with maximum HGL if backwater or tailwater governs.

Escalating beyond bulk indicators #

When Group 1 or Group 2 shows that unresolved local behaviour may matter, use the workflow and modelling hierarchy in The Superelevation Equation for Flow Through Curved Conduits. The companion document defines when the preliminary geometric screen is available, when full-momentum 2D modelling is appropriate, and when vertical jets, impact, air entrainment, pressure, or structural loading require validated 3D CFD, measurements, or physical modelling.

A proportionate engineering workflow #

Here is how we would sequence it when no higher-fidelity local model is available:

1. Use Group 1 for network decisions — report HGL, surcharge, flooding, and margins with the adopted preliminary loss clearly identified.
2. Use Group 2A only with a declared acceptance basis — a represented head loss, Froude number, or specific energy supports an established check only once MiTS identifies the applicable allowance, flow-regime condition, or energy criterion.
3. Use Group 2B as diagnostics, not invented checks — compare alternatives and select the next assessment; do not translate into unverified erosion, vibration, wall-force, or pass/fail conclusions.
4. Hand local-rise questions to the companion method — retain its geometric-radius construction, subcritical gate, fixed C_\text{ref}=0.5 , and preliminary-reference label.
5. Escalate according to consequence — use measurements, physical modelling, full-momentum 2D modelling, or validated 3D CFD when the decision depends on actual local levels, pressures, forces, splash, or overtopping.
6. Define higher-dimensional outputs before modelling — the sections, wall areas, free-surface traces, spatial integrations, time statistics, and convergence criteria that will convert a 2D or 3D field into engineering results.

This preserves the value of simple calculations without overstating what they know.

Interpretation summary #

References #

• Federal Highway Administration, Urban Drainage Design Manual, HEC‑22, Fourth Edition, FHWA‑HIF‑24‑006, Sections 9.1.6.6–9.1.6.7.
• U.S. Environmental Protection Agency, Storm Water Management Model User’s Manual, Version 5.2, EPA‑600/R‑22/030.
• U.S. Environmental Protection Agency, Storm Water Management Model Reference Manual, Volume II: Hydraulics.
• U.S. Army Corps of Engineers, Hydraulic Design of Flood Control Channels, EM 1110‑2‑1601, Section 2‑5.
• Federal Highway Administration, Junction Loss Experiments: Laboratory Report, FHWA‑HRT‑07‑036.