Every engine against a reference.
A surge study is only worth the trust you can put in it. Each SurgeSpace engine is compared with an independent reference — EPANET 2.2, published benchmarks, closed-form theory, separate ODE solvers — and the comparisons run in continuous integration on every change.
The hydraulics.
| Check | Reference | Result |
|---|---|---|
| Joukowsky rise, wave period, line packing | Closed form | Joukowsky within 0.05 % |
| Transient benchmarks Tnet0 and Tnet3 | Published benchmark results | Tnet0 exact; Tnet3 92 % within 3 m |
| Transient, Net3 valve closure | TSNet | Steady state and first waves match |
| Steady heads and flows (Net3, KY4, Net6) | EPANET 2.2 | Worst 0.012 m |
| PRV, PSV, FCV, PBV, GPV, TCV | EPANET 2.2 | Heads 0.003 m, flows 0.01 L/s |
| Check-valve and air-valve slam | a·v/g | As theory |
| Surge tank mass oscillation | Jaeger's formula; wave engine | Within 5 %; engines agree within 1 % |
| Differential surge tank | SciPy ODE integration (independent code) | Upsurge equal to 0.01 % |
| Motor start, direct on line | SciPy ODE integration | Rigid 0.03 %, wave 1.2 % |
| Pipe force on a straight pipe | ρ·a·V·A | Within 12 % (line packing) |
| Rigid-column sparse solve | Dense solve | Identical; 3,000 pipes at about 0.1 s a step |
| Built-in examples | Stored baseline | 880 values identical |
Against EPANET 2.2.
| Check | Reference | Result |
|---|---|---|
| Extended period, Net1, Net2, Net3 | EPANET 2.2 | Heads within 0.04 m |
| Rule-based controls (Net1, four rules) | EPANET 2.2 | Heads within 0.008 m |
| Emitters (leakage), Net1 over 24 h | EPANET 2.2 | Heads 0.008 m, demands 0.003 L/s |
| Pressure-driven demand, Net1 over 24 h | EPANET 2.2 | Heads 0.007 m, demands 0.005 L/s |
| Chlorine, Net1 | EPANET 2.2 | Within 0.003 mg/L |
| Water age, Net1, Net2, Net3 | EPANET 2.2 | 95 % within 0.2 h |
| Calibration | Synthetic field data | Recovers roughness ×3 and demand ×1.2 |
The model from your drawings, checked too.
KY4 — a real distribution network of 959 junctions and 1,156 pipes with 2,812 bends, 2 pumps, 4 tanks and a reservoir — was drawn as a CAD plan in feet, with level, size and water-level texts, symbols and clutter. SurgeSpace read it back and was compared item by item with the network it was drawn from.
| Check | Reference | Result |
|---|---|---|
| Nodes and pipes found between the right nodes | KY4 source network | 99.8 % and 99.5 % |
| Levels, sizes, lengths along the bends | KY4 source network | 99.8 % within 1 cm; 99.5 %; 99.8 % |
| Steady heads of the model read from the plan | An independently built model | Median 0.03 m; 95 % within 0.29 m |
| Network from schedules | EPANET Net3 as Excel schedules | Heads within 0.0004 m |
| Pump curve digitised by hand | A published curve: 2,000 gpm at 80 ft | Read as 79.8 ft |
| Pump curve fitted to points | An exact quadratic | Recovered exactly |
The rest of the KY4 differences are where pipes in the source data pass within a metre of a junction they do not join, which a drawing cannot tell apart. Methods and the full list: Theory and validation.
Files other programs trust.
| Check | Reference | Result |
|---|---|---|
| DXF export | ezdxf (independent reader) | No audit errors |
| QGIS plug-in | QGIS 3.34 | Export, import and results layers |
| UTM conversion | PROJ | Within 0.2 mm |
| EPANET .inp round trip | EPANET 2.2 | Read and written, with vertices |