Switching arc simulation & high-voltage consulting

An independent CFD code for switching arc simulation.

A failed interruption test costs a high-power laboratory slot and weeks of project delay. NAVARC simulates the milliseconds that decide the outcome — arc, gas blast, nozzle ablation, moving contacts — before the hardware exists.

Discuss a case See the reference result

The mark above is not an illustration: it is six isotherms of a computed arc temperature field, extracted by marching squares. The two hottest levels carry the warm colour — on this site, colour only ever marks data.

Reference result — a 12 ms self-blast interruption duty

A complete breaker duty in one continuous calculation: a self-blast breaker filled with CO₂ + 20% O₂ (an SF₆-free quenching gas) at 8 bar, compression volume pre-charged to 10.5 bar, with PTFE nozzle ablation, three force-coupled valves and all arc models active. The arc rides the 1 ms current zero unaided through a full 56.7 kA loop, interrupts cleanly at the 11 ms zero, withstands the recovery voltage, and the calculation verifies a further millisecond of cooldown.

56.7 kA

peak fault current, 50 Hz

23–27 kK

radiation-limited arc core

75 bar

peak pressure under the self-blast build-up

11 ms

clean interruption at the second zero, recovery voltage held

+0.9 %

mass balance over 12 ms (= 4.9 g of injected ablation vapour)

Six temperature-field snapshots of the 12 ms duty: first loop, 56.7 kA peak, declining loop, approach to the 11 ms zero, just interrupted, cooldown — Temperature field at six moments (log scale, solids in grey): the first loop, the 56.7 kA peak, the declining loop, the thinning column approaching the 11 ms zero, the interrupted gap under recovery voltage, and the cooldown. Note the assembly motion between frames.

Time-history panels: current, arc voltage, peak temperature, peak pressure and PTFE ablated mass over the 12 ms calculation — Electrical and thermal history from the run's monitoring log. The current rides the 1 ms zero (dotted) and is interrupted at 11 ms (orange); the arc voltage runs at hundreds of volts to the kV range during arcing, then jumps to the 8 kV recovery clamp; the core holds its 23–27 kK plateau; peak pressure reaches 75 bar near the current peak; ablated PTFE mass climbs to 4.9 g, its slope following |I|. One clean interruption, no spurious extinctions, monotonic post-zero decay.

What that buys in practice

Design findings before hardware

In the featured self-blast configuration, the calculation showed that a check valve backed by its support has unequal wetted areas front and back — it needs about 23 % overpressure to open, exactly as the real part would; that number now sets the pre-charge. An earlier relief-valve preload would never have cracked at all: the wetted-area force balance surfaced it, and the valve was retuned before any hardware existed.

Overnight design iterations

The full 12 ms interruption duty runs in roughly 12 h on a 4-core laptop — one design variant per night, no cluster, no licence server. A geometry change is a re-export from CAD, not a new meshing project.

Valve lift versus time and the mean pressures of the heating and compression volumes over the 12 ms self-blast duty — Self-blast chain, resolved end to end: arc heating → ablation → pressure build-up → valve dynamics → blast flow. The thermal check valve breathes under the pre-charge imbalance (peak lift 1.3 mm) and re-seats at the Δp reversal at 3.6 ms; the discharge relief and fill check valves stay correctly seated throughout the duty. The heating volume peaks at 24.6 bar while the piston stroke raises the compression volume to 23 bar — the self-blast pressure reserve for the recovery phase. Valve behaviour is an output of the calculation, not an input.

What the code models

Compressible real-gas flow

2D axisymmetric Navier–Stokes with real-gas thermodynamics for SF₆, CO₂/O₂, N₂, air and C₂F₄ up to 100 kK — computed by an in-house 21-species equilibrium property pipeline (composition, transport, radiation), validated point-by-point against published tables.

Arc electrical model

Imposed current waveform, conductance-sliced arc column, series arc voltage, and an extinction / re-ignition model that mirrors how a real breaker recovers voltage at current zero.

Radiation & self-pinch

Net-emission-coefficient radiation with re-absorption in the cooler fringe and at walls — occlusion-aware — plus the magnetic self-pinch force of the column.

PTFE nozzle ablation

Wall heat flux drives self-limiting vapour blow-off; the vapour is injected hot with its mass, momentum and enthalpy, and blends into the gas as a fully composition-aware C₂F₄ species — including its exothermic oxidation in CO₂/O₂.

Moving & eroding geometry

Contacts and nozzles move on a fixed Cartesian grid with cut-cell embedded boundaries — no remeshing as parts translate or walls recede. Geometry comes straight from CAD assemblies.

Force-coupled valves

Spring-loaded and free check valves as 1-DOF bodies driven by the pressure integrated over their actual wetted faces — including valves that ride on moving carriers.

Built like an instrument

NAVARC is written in Rust — memory-safe by construction, with a zero-unsafe policy and a deliberately small dependency footprint (five runtime libraries; no legacy black box, no licence server). Finite-volume HLLC fluxes, second-order MUSCL reconstruction and SSP-RK3 time integration, with adaptive mesh refinement, SIMD kernels and multi-core parallelism where it counts.

Mass, momentum and energy are conserved to machine precision — including across refinement boundaries and moving walls — and every numerical scheme is guarded by property tests on its conservation invariants: 489 automated tests in the current build. Runs are checkpointed and restartable; every output is plain VTU/CSV, readable in ParaView or any tool you already use.

Verification

Every building block is verified against analytic solutions and published benchmarks, with exact conservation as the standing requirement. Representative results:

What was tested	Result
Shock-tube (Sod) density accuracy	≈ 1.4 % L2 error
Piston-driven pressure rise vs acoustic prediction	within 0.5 %
Viscous shear decay vs analytic solution	within 0.8 %
Closed-box mass drift over 500 steps	≈ 2×10⁻¹⁴
Two-level mass & energy conservation across refinement	≈ 10⁻¹⁷
PTFE ablation mass balance (solid removed = vapour injected)	machine precision
Arc source energy-balance closure (Joule + radiation)	≈ 10⁻¹²
Gas property data vs published CO₂ tables and Saha checks	20/20 anchor checks

Where it stands — stated plainly

Simulation earns trust by declaring its own boundaries. The current build:

2D axisymmetric — no kink or azimuthally asymmetric arc motion.
Local thermodynamic equilibrium composition; no electrode-sheath physics (the standard assumption for this class of circuit-breaker CFD).
Laminar — no turbulence model, a deliberate choice consistent with most published circuit-breaker CFD.
Imposed current waveform, with the network represented by a recovery-voltage ceiling at interruption.
Correlation-based transport and radiation data, with quantified uncertainty (tens of percent near the transport peaks; a factor 2–3 on net emission — typical of the method).
Verified to date against analytic solutions and published benchmarks; quantitative validation against test measurements is the next step — and an open invitation to laboratories holding pressure and arc-voltage records.

The workbench

Cases are set up in a typed, schema-checked browser workbench — every parameter documented in place, validated before launch. Runs report live electrical and thermal telemetry, and results are browsed as field snapshots and multi-run comparisons.

Case setup tab: schema-generated configuration form with units and inline documentation — Case setup — typed configuration, validated before launch. No meshing step.

Monitor tab: live plots of current, arc voltage, peak temperature, time step, mass and conductance during a run — Live monitoring — current, arc voltage, peak temperature, conservation, time step.

Results tab: two-dimensional temperature field of the breaker cross-section with time slider and multi-run comparison — Results — field snapshots with time slider, and metric overlays across runs.

Behind the code

NAVARC is developed by Karim Bousoltane — twenty years of high-voltage circuit-breaker development, from arc physics and interruption testing to R&D leadership, across major equipment manufacturers. The solver is the working instrument of an independent consulting practice: arc simulation studies, interruption-performance assessment, and expert support to breaker development teams.

The name reads Navier + arc: the equations, and the object they serve.

Ways of working

Simulation studies on your geometry — CAD in, quantified interruption behaviour out.
Development support — nozzle and chamber design, gas choice (SF₆ and SF₆-alternatives), test preparation and failure analysis.
Independent expert review — second opinion on designs, simulations and test results.

Contact

The most useful first exchange is a concrete case: a geometry, a gas, a test that behaved unexpectedly. Confidential by default.

karim@navarc.fr

Karim Bousoltane · Lyon area, France · Interventions on site or remote, English and French.