Skip to content

Worked example: Pump control

The honest question a professional asks about a modelling tool in a web page is: why should I trust it? This example answers it - not with adjectives, but with the standard case the whole industry uses to shake out a routing engine.

It is the EPA SWMM 5.2 pump-control example: a wet well with a pump switched on and off by real-time control rules, under a day of dry-weather inflow. It is a generic, schematic model - no real location - chosen because it is deliberately hard on the solver. A pump snapping on and off against a wet well is a discontinuity, and discontinuities are exactly where routing engines disagree.

SWMM Web runs the EPA SWMM 5.2 engine - the same proven C source as the desktop application, not a reimplementation of it. InfoSWMM and PCSWMM run that same EPA numerical core, and SewerGEMS offers it as one of its solvers. (InfoWorks ICM imports SWMM files but solves with its own engine, so expect close rather than identical numbers there.) So the claim here is not “our results are as good as theirs.” It is stronger and simpler:

These are the same results, computed in a different place.

The way to check a claim like that is to run the identical input file in two engines and compare the numbers. This page gives you the numbers our engine produces, on the standard case, so you can do exactly that.

Wet wellSU1 - a cylindrical storage node
PumpPUMP1, on a depth-flow pump curve (peaks at 0.9 CFS)
Controltwo rules: start when depth ≥ 4 ft, stop when depth < 1 ft ([CONTROLS])
Forcinga diurnal dry-weather inflow ([DWF] + a daily pattern) filling the well
Routingdynamic wave, 20-second step, over 24 hours

It opens in Schematic view - it is a generic benchmark, so there is no map.

Run it (the green Simulate button). Our engine produces:

QuantityValue
Flow-routing continuity error-0.017%
Pump start-ups over the day9
Pump utilisation94.7% of the time
Peak pump flow0.89 CFS
Wet-well level rangecycles between 1.0 and 4.1 ft

Every one of those is a number a reviewer can reproduce. The continuity error is the headline: −0.017% on a case built to break solvers means the engine conserved water almost exactly through nine on/off transitions. Open the same file in EPA SWMM 5.2 on the desktop, or in a commercial package, and you should read the same values back - if you don’t, that is worth reporting.

A pump turning on is a step change in flow. A naive fixed-step solver either chatters (the pump flips on and off every step) or smears the transition and loses mass. The settings that keep it clean are in Calculation options:

  • Dynamic-wave routing with a short 20-second routing step - fine enough to resolve each switch.
  • MAX_TRIALS 8 and HEAD_TOLERANCE 0.005 - the iterative solver converges each step before moving on.
  • A Hazen-Williams force-main equation for the rising main downstream of the pump.

Change these and re-run to see the continuity error move - that is the fastest way to feel why the defaults are set where they are.

  1. Run it here and record the five numbers above.

  2. Download pump-control.inp - it is a plain EPA SWMM 5.2 input file, nothing browser-specific.

  3. Open it in your own tool - desktop EPA SWMM, InfoSWMM, PCSWMM or SewerGEMS with its SWMM solver - run it, and compare the pump start-ups, peak flow and continuity. They should match, because underneath it is the same engine.

  • We do not ship a side-by-side table against the commercial packages. That would be our numbers next to numbers we typed in - not evidence. The honest form of the benchmark is the reproducible input file plus our results, so you run the comparison in a tool you already trust.
  • It is generic and schematic - a standard test case, not a real pump station, so it opens without a map.
  • It is a single dry-weather day. It exercises the control logic and the solver, not wet-weather capacity or long-term energy cost - the operational-cost angle on pumping (start/stop levels that move the same volume for less energy) is a separate story, told in the Bello example.