Resistance Spot Welding: The 4,000 Invisible Welds in a Car Body, the 430 J That Melts a 5.5 mm Nugget, and the 2.4 kA Window AHSS Keeps Narrowing

Engineering guide to resistance spot welding: I²Rt heat generation, dynamic resistance, why only 13% of input energy melts the nugget, 4√t nugget criteria, weld lobe construction, RWMA Class 2 CuCrZr electrode life on galvanised AHSS, liquid metal embrittlement, AC vs MFDC, and the per-weld cost stack for a 4,000-weld body-in-white.

·

Resistance Spot Welding: The 4,000 Invisible Welds in a Car Body, the 430 J That Melts a 5.5 mm Nugget, and the 2.4 kA Window AHSS Keeps Narrowing

India built 34,708,984 vehicles in FY2025-26. Passenger vehicle wholesales alone reached 4,643,439 units, up 7.9%, and total vehicle exports rose 24% to 6,647,685 units, according to SIAM's April 2026 release. Almost every one of those steel bodies — hatchback, SUV, truck cab — is assembled by a joining process that leaves no visible seam, no filler metal, no shielding gas and no weld bead: two copper electrodes squeeze the sheets, 8,000–10,000 A flows for 200 ms, and a lens of molten steel forms between the panels and solidifies before the electrodes open.

A modern body-in-white carries roughly 3,000–6,000 of these welds, placed by robots at about 1.5 s per spot. They are never seen by the customer, never painted over as a feature, and almost never inspected non-destructively on the line. Quality is proven by tearing the weld apart and measuring the button.

That is the whole engineering problem in one sentence: the most safety-critical joint in the vehicle is also the one you cannot look at.

The circuit: seven resistances, one of which matters

A spot weld is a series circuit. Current leaves the transformer, passes through the gun arm, the electrode cap, the electrode-to-sheet interface, the top sheet, the faying interface between the sheets, the bottom sheet, the second interface, the second electrode, and back. Each element has a resistance, and Q = I^2Rt sums them.

Element · Typical resistance · Character

Electrode bulk + shank · 10–30 µΩ · Copper, low and stable

Electrode-to-sheet interface · 30–100 µΩ · Drops as asperities collapse

Sheet bulk (per 1.0 mm) · 30–60 µΩ · Rises steeply with temperature

Faying interface (sheet-to-sheet) · 100–1,000 µΩ · Dominant at t = 0; vanishes once molten

Cables, shunts, gun frame · 10–50 µΩ · Fixed loss

The faying interface dominates at the start of the cycle because two mill-finish or zinc-coated surfaces touch only at asperity peaks. The real contact area is a small fraction of the apparent area under the electrode face, so the constriction resistance is high and the first few milliseconds of current heat exactly the place that needs to melt.

As the asperities collapse and the oxide film breaks down, contact resistance falls. Bulk resistance rises because steel's resistivity climbs with temperature — from about 1.4 \times 10^{-7}\ \Omega\cdot m at 20 °C to roughly 1.4 \times 10^{-6}\ \Omega\cdot m at 1,500 °C, an order of magnitude. The process crosses over: what began as interface heating becomes bulk heating, and the hottest point moves from the faying plane into a lens around it.

Plotted as dynamic resistance against time, the curve has a signature that every weld engineer learns to read: an initial spike and sharp drop (film breakdown), a rise as bulk resistivity climbs, a peak, then a fall as the nugget melts and grows because molten steel is a larger conductor than the solid path it replaced.

The diagnostic value is in the deviations. A curve that never peaks means the weld never melted. A sudden mid-cycle collapse means expulsion. A curve that peaks early and then rises again means the electrode face has mushroomed and current density has fallen below the melting threshold. Modern controllers sample current and voltage every millisecond, compute R(t) in real time, and can reject a bad weld before the gun opens.

Where the heat goes: 13% melts metal, the rest is loss

Take a representative automotive joint: two 1.2 mm galvanised sheets, a 5.5 mm nugget, 9 kA, 200 ms, total circuit resistance around 200 µΩ.

That is 16.2 kW delivered for a fifth of a second, at a secondary voltage of about 1.8 V. Now compute what it takes to actually melt the nugget. A 5.5 mm diameter lens about 1.68 mm thick has a volume of 39.9 mm³ and a mass of 0.313 g:

Heating that mass from 20 °C to 1,500 °C at an average specific heat of 0.75 J/g·K, then adding the latent heat of fusion of 270 J/g:

So of the 3,240 J dissipated in the circuit, about 430 J — 13.3% — ends up as molten weld metal. The rest is conducted into the water-cooled copper electrodes, spread into the surrounding sheet, and lost as radiation and convection from the hot surfaces. This is not inefficiency to be engineered away; it is the mechanism that keeps the nugget from growing without bound. The electrodes are a heat sink by design. If they stop removing heat, the nugget grows until it breaks through the sheet surface and the weld expels.

The ratio also explains a counter-intuitive production fact. Shortening the weld time does not simply reduce energy proportionally, because conduction losses scale with time while I^2Rt scales with time at fixed current. Published weld-schedule data for 0.8 mm steel shows roughly 3.5 kJ consumed at 400 ms and about 500 J at 50 ms for the same joint — a seven-fold reduction in energy for the same nugget. The short, high-current pulse puts heat in faster than the electrodes can pull it out. That is the entire argument for MFDC and for high-current short-time schedules.

It is also the argument for restraint. Heat generation goes as the square of current. A 10% current error is a 21% heat error; a 20% error is 44%. Current is the strongest lever on the machine and the one most worth calibrating.

The four knobs, and what they actually control

A spot welding schedule is four numbers: current, time, force, and electrode geometry. Everything else — coating, fit-up, shunting, electrode wear — modulates those four.

Current density is what melts metal, not current. A 6 mm electrode face has an area of 28.3 mm², so 9 kA is 318 A/mm². The same current through a face mushroomed to 8 mm is 179 A/mm², which may be below the melting threshold. This single number explains most "the welds went small at the end of the shift" complaints.

Force does three things: collapses asperities to create a defined contact patch, contains the molten nugget against internal pressure, and forges the solidifying weld. A 3 kN force on that 6 mm face is 106 MPa; 5.8 kN is 205 MPa. Too little force gives high contact resistance, early melting, and expulsion. Too much force gives a larger contact area, lower current density, and an undersized nugget unless current is raised to compensate.

Time is quantised by the power supply. A 50 Hz AC controller can only switch at zero crossings, so the minimum increment is one cycle — 20 ms. A 60 Hz machine gives 16.7 ms. An MFDC inverter at 1 kHz gives 1 ms. That difference matters when the entire weld is 120 ms long.

Geometry — tip diameter, tip shape, radius — sets current density and access. ISO 5821:2025 standardises female caps with a 1:10 taper, type B0 caps at 16 mm shank diameter and 6, 7 or 8 mm face diameter. A 16/6 cap on 1.2 mm sheet is the default automotive choice; going to 16/8 widens the current path and softens the thermal gradient, at the cost of current.

Parameter · 0.8 mm mild · 1.2 mm mild · 1.2 mm DP980 · 1.6 mm DP980

Current · 6–8 kA · 8–10 kA · 4.9–7.3 kA · 7–10 kA

Weld time · 8–10 cycles · 10–14 cycles · 12–15 cycles (250 ms) · 14–18 cycles

Electrode force · 1.8–2.5 kN · 2.5–3.5 kN · 3.5–4.0 kN · 5.0–5.8 kN

Nugget target · 3.6 mm · 4.4 mm · 4.4 mm · 5.1 mm

The AHSS numbers are lower in current and higher in force than the mild steel numbers at the same thickness, which is the opposite of what most people assume. Higher resistivity means less current is needed to reach melting; higher yield strength means more force is needed to establish contact and contain the nugget.

The weld lobe is the only diagram that matters

A weld lobe plots current against time at fixed force, with two boundaries. The left boundary is the current at which the nugget just reaches the minimum acceptable diameter. The right boundary is the current at which expulsion begins. Everything between is the usable window.

The window is not wide. For 1.6 mm hot-dip galvannealed dual-phase steel, published weldability studies measured a useful current range of 2.2 kA for DP 340/590 and DP 420/780, and 2.5 kA for DP 550/980. For 1.2 mm uncoated DP980 at 250 ms and 400 kgf, the window ran from 4.9 kA to 7.3 kA — 2.4 kA. Set the machine at 6.5 kA and you have ±1 kA of margin, which is roughly the drift you get from electrode wear over a few hundred welds.

Two levers widen the lobe, and the AHSS Guidelines quantify both:

This is why AHSS schedules use more force and longer times than the mild steel schedule they replace, even though they use less current. The force is not there to squeeze harder for its own sake; it is there to buy process margin.

Heat balance is the second constraint. When the two sheets differ in thickness or grade, the nugget grows preferentially in the higher-resistivity sheet, and the weld sits off-centre. The practical rule is that thickness ratios up to 2:1 weld acceptably on AC; beyond 2:1, DC power sources and modified schedules give better control. Above 3:1, the stack-up is usually redesigned. Three-layer stacks are routine in body structures but need their own lobe — the middle sheet is a heat sink and a shunt path at the same time.

Nugget size: the arithmetic of 4√t

The universal acceptance criterion is a minimum nugget diameter expressed as a multiple of the square root of sheet thickness:

with d in mm and t in mm. AWS D8.1M, SAE and ANSI all use this form. ISO 14373:2024 grades results in bands at 3.5\sqrt{t} and 5\sqrt{t}, and European aluminium practice (VDA 238-401, ISO 18595) uses 4\sqrt{t} minimum with 5\sqrt{t} nominal.

Sheet thickness · 3.5\sqrt{t} · 4\sqrt{t} (min) · 5\sqrt{t} (nominal) · Nugget area at 4\sqrt{t}

0.7 mm · 2.93 mm · 3.35 mm · 4.18 mm · 8.8 mm²

0.8 mm · 3.13 mm · 3.58 mm · 4.47 mm · 10.1 mm²

1.0 mm · 3.50 mm · 4.00 mm · 5.00 mm · 12.6 mm²

1.2 mm · 3.83 mm · 4.38 mm · 5.48 mm · 15.1 mm²

1.6 mm · 4.43 mm · 5.06 mm · 6.32 mm · 20.1 mm²

2.0 mm · 4.95 mm · 5.66 mm · 7.07 mm · 25.1 mm²

3.0 mm · 6.06 mm · 6.93 mm · 8.66 mm · 37.7 mm²

The square-root form deserves a moment's scepticism. Equating the shear capacity of the nugget to the tear-out capacity of the sheet around it gives a linear relationship:

With \tau_n \approx \sigma_s, that is d = 4t — 4.8 mm for 1.2 mm sheet, close to the 4.38 mm from 4\sqrt{t}. But at 0.5 mm the two forms diverge badly: 4\sqrt{t} gives 2.83 mm (5.7× the thickness), and at 3.0 mm it gives 6.93 mm (2.3×). The \sqrt{t} relation came from riveting practice, and different material families fit better with different exponents — published correlations give d \approx 1.9t for HSLA, 3.2t for stainless and 2.9t for aluminium. The standards use 4\sqrt{t} because it is conservative in the automotive thickness band, not because it is a law of nature.

Strength follows the same geometry. AWS C1.1 gives a minimum shear-tension strength for a spot weld as a function of base metal tensile strength S (MPa) and thickness t (mm):

Grade and thickness · ST_{\min} · Nugget at 4\sqrt{t}

340 MPa, 0.8 mm · 1.82 kN · 3.58 mm

590 MPa, 1.2 mm · 6.20 kN · 4.38 mm

780 MPa, 1.2 mm · 8.29 kN · 4.38 mm

980 MPa, 1.6 mm · 15.70 kN · 5.06 mm

1180 MPa, 1.6 mm · 17.82 kN · 5.06 mm

1480 MPa, 2.0 mm · 26.17 kN · 5.66 mm

Note what the formula implies: doubling strength at fixed thickness nearly doubles the required load, but the nugget diameter does not change. The weld is not made stronger by the base metal; it is made more highly stressed by it. That is why ultra-high-strength grades fail at the weld interface rather than by pull-out, and why "interfacial failure" is a rejectable failure mode in most automotive specifications regardless of the load carried.

The distinction between weld diameter and nugget diameter matters when reading a test report. The button measured on a peeled coupon is the weld diameter and is typically larger than the metallographic nugget diameter measured on a cross-section, because the fracture follows the heat-affected zone rather than the fusion boundary. Quoting one against a specification written for the other is a common and expensive mistake.

Electrodes: the consumable that sets the cost

The electrode is where the economics live. Automotive caps are CuCrZr — 0.5–1.2 wt% chromium, 0.03–0.3 wt% zirconium, precipitation hardened. Hardness runs 70–160 HV, electrical conductivity up to 75% IACS, and softening resistance holds to about 500 °C. That combination is the compromise: pure copper conducts better but deforms under 3–6 kN at 400 °C; beryllium copper holds hardness better but is expensive and a health-and-safety problem in grinding dust.

The cap is water-cooled through an internal chamber, typically at 4–6 L/min. Even so, the cap itself dissipates real power: at 9 kA through 20–40 µΩ of cap and interface resistance, that is 1.6–3.2 kW concentrated in a 16 mm diameter copper part. If the water stops, the cap anneals, mushrooms and the weld goes small within a few hundred cycles.

Electrode wear has a signature. The face grows from 6 mm to 8 mm, current density falls by 44%, and the nugget shrinks until it drops below 4\sqrt{t}. The controller compensates by stepping current up — typically 2–4% per few hundred welds — until the machine reaches its ceiling. Then the cap is dressed back to diameter on the gun, or replaced.

Published life figures for galvanised sheet are blunt: 300–500 welds without surface cleaning or current stepping, extendable to 2,500 welds with tip dressing and current stepping. Uncoated steel does better, often 1,000–2,000 welds between dresses.

Cap price · Life · Cost per weld · Cost per 4,000-weld body

₹180 · 400 welds · ₹0.45 · ₹1,800

₹300 · 500 welds · ₹0.60 · ₹2,400

₹300 · 600 welds · ₹0.50 · ₹2,000

₹300 · 2,500 welds · ₹0.12 · ₹480

₹450 · 2,500 welds · ₹0.18 · ₹720

The gap between ₹2,400 and ₹480 per body is not a purchasing decision. It is a dressing-discipline decision, and on a 60 JPH line it is worth more than the electricity.

Zinc, aluminium, and the coating problem

Zinc coating changes the electrical boundary condition. Zinc has roughly one-sixth the resistivity of steel and a lower melting point (419 °C), so a galvanised stack has lower contact resistance and needs more current and longer time to reach the same nugget — typically 10–25% more current than the same gauge uncoated.

The metallurgical problem is worse than the electrical one. At welding temperature, zinc alloys with the copper electrode face and forms brass phases. Those phases stick to the cap and are torn off when the gun opens, pitting the face and accelerating erosion. The severity depends on the coating: Zn-Al coatings (55% Al-Zn, sold as ZINCALUME-type) are the most aggressive, followed by hot-dip galvanised, then galvannealed, then zinc-iron alloy coatings, which are the gentlest because the iron-rich layer is already stable.

Then there is liquid metal embrittlement. In galvannealed AHSS, liquid zinc can penetrate austenite grain boundaries in the heat-affected zone and crack the sheet — a failure that appears only after welding and sometimes only after forming. Nippon Steel's published work on 1.6 mm 980 MPa galvannealed sheet found that with a standard 6 mm electrode tip, LME cracks appeared at currents above 9.5 kA regardless of expulsion. With 8 mm and 10 mm tips, no cracks formed across the entire current range tested. Widening the tip lowered the thermal gradient enough to keep the surface out of the critical temperature window.

The countermeasures are all thermal: larger tip diameter, longer hold time so the coating resolidifies under load before the electrode releases, lower current with longer time, a tempering post-pulse, or a change of coating. What does not work is welding faster and hotter and hoping the crack stays subcritical.

Aluminium is a different process wearing the same name. Resistivity is low, thermal conductivity is high, and the oxide layer is refractory, so aluminium RSW needs 20–40 kA, weld times of 1–3 cycles, domed tips, and often a mid-frequency DC supply to hold current through the fast-rising resistance. Electrode life is measured in hundreds, not thousands, of welds. That economics is why aluminium bodies use self-pierce riveting and adhesive bonding for most joints and reserve RSW for a minority — each process adds either a consumable or a cycle-time penalty, and the body shop picks whichever costs less per joint at the required strength.

AHSS: same weld, one-third the window

Advanced high-strength steels are the reason spot welding got hard again. Dual-phase, TRIP, complex-phase, martensitic and quench-and-partition grades all have higher alloy content and higher resistivity than mild steel. DP and TRIP sheet resistivity can be twice and three times that of mild steel, which means the nugget forms at lower current.

The catch is everything else. Yield strength is higher, so electrode force must rise — the AHSS Guidelines recommend starting at 20% above the mild steel force for the same gauge and increasing from there. Ductility is lower, so the nugget cannot deform as far before the surrounding sheet tears, which pushes failure toward the interface. The weld lobe narrows from both sides: the left boundary moves left because resistivity is higher, and the right boundary moves left faster because the plastic shell containing the nugget is stronger but less ductile.

A 1.6 mm quench-and-partition 980 grade, for example, needed 5.8 kN of force to produce a stable lobe with full-button fracture. That is nearly double the force a 1.6 mm mild steel joint would use.

There is a second, subtler problem in the heat-affected zone. Welding tempers the martensite in the subcritical HAZ, producing a soft band — sometimes 200–250 HV against a 450–500 HV base metal in hot-stamped boron steel. In a crash, that soft band is where the component bends. It does not fail the weld; it changes the crash mode of the part. Body engineers model it explicitly, and the weld schedule that produces the best-looking nugget is not always the one that produces the best crash behaviour.

Practical AHSS schedule changes, in the order they are usually applied:

  1. Increase electrode force by 20% or more, proportional to yield strength.
  2. Increase weld time rather than current.
  3. Move to a multi-pulse schedule — several current pulses with cooling intervals, or a post-weld temper pulse.
  4. Increase tip diameter, which lowers current density and thermal gradient.
  5. Increase the minimum nugget size, if the specification allows.

AC or MFDC: the power supply decides your resolution

A single-phase AC gun runs at line frequency. The current is a train of half-cycles, the transformer is large, the power factor is around 0.6, and each half-cycle is 10 ms at 50 Hz. Control resolution is a cycle: 20 ms. Reactive losses in the gun arm grow with throat depth, so a long C-gun on a big panel wastes more energy than a short one.

An MFDC supply rectifies three-phase mains, chops it at about 1,000 Hz with IGBTs, steps it down through a much smaller transformer, and rectifies the secondary to DC. The output has no zero crossings, the power factor is above 0.95, and the control resolution is 1 ms. The transformer is small enough to mount on the gun, which is why modern servo guns weigh 80–120 kg and ride on 165 kg-payload robots rather than 250 kg ones.

The measured differences are consistent across independent studies: AC draws roughly 10% more total energy than MFDC for the same weld size, and at low current the weld sizes diverge more because the AC waveform spends part of each half-cycle below the melting threshold. Vendor claims of 30–50% energy savings usually compare a new MFDC installation against an old AC line with long cables and poor power factor — real, but not a like-for-like machine comparison.

Measuring what you cannot see

Production quality control rests on three layers.

Destructive testing. Every shift, a body or a coupon set is taken to the lab and the welds are tested:

In-process monitoring. Current, voltage, force and electrode displacement are logged for every weld. Dynamic resistance is the classical signal; electrode displacement is often more sensitive to expulsion and to fit-up gaps. Adaptive controllers close the loop within the weld, adjusting current in real time to hit a target resistance trajectory. Modern lines log a weld ID against the body serial number, so a field failure can be traced back to the exact schedule, gun and electrode used.

Statistical control. Nugget diameter is tracked as a Cpk against the specification. IATF 16949 practice sets 1.33 as the general minimum and 1.67 for safety-critical joints identified in the DFMEA. A line running at Cpk 1.0 is producing roughly 0.3% undersized welds — 12 bad welds in a 4,000-weld body, scattered where nobody will find them until a crash test or a warranty claim does.

Design rules for parts that will be spot welded

Most spot welding problems are designed in, not welded in. The rules that matter at the drawing stage:

The per-weld cost stack

A spot weld looks almost free — no wire, no gas, no consumable per joint except the electrode. The capital and the consumables tell a different story. Using a robotic cell with a servo gun, a 4,000-weld body and a 60 JPH line:

Cost element · Basis · Cost per weld

Cell capital · ₹1.2 crore cell, 8-year life, 6,000 h/yr, 1,200 welds/h · ₹0.21

Electrode caps · ₹300 cap, 600 welds · ₹0.50

Labour · 0.2 operator per cell at ₹400/h · ₹0.07

Maintenance · 5% of capital per year · ₹0.05

Energy · 3.2 kJ per weld at ₹8.5/kWh · ₹0.008

Direct total · · ₹0.84

That puts the welding cost of a 4,000-weld body at roughly ₹3,400, of which the copper is more than half. The number scales with wage rates and cell utilisation, but the structure does not: electrode life is the largest controllable line item, and it is set by coating, cooling, force and dressing discipline.

The line arithmetic is equally unforgiving. Four thousand welds at 1.5 s each is 6,000 gun-seconds per body. On a 60 JPH line — one body every 60 s — that requires 100 welding guns working continuously, which is why a body shop has 400–600 robots and why a 5% cycle-time overrun at one gate stage propagates down the whole line. A plant with 400 guns also carries a substation sized in the tens of megavolt-amperes, because a 9 kA weld at 1.8 V is 16 kW instantaneous and the transformer feeding it is rated 100–200 kVA on a duty cycle of a few per cent. Power-factor correction is a capital project, not a utility bill.

What India's AHSS shift changes

The material transition is already visible in the supply chain. Tata Steel's Kalinganagar hot strip mill rolls up to 1,200 MPa, and its JCAPCPL continuous annealing line in Jamshedpur supplies cold-rolled automotive grades up to 590 MPa with capability to 980 MPa. JSW Steel operates India's largest auto-grade steel complex at Vijayanagar at 2.3 MTPA, developed 182 new products and grades in FY2025-26, and has a 0.4 MTPA continuous galvanising line dedicated to high-strength automotive grades commissioning by June 2028. In August 2026 JSW ordered a 500,000 t/yr ANDRITZ galvanising line at Toranagallu for strip up to 1,900 mm wide and 2.60 mm thick, including AHSS grades to 1,480 MPa, with start-up in 2028.

Two forces are pulling that investment. Bharat NCAP's crash protocols reward body structures that absorb energy without adding mass, and every kilogram removed from the body is a kilogram of battery or payload gained. The industry estimate cited in JSW's FY26 annual report puts India's automotive steel market at $7.6 billion by 2030, growing at 7.7% a year, with the growth concentrated in exactly the grades that are hardest to weld.

For a fabrication business, the opportunity is not in making the steel. It is in the ecosystem the steel creates:

None of that is a "quote and ship" business. An automotive customer will not accept a new fixture or a new weld schedule without PPAP, IATF 16949 evidence and months of capability data. The barrier is qualification, and qualification is where the margin is.

The specification checklist

If you are quoting a spot-welded assembly, or buying one, these are the numbers that belong on the drawing and the RFQ:

A spot weld is a 200 ms thermal event inside a 0.3 g volume of steel, measured after the fact by tearing it apart. Everything above is an attempt to make that event repeat four thousand times a day, on a coated high-strength alloy, at the edge of a process window that keeps getting narrower. The industry has spent fifty years getting good at it. The next fifty are about doing it with a third less steel and a third more strength — which, in practice, means a third less margin.

More FabFlow blog posts