Switching modelling tools puts years of calibrated models at stake. So the first question is always the same: do I get the same results? Here is the answer, measured.
In short — ORCANET reads and writes EPANET .inp files and gives the same results as EPANET 2.2: 18 of 20 test networks in strict parity, with a maximum head difference of 1 cm. It adds optimisation, calibration, water hammer and the digital twin.
The test bench runs EPANET 2.2 itself on every file, reads the report it produces and compares head by head, flow by flow, hour by hour. ORCANET is never compared with itself: an engine compared with its previous version only measures its reproducibility.
| Mechanism tested | Head difference | Flow difference |
|---|---|---|
| Real looped network, 163 nodes, 2 pressure-reducing valves | 0.010 m | 0.080 L/s |
| All six valve types (PRV, PSV, PBV, TCV, FCV, GPV) | 0.010 m | 0.070 L/s |
| Check valves | 0.000 m | 0.000 L/s |
| Leakage flows (emitters) | 0.000 m | 0.000 L/s |
| Rule-based controls [RULES] | 0.010 m | 0.010 L/s |
| Simple controls [CONTROLS] | 0.010 m | 0.020 L/s |
| US units (GPM, feet) | 0.004 m | 0.039 L/s |
| Darcy-Weisbach and Chézy-Manning head loss | 0.020 m | 0.000 L/s |
| 3-point and 4-point pump curves | 0.000 m | 0.000 L/s |
On an SI-unit file, EPANET delivers 1.341 times the declared power: it converts horsepower to kilowatts the wrong way round. A pump declared at 50 kW delivers 67 in its calculation.
EPANET freezes the level at 5.75 m while the declared maximum is 6.00 m, and loses 162 m³/h that appear nowhere. ORCANET caps the level at the declared maximum and accounts for the overflow volume.
Copying these two behaviours would have shown “20/20”. A compliance table obtained by reproducing the reference's errors is worthless to the engineer who signs the study.
Take your hardest network and compare it with the EPANET report you already have.
Yes, to within one centimetre. On 20 test networks compared with EPANET 2.2 — which produces the reference itself — 18 are in strict parity, with a maximum head difference of 1 cm: the last digit EPANET prints. The two remaining differences are documented EPANET defects, which ORCANET does not reproduce.
Yes, as it is, and it comes back out in the same format. Every section is read — junctions, reservoirs, tanks, pipes, pumps, all six valve types, curves, patterns, controls, rules, quality, sources, reactions and options — in SI and US units. You can work in ORCANET and keep delivering an .inp file.
EPANET computes steady state, extended-period simulation and water quality; ORCANET gives the same results and adds what EPANET does not do: pipe sizing and pump scheduling optimisation, automatic calibration, water hammer, a SCADA digital twin, satellite background, shared projects and thirteen languages — in a browser, with nothing to install.
Because EPANET gets them wrong. On an SI-unit file, it delivers 1.341 times the declared power of a constant-power pump; on a saturated through-flow tank, it freezes the level below the declared maximum and loses 162 m³/h without a trace. ORCANET follows the physics and documents the difference rather than reproducing the error.
Essentially, yes. Across 21 water-quality cases — chlorine, water age, source tracing, sources, reactions — complete mixing (MIXED), the default tank model, is in parity. Two stratified tank models, 2COMP and LIFO, are not yet aligned: if a study depends on tank stratification, cross-check it with EPANET.