Friction Stir Welding: The 0.48 mm/rev Window, the Invisible Kissing Bond, and the Arithmetic of a 22-Metre Battery Tray

Engineering guide to friction stir welding: heat input, weld pitch, tool geometry, HAZ softening, kissing-bond inspection, ISO 25239, and cost per metre.

·

Friction Stir Welding: The 0.48 mm/rev Window, the Invisible Kissing Bond, and the Arithmetic of a 22-Metre Battery Tray

A 1,300 × 1,700 mm EV battery tray is a flat aluminium box that must do three things at once: hold 400–700 kg of cells, carry coolant at 2–3 bar without weeping into the pack, and keep the cell mounting surface flat enough that every module sits at the same height. The benchmark design uses 17 extrusions in AA 6082-T6, weighs 72 kg, and needs 22 m of leak-tight weld if the floor plate is mechanically interlocked on the inside — 38 m if it is welded from both sides.

The process that makes it is friction stir welding. A non-consumable tool rotates at 600–1,200 rpm, plunges a pin into the joint line until the shoulder bears on the surface, and traverses at 2–4 m/min. The metal reaches 400–500 °C — roughly 70–80% of its melting point — and never becomes liquid. There is no filler wire, no shielding gas, no arc, no fume, and no solidification shrinkage. The weld is made by extrusion and forging, not by casting a joint.

That is why FSW owns the battery tray, the rail car body, the rocket propellant tank, and the aluminium bridge deck. It is also why FSW has a failure mode that fusion welding does not: a root defect measured in hundreds of micrometres, mechanically closed, metallurgically unbonded, and effectively invisible to the radiography that most Indian job shops reach for first.

This is the engineering of the process: where the heat comes from, why the weld pitch is the variable that matters most, what happens to a T6 temper in the heat-affected zone, which defects you can and cannot inspect, and what a metre of FSW actually costs.

1. Where the heat comes from

FSW has four stages: plunge, dwell, traverse, retract. The tool is a single piece of heat-treated steel or PCBN with a shoulder (the flat, usually concave face that bears on the workpiece) and a probe or pin (the feature that enters the joint). The shoulder does two jobs: it generates most of the heat, and it forges the plasticised material back into a consolidated joint.

The mechanical power going into the workpiece is torque times angular velocity:

and the heat input per millimetre of weld is that power divided by traverse speed:

A 6 mm AA 6082-T6 butt weld run at 800 rpm and 400 mm/min with a measured spindle torque of 45 N·m gives:

That is a typical number for 6 mm aluminium: 0.4–0.8 kJ/mm. Thick-section work is much hotter. TWI's 25–50 mm trials on AA 6082 recorded 240 N·m and 65 kN downforce at 300 rpm, with heat inputs of 1.8–4.4 kJ/mm depending on alloy. For comparison, a 6 mm steel MIG weld runs at roughly 0.5–1.0 kJ/mm but deposits filler and melts the parent; the FSW number is achieved with no filler and a peak temperature several hundred degrees lower.

If you model the shoulder as a sliding annulus under uniform pressure P_s, the frictional heat generated between probe radius R_p and shoulder radius R_s is:

The cubic term is why shoulder diameter dominates the thermal field. Going from a 12 mm to an 18 mm shoulder at the same rotation speed more than triples the frictional power, which is useful for thick plate and disastrous for thin sheet: the surface overheats, the root lags, and the weld cups.

In practice, 70–80% of the heat is generated under the shoulder, with the remainder from the probe sides and tip plus plastic dissipation in the shear layer. The partition between tool and workpiece matters too: H13 conducts about 25–30 W/m·K, aluminium 170–200 W/m·K, so most of the heat flows into the aluminium, which is exactly what you want. PCBN tools used on steel have the opposite problem — they conduct heat readily and run hot, which is why PCBN tool life in steel is measured in tens of metres rather than thousands.

Peak temperature is bounded by the solidus. For 6xxx alloys the window is comfortable: solidus around 555–582 °C, typical FSW peak 350–480 °C. For 7075-T6 the solidus is only about 477 °C because of zinc-rich grain-boundary eutectics, so the process window narrows sharply; exceed it and you get liquation cracking in the HAZ that looks like a fusion weld defect in a process that is supposed to avoid melting.

2. Weld pitch is the master variable

Rotation speed and traverse speed are usually tabulated separately, which hides the variable that actually controls material flow. Divide traverse speed by rotation speed and you get the weld pitch — the distance the tool advances per revolution:

At 600 rpm and 300 mm/min, p = 0.50 mm/rev. At 1,200 rpm and 600 mm/min, p is also 0.50 mm/rev, but the heat input is different because torque and strain rate are different. The pitch tells you how thick a layer of plasticised metal the probe must deposit behind itself on every revolution.

For 3.18 mm AA 6061-T6, the experimental optimum is 0.48 mm/rev. Wanjara et al. found that pitch values below 0.12 mm/rev over-stir the nugget — dark banding, heavy flash, plate thinning — while pitches above 0.60 mm/rev produce wormholes and tunnel defects on the advancing side. Liu et al. put the threshold at 0.53 mm/rev for 5 mm 6061-T6. The optimum is not a universal constant, but the shape of the window is: too hot and you over-stir, too cold and you fail to consolidate.

The volume deposited per revolution is the weld pitch times the cross-section of the stirred zone:

For a 6 mm plate with a 12 mm wide stirred zone at 0.5 mm/rev, that is 72 \times 0.5 = 36 mm³ per revolution, or a volumetric transport rate of about 480 mm³/s at 800 rpm. The probe's thread or flute has to move that volume from the leading side to the trailing side every revolution. If the thread pitch is finer than the weld pitch, the thread cannot keep up: material stalls on the advancing side and a void opens. This is the physical reason high weld pitch produces wormholes, and why tool designers match thread pitch to the intended advance per revolution rather than treating the probe as a generic stirrer.

Strain rate in the shear layer scales with tool radius and rotation speed:

where \delta is the effective shear-layer thickness, typically 0.1–0.5 mm. At 800 rpm, R_p = 3 mm, and \delta = 0.3 mm, that is about 840 s⁻¹ — three orders of magnitude above a tensile test, which is why the nugget recrystallises rather than work-hardens.

The Zener–Hollomon parameter ties strain rate and temperature to the resulting grain size:

With Q \approx 142 kJ/mol for aluminium self-diffusion and a peak temperature of 450 °C, Z lands around 10^{13} s⁻¹ and the recrystallised grain size in the nugget falls to 2–10 µm. That refinement is real, and it is why the nugget is often harder than a fusion weld's cast structure — but it does not compensate for the loss of precipitation hardening in the surrounding HAZ.

3. Tool geometry: the 3:1 rule and its exceptions

A conventional FSW tool is defined by five dimensions: shoulder diameter, shoulder profile, probe diameter, probe length, and probe feature. The ratios matter more than the absolute numbers.

Tool material is selected by workpiece temperature, not by hardness alone:

The economic consequence is stark. Aluminium FSW tooling costs ₹15–30 per metre of weld. Steel FSW tooling costs ₹3,000–15,000 per metre. That single ratio explains why FSW is ubiquitous in aluminium and niche in steel, and why the steel FSW papers you read are mostly about tool wear rather than weld metallurgy.

Tool geometry is an essential variable in every FSW procedure specification. ISO 25239-4 requires the tool and the parameter set to be qualified together; a procedure developed with one shoulder diameter cannot be transferred to a nominally identical tool from another supplier without requalification. AWS D17.3 goes further and lists tool geometry, rotation speed, travel speed, tilt angle (±0.5°), plunge depth (±5%), dwell time (±5%), and backing bar as essential variables.

4. The four zones, and why the HAZ fails first

A friction stir weld has four microstructural regions:

  1. Nugget / stir zone (SZ): fully recrystallised, equiaxed grains of 1–10 µm, heavily deformed, temperature 400–500 °C. In 6061-T6 the strengthening precipitates are partly dissolved, so the nugget is softer than the base metal but harder than the overaged HAZ.
  2. Thermo-mechanically affected zone (TMAZ): deformed but not recrystallised, with a sharp transition into the nugget. This is where strain localisation and fracture initiation often begin.
  3. Heat-affected zone (HAZ): no deformation, but enough temperature (250–450 °C) to coarsen the Mg₂Si or MgZn₂ precipitates. This is the weakest region in any precipitation-hardened alloy.
  4. Base metal: unaffected beyond about 20–30 mm from the weld centreline for 6 mm plate.

The HAZ softening is straightforward physical metallurgy. Peak-aged 6061-T6 gets its 276 MPa yield strength from a dense distribution of nanoscale β″ precipitates. Hold it at 350–400 °C for a few seconds and those precipitates coarsen; coarsening increases the inter-particle spacing \lambda, and the Orowan bypass stress falls:

Doubling \lambda roughly halves the strengthening contribution. The HAZ therefore drops from 276 MPa yield to 130–160 MPa, and the joint efficiency — weld UTS divided by base UTS — lands at 70–80% for 6061-T6 and 6082-T6 in the as-welded condition.

Measured numbers from the literature:

The alloy matters as much as the process. AA 5083 is strain-hardened, not precipitation-hardened, so it has no HAZ overaging problem; joint efficiency is close to 100%, and it is the default for shipbuilding and cryogenic tanks. AA 7075-T6 is the opposite: high strength, but a solidus of 477 °C and a HAZ that overages severely, giving 65–75% joint efficiency and a narrow process window. AA 2219 and 2195 Al-Li behave better at cryogenic temperatures and are the reason NASA adopted FSW for the External Tank.

Post-weld heat treatment can recover part of the loss, but be precise about what is achievable. Artificial aging at 175 °C for 8 h re-precipitates some of the solute that was taken into solution in the nugget and can lift joint efficiency by 5–10 percentage points. Full T6 properties require solution treatment at about 530 °C, a quench, and then aging — which is impractical for a 1.7 m battery tray and reintroduces exactly the distortion FSW was chosen to avoid. Design the joint for the as-welded HAZ strength, not for the parent temper.

5. Defects: what you can see, and what you cannot

FSW defects divide into flow-related and geometry-related families. The important ones:

Wormhole / tunnel. A volumetric void running along the weld, usually on the advancing side near the root or mid-thickness. Caused by insufficient heat input, excessive weld pitch, wrong probe thread direction, or inadequate forge force. Detectable by phased-array ultrasonic testing and visible on a macro section. This is the defect everyone learns first.

Kissing bond / root flaw. A planar discontinuity at the original faying surfaces where the oxide film was never broken and no metallurgical bond formed. The surfaces are in intimate contact — hence "kissing" — but there is no grain growth across the interface. Sizes run from 100 to 700 µm. Radiography passes it because there is no density difference. Dye penetrant cannot reach it because it is not open to the surface. It fails in fatigue and in root bend tests, often at loads well below the static joint strength.

Lazy S / joint line remnant. A zigzag line of entrapped oxide following the weld centreline. Below a threshold it is tolerated; connected to the root, it becomes a kissing bond. The distinction matters because the acceptance criteria in ISO 25239-5 and AWS D17.3 treat planar root discontinuities far more severely than volumetric ones.

Lack of penetration (LOP). The probe did not reach the root, leaving unjoined parent metal. Geometry-related, easy to produce with a short pin or a worn shoulder, and easy to see on a macro.

Hook defect. In lap joints, the upward flow of material lifts the interface, creating a sharp oxide-filled hook that reduces the effective sheet thickness. The hook is a stress concentrator; its severity depends on tool direction and probe feature.

Exit keyhole. The tool leaves a hole at the end of every weld. Production parts either run the weld onto a sacrificial runoff tab that is later machined off, or use a retractable-pin tool that withdraws the probe before lifting the shoulder. For hermetic battery trays, a keyhole is a leak path and must be closed — refill friction stir spot welding is the standard repair, with joint efficiencies around 78–82%.

Inspection is where FSW projects fail commercially rather than metallurgically. A typical specification says "radiograph the weld", which is the wrong tool for a planar oxide bond. What works:

The practical rule: specify the defect type and the acceptance standard, not the inspection method. "No planar root discontinuities greater than 0.5 mm, inspected by PAUT to ISO 25239-5 quality level B, with bend tests on the first article" gives a supplier something to comply with. "X-ray the weld" does not.

6. Joint design and the standards that govern it

FSW is a square-butt process. There is no bevel, no root gap, no filler, and no shielding gas. That simplifies joint preparation and complicates fit-up: the tool needs sound material on both sides of the joint line, and the two edges must be within about 0.1–0.2 mm of each other for 3–6 mm sheet. Extruded profiles are usually better than rolled sheet because the dimensional tolerance is tighter and the joint line can be designed into the extrusion.

The standards stack:

Design rules that prevent most field failures:

7. Variants that solve specific problems

Bobbin tool (self-reacting FSW). Two shoulders connected by a probe, one on each side of the workpiece. The axial forces balance inside the tool, so no backing bar is needed and closed profiles — pipes, hollow extrusions, battery tray frames — can be welded. The trade-off is a more complex tool and a keyhole that is harder to close. Linear energy is around 15 kJ/m for a 3 mm aluminium weld, dropping to about 12 kJ/m with a stationary upper shoulder.

Stationary shoulder FSW (SSFSW). The shoulder does not rotate; only the probe does. Heat input drops by about 30%, the HAZ narrows, and the sliding shoulder irons the surface, improving roughness by up to a factor of five. It is the variant of choice for thin sheet, titanium, and any part where post-weld distortion or surface finish is critical.

Refill friction stir spot welding (RFSSW). A three-part tool — clamping ring, sleeve, probe — plasticises a spot, refills the cavity, and leaves a flush surface. Used to replace rivets in skin-stringer joints and to close FSW keyholes. Joint efficiency around 78–82%, which is acceptable for a spot joint but not for a primary butt weld.

Friction stir processing (FSP). The same hardware, used to modify a surface rather than join two parts: grain refinement, homogenisation of cast structures, or local composite fabrication. It is a legitimate production tool for castings and for repairing porosity in aluminium castings.

Robotic FSW. A 6-axis robot with force control and deflection compensation can weld large, curved structures that a gantry cannot reach. The limitation is stiffness: a robot with 0.5 mm compliance under 10 kN will not hold plunge depth without closed-loop force control and, ideally, a 6D force/torque sensor between flange and spindle. KUKA supplied exactly such a research cell to IIT Kharagpur, and the configuration is now common in automotive integration.

8. The industrial record

FSW was invented at The Welding Institute in 1991 by W. M. Thomas and colleagues. The first production application that mattered was aerospace: NASA adopted it for the Space Shuttle External Tank after fusion welding of Al-Li 2195 proved prone to hot cracking in repair welds. The first production FSW welds on the External Tank were made in 2001; the first tank to fly with them was ET-132 in 2009.

The scale today is routine. NASA's Space Launch System uses FSW for the core stage's liquid hydrogen tank — a 39 m structure welded in a 52 m tall facility — and for Orion crew module elements. SpaceX has built Falcon 9 tanks with all-FSW Al-Li construction since the vehicle entered service, welding material from 1.6 mm to more than 12 mm with a retractable-pin head. Eclipse Aviation certified the Eclipse 500 business jet with FSW replacing thousands of rivets in cabin and aft-fuselage assemblies. ISRO flew its first FSW propellant tanks on PSLV-C41 in April 2018, and reported improved productivity and payload capability from the switch.

In India, the capability is real but concentrated. ETA Technology in Bengaluru builds the WSV-G FSW machine with up to 50 kN Z-axis thrust, retractable-pin tooling, stationary-shoulder capability, and in-process monitoring for automotive, rail, and aerospace work. KUKA's Indian operation supplied the IIT Kharagpur FSW research cell. But the number of Indian job shops that can quote a qualified 6xxx FSW procedure with PAUT and bend-test coupons is small — which is exactly the kind of capability gap a marketplace is supposed to expose.

9. What a metre of FSW costs

The cost model has five components: tool wear, machine time, labour, power, and inspection. The following uses a 22 m battery tray, 6 mm AA 6082-T6, 3 m/min traverse, 10.2 min per part including handling and plunge/retract, and a gantry cell with 50 kN axial capacity.

Cost element · Basis · ₹/m

Tool wear · H13 tool ₹25,000, 1,500 m life · 15–25

Machine · ₹2.5–6 crore cell, 5 years, 4,000–6,000 h/yr, 20–30 m/h · 50–150

Labour · One operator + one helper, ₹500–800/h · 17–40

Power · 20–40 kW at ₹8–10/kWh · 7–20

Inspection · PAUT sampling + leak test · 50–250

Fixture/tooling amortisation · Clamps, backing bars, runoff tabs · 20–50

Total · · 160–490

At the low end — a captive line running 200,000 trays a year, full PAUT sampling optimised, tool life at the top of the range — the weld cost is ₹120–180/m, or ₹2,600–4,000 per 22 m tray. At the high end — a job shop quoting a one-off batch with full inspection and setup amortised over a few parts — ₹350–490/m, or ₹7,700–10,800 per tray.

Compare MIG: 6 mm aluminium MIG runs at ₹80–150/m in consumables, labour, and power, apparently cheaper. But MIG on a 1.7 m panel introduces 1–3 mm/m of angular distortion, requires straightening, and produces porosity that must be leak-tested and sometimes reworked. The true comparison is not ₹/m of weld; it is ₹/acceptable part. FSW's low distortion is why the battery tray line uses it despite the higher nominal weld cost.

The capital threshold is the real barrier. A gantry FSW cell with 50 kN force capacity, a 3 m × 2 m work envelope, and PAUT integration costs ₹2.5–6 crore. To amortise that, you need either high-volume automotive work (20,000+ trays/year), aerospace or rail work with a quality system that supports the procedure qualification, or a niche where distortion-free aluminium welding commands a premium. A general machine shop cannot justify it on occasional work.

10. Specifying an FSW job

If you are sourcing FSW on FabFlow, the RFQ needs more than a drawing and a quantity. The minimum information set:

Questions worth asking before you award:

  1. What tool material and shoulder/probe geometry will you use, and what is the qualified life in metres?
  2. What is the machine's rated axial force and Z-axis stiffness at the working height?
  3. Do you have retractable-pin capability, or will you use a runoff tab?
  4. How will you inspect the root — PAUT angle, scan step, from which side — and what is your demonstrated flaw detection capability?
  5. Can you supply a procedure qualification record to ISO 25239-4 or AWS D17.3 for this alloy and thickness?
  6. What is your evidence for as-welded joint efficiency, and are you designing to the HAZ strength or the parent temper?

Red flags: a quote that promises "T6 properties in the weld" without post-weld solution treatment; an inspection plan that relies on radiography alone; no mention of tool geometry or weld pitch; and a fixture that is not designed for the axial force. FSW looks like milling with a funny tool, and that resemblance causes more failed projects than any metallurgical issue.

The process is mature, the standards exist, and the economics work at volume. The engineering work is in the process window, the tool, and the inspection plan — and those three are decided before the first arc-free weld is made.

More FabFlow blog posts