Wire EDM

Cutting metal with a controlled lightning storm

Every conventional machining process works by force. A milling cutter, a lathe tool, a drill bit, a saw blade — each one pushes a hard wedge of carbide or high-speed steel into the workpiece and shears chips off. That logic has one fatal limit: the tool has to be harder than the work, and it has to survive the contact force. Try to mill a fully hardened die at 62 HRC and the cutting edge chips on the first pass; try to drill a 0.3 mm slot through a 100 mm block and the bit snaps.

Wire EDM throws out the force model entirely. The "tool" is a thin wire — usually 0.25 mm brass — that never touches the metal it is cutting. Instead, the wire and the workpiece are connected to a pulse generator and submerged in a dielectric fluid, almost always deionized water. A CNC servo holds the wire a few microns from the metal. When the voltage across that gap gets high enough, the dielectric breaks down and a spark jumps across, exactly as lightning jumps between a cloud and the ground. That single spark — a plasma channel hotter than the surface of the sun for a few microseconds — melts and vaporizes a microscopic crater of metal. Then the spark collapses, the flowing water flushes the molten debris out of the gap and re-insulates it, and the next spark fires. Repeat that spark erosion cycle a hundred thousand to a million times a second and the wire eats its way through the metal along whatever path the CNC commands.

The headline consequence is that hardness is irrelevant. A spark does not care whether the metal is annealed aluminium or 65 HRC powder-metallurgy tool steel; it melts both the same way, governed only by electrical conductivity and melting point. That single fact is why wire EDM exists as an industry: it is the process you reach for after a part is hardened, when its final precise geometry has to be cut and nothing with a cutting edge can survive.

The discharge cycle, spark by spark

A single discharge has four phases, and a good machine controls each one independently:

• Ignition. The generator raises the open-circuit voltage across the gap to 60–300 V. The electric field at the narrowest point of the gap ionizes a column of dielectric and any conductive debris suspended in it, forming a plasma bridge.
• Discharge. Once the bridge forms, voltage collapses to a low arc voltage (~20–30 V) and current surges — peak currents of tens to hundreds of amps for a few microseconds. The plasma channel reaches roughly 8,000–12,000 °C, melting and partly vaporizing metal on both the wire and the work.
• Ejection. The vapour bubble around the channel collapses when the current shuts off; the implosion and the rushing dielectric eject the molten pool, leaving a crater a few microns across. A tiny amount of melt re-solidifies on the surface as the recast layer.
• Deionization. The generator holds the gap at zero volts for a controlled off-time so the dielectric recombines and re-insulates. Skip this and the channel persists as a continuous DC arc, which gouges the surface and breaks the wire.

The engineer's control knobs map directly onto these phases. On-time (pulse duration, t_on) and peak current set how much energy goes into each spark, and therefore the crater size: bigger sparks cut faster but leave a rougher surface and a thicker recast layer. Off-time (t_off) governs flushing and deionization: too short and you arc, too long and you waste cutting speed. The ratio of on-time to total cycle time is the duty cycle. Material removal per unit time scales roughly with the energy per spark times the spark frequency:

Energy per discharge:   E = ∫ u(t)·i(t) dt  ≈  U_arc · I_peak · t_on
Crater volume:          V_crater ∝ E^k          (k ≈ 0.6–0.8, empirical)
Material removal rate:   MRR ≈ V_crater · f_spark · η_eject

where  f_spark = 1 / (t_on + t_off)   discharge frequency
       η_eject = fraction of melt actually flushed clear  (0.1–0.3)

The low ejection efficiency — only 10–30 % of the melted metal is actually carried away, the rest re-solidifies — is the central inefficiency of EDM and the reason finish passes are needed. It also tells you why flushing matters so much: better debris removal raises η_eject, and a clean gap fires cleaner sparks.

The wire and the dielectric

The wire is consumable. Each spark erodes the wire as well as the work, so the wire is continuously fed from a spool, threaded through the part, and discarded at speeds of 5–15 m/min — a fresh, undamaged electrode is always in the cut. The wire is held in tension (typically 10–25 N) between upper and lower guides and runs through water nozzles that flush both faces of the cut.

• Plain brass (CuZn37). The workhorse. Brass at 0.25 mm cuts most steels well; the zinc helps cutting speed because it vaporizes readily and stabilizes the discharge.
• Coated ("stratified") wire. A brass or steel core with a zinc-rich outer layer. The zinc skin boosts cutting speed 15–40 % and improves flushing of thick sections; standard for high-throughput production.
• Tungsten and molybdenum fine wire. Down to 0.02–0.05 mm for micro-EDM — cutting watch gears, fuel-injector orifices, and microfluidic features with kerfs narrower than a human hair.
• Hard brass for tall parts. Higher tension capacity resists wire vibration ("bow") in 200–400 mm thick blocks.

The dielectric is deionized water on essentially every modern wire machine. A resin deionizing bed keeps conductivity controlled (commonly 5–25 µS/cm); too conductive and the water leaks current everywhere instead of breaking down only at the gap, widening the kerf and slowing the cut. Water's low viscosity makes it flush narrow kerfs well, and it carries away heat about four times better than the hydrocarbon oils used in sinker EDM — important at the high spark rates wire machines run. The price is electrolysis: water-cut surfaces can corrode and, on tungsten carbide, leach the cobalt binder. Anti-electrolysis (AE) generators apply a bipolar or net-zero-DC waveform to suppress this for carbide and corrosion-sensitive alloys.

Worked example — cutting a stamping die punch

Consider a real job: cut a progressive-die punch profile through a 50 mm thick block of D2 cold-work tool steel, hardened to 60 HRC, to a final tolerance of ±5 µm and a 0.4 µm Ra finish. Here is how the process plans out.

Wire:            0.25 mm coated brass, 18 N tension
Start hole:      Ø1.0 mm EDM-drilled before hardening (wire threads through)
Kerf width:      wire Ø + 2 × spark gap = 0.25 + 2 × 0.025 ≈ 0.30 mm

Rough pass (1):  high energy, ~400 mm²/min, leaves ~2.5 µm Ra, ~10 µm recast
                 offset = kerf/2 + finishing stock = 0.15 + 0.04 = 0.19 mm
Skim pass (2):   medium energy, removes ~25 µm, → 1.0 µm Ra
Skim pass (3):   low energy, removes ~8 µm,  → 0.4 µm Ra, recast < 2 µm

Cut length:      perimeter 120 mm
Rough time:      (120 mm × 50 mm) / 400 mm²/min = 15.0 min
Skim 2 + 3:      ≈ 0.6 × rough each → ~18 min combined
Total cycle:     ≈ 33 min for a finished, hardened, distortion-free punch

Two details carry the whole job. First, the start hole is drilled before hardening, because once the block is at 60 HRC nothing but EDM will make a hole in it — so the process sequence (drill, harden, wire-cut) is locked by the physics. Second, the part is cut after hardening, which is the entire point: heat treatment distorts steel by tenths of a millimetre, so any feature machined before hardening drifts out of tolerance. Wire EDM cuts the final geometry after the distortion has already happened, with zero cutting force to add more.

Where the precision comes from — and its limits

Several things conspire to make wire EDM accurate. There is no cutting force, so the part cannot deflect and the wire cannot push it out of position. The wire is always fresh, so unlike a milling cutter it never wears in a way that drifts the dimension mid-cut. The machine's glass scales and thermally-stabilized structure resolve sub-micron motion. And the multi-pass strategy lets a rough cut do the fast bulk removal while skim passes, offset by a few microns each, walk the surface in to its final size and finish.

The limits are equally physical. The smallest internal corner radius equals the wire radius plus the spark gap — a 0.25 mm wire bottoms out near 0.15 mm internal radius, so a truly sharp internal corner is impossible without a smaller wire. Tall parts suffer wire bow: the flushing pressure and discharge forces deflect the unsupported middle of the wire, so a 300 mm thick cut is slightly barrel-shaped unless tension and flushing are tuned and skim passes correct it. And taper cuts — angling the upper guide relative to the lower — are limited by how far the wire can be deflected before it fatigues and snaps, typically up to ±30° on dedicated taper machines.

Wire EDM versus the alternatives

The honest way to position wire EDM is against the processes a tool shop would otherwise reach for. It is slow in raw volume terms and only cuts conductors — but it owns the hardened, high-precision, zero-force corner of the design space that nothing else can touch.

Property	Wire EDM	CNC milling	Laser cutting	Abrasive waterjet
Mechanism	Spark erosion	Mechanical shearing	Thermal melt/vaporize	Abrasive erosion
Cutting force	Zero	High	Zero	Low (jet reaction)
Cuts hardened steel?	Yes, any hardness	No (chips the tool)	Yes, thin only	Yes
Material limit	Conductors only	Any	Most	Almost any
Tolerance	±2–5 µm	±10–25 µm	±50–100 µm	±100–200 µm
Surface finish (Ra)	0.1–2.5 µm	0.4–3.2 µm	3–12 µm	3–6 µm
Kerf width	0.05–0.35 mm	tool dia.	0.1–0.5 mm	0.8–1.5 mm
Heat-affected zone	2–15 µm recast	Negligible	0.1–0.5 mm	None
Bulk removal speed	Slow	Very fast	Fast (sheet)	Medium
Sweet spot	Hardened precision profiles	3D bulk shaping	Sheet-metal profiles	Thick stacks, any material

Where wire EDM actually shows up

• Stamping and progressive dies. The classic application. Punches and matching die plates in hardened tool steel are wire-cut as a matched pair, with the wire offset adjusted to set the punch-to-die clearance to within a couple of microns — the parameter that decides whether the stamped part has a clean shear or a burr.
• Extrusion dies. Aluminium and plastic extrusion dies have intricate hollow profiles cut through hardened H13 — exactly the through-profile geometry wire EDM is built for.
• Injection-mold inserts. Slides, ejector-pin holes, and sharp parting-line details that sinker EDM and milling cannot reach are finished on the wire.
• Aerospace turbine components. Fir-tree root slots in nickel-superalloy turbine disks and blades, where the alloy is nearly unmachinable by conventional tools and the slot tolerance is a few microns.
• Medical and dental. Surgical-implant features, bone-plate profiles, and dental components in titanium and cobalt-chrome, where burr-free, low-stress edges matter.
• Gears, splines, and timing components. Master gears and form tools cut after hardening, when the involute profile has to be exact and the blank cannot move.
• Micro and fine work. Fuel-injector orifices, watch movements, and microfluidic plates cut with 0.02–0.05 mm wire to features smaller than a human hair.

Failure modes and trade-offs

• Wire breakage. The single biggest production disruptor. Caused by inadequate flushing in a deep cut, too-aggressive energy settings, a hard inclusion in the metal, or dielectric that has drifted too conductive. Modern machines auto-detect a break, retract, re-thread the wire through the kerf with a water jet, and resume — but every break costs minutes.
• Recast layer and micro-cracks. The re-solidified white layer is hard, brittle, and micro-cracked. Harmless on a die, but a fatigue-crack initiator on an aerospace or implant part — these get skim passes and a final polish or etch to remove it.
• Wire bow on thick sections. Flushing and discharge forces deflect the unsupported wire midspan, barreling a tall cut. Controlled with higher tension, better flushing, and corrective skim passes.
• Electrolytic corrosion and cobalt leaching. Water dielectric can rust ferrous surfaces and dissolve the cobalt binder out of tungsten carbide, weakening the cut surface. Anti-electrolysis generators with a net-zero DC waveform suppress it.
• Conductivity drift. If the deionizing resin saturates and water conductivity climbs, current leaks along the whole wire instead of only at the gap — the kerf widens, accuracy drops, and cutting slows. Conductivity is monitored and the resin swapped on schedule.
• The speed/finish trade-off. There is no free lunch: big sparks cut fast but rough, small sparks finish beautifully but slowly. Every extra skim pass roughly doubles cycle time, so production parts are cut to the loosest finish the application allows.
• Conductivity requirement. The hard limit. The work must conduct electricity — wire EDM cannot cut ceramics, glass, plastics, or composites (though it can cut some semiconducting ceramics like certain conductive carbides).