Continuous Casting: Freezing Molten Steel Into an Endless Strand

The machine: ladle to cut length

A continuous caster is a single path that steel travels once. A ladle of ~150–300 t of refined steel is hoisted onto a rotating ladle turret and opened; the stream falls through a shroud into the tundish, a refractory bathtub of ~20–80 t that buffers ladle changes and gives inclusions minutes to float out. From its floor a submerged entry nozzle (SEN) — an alumina–graphite tube with ports buried below the steel surface — feeds the mold: an open-ended, water-cooled copper tube or set of four plates, typically 700–900 mm long, liquid entering the top and a solid-skinned strand leaving the bottom.

Below it the strand runs through the secondary cooling zone, sprayed between closely spaced containment rolls; curved-mold machines cast it already bent while vertical-bending machines bend it below the mold, and both later straighten it back to horizontal in driven withdrawal rolls; travelling oxy-fuel torches (or shears, for billets) then ride along with the strand and cut it to length. Because the turret swings a fresh ladle over the tundish without stopping the flow, casters run sequences of ten, twenty or more heats back-to-back — the strand genuinely never stops.

The square-root law: why the machine is as long as it is

Everything about a caster's geometry follows from one piece of physics. To freeze more steel, the latent heat released at the solidification front (~270 kJ/kg) must be conducted outward through the shell already formed. As that shell thickens its thermal resistance grows, so the growth rate falls in inverse proportion to thickness. Integrating ds/dt ∝ 1/s gives the classic parabolic result:

s = k √t

with k ≈ 20–30 mm/min^0.5 in a commercial caster. Growth is fast, then stubbornly slow: the first 25 mm takes about a minute, the next 25 mm three more.

Since the strand moves at casting speed v_c, time maps onto distance. Solidifying a section of thickness D takes t_f = (D/2k)², so the metallurgical length — meniscus to the tip of the liquid core — is L = v_c(D/2k)². A 220 mm slab at 1.2 m/min with k = 25 needs about 19 minutes and roughly 23 m of machine. Halving cast thickness therefore cuts solidification time by four — which is why a 50 mm thin slab runs at 5 m/min in a few metres of machine and a 250 mm slab cannot. Casting speed is capped by the metallurgical length the machine has, and by the need for the shell leaving the mold (typically 10–25 mm) to hold what is inside it.

Inside the mold: oscillation, negative strip, and flux

The mold must extract enough heat, fast enough, to build a self-supporting shell in under a minute. Copper plates ~40–50 mm thick, plated against wear on the lower region, carry cooling water at 6–12 m/s; average heat flux is ~1–2.5 MW/m², peaking near 3–5 MW/m² just below the meniscus, under automatic level control.

Left alone, the freezing shell would weld itself to the copper. Junghans's 1930s insight was to oscillate the mold vertically, and effectively every mold-type caster still does: strokes of a few millimetres up to ~10 mm at 100–300 cycles per minute. The critical parameter is negative strip time, the fraction of each cycle in which the mold descends faster than the strand — typically 0.1–0.2 s, during which the copper moves down past the shell, compressing the fragile skin so any incipient tear is squeezed shut and pumping lubricant into the gap. For a sinusoidal stroke s at frequency f, negative strip exists whenever the peak mold velocity πfs exceeds v_c.

That lubricant is mold flux, a CaO–SiO₂-based casting powder (with Al₂O₃, Na₂O, CaF₂ and free carbon) fed onto the steel surface. It melts into a pool ~10 mm deep and is dragged into the mold–shell gap, where it lubricates, absorbs alumina inclusions, insulates the meniscus, and controls heat transfer — the film partly crystallises as cuspidine, and a crystalline film deliberately slows heat extraction. It also leaves a fingerprint: every cycle stamps a shallow transverse oscillation mark, spaced v_c/f (roughly 5–15 mm) and a few tenths of a millimetre deep. Those marks are stress raisers, and deep ones start transverse cracks.

Below the mold: containing a bag of liquid steel

What leaves the mold is not a solid bar but a rectangular tube of steel perhaps 15 mm thick, glowing at ~1100–1200 °C and holding many tonnes of liquid. That liquid exerts ferrostatic pressure ρgh — roughly 1.4 MPa at 20 m below the meniscus — enough to balloon the hot shell outward. The cage of containment rolls fights it: pitched close (~150–250 mm) just below the mold where the shell is thinnest, and spaced wider as it thickens. Bulging between rolls does more than distort the section; it pumps enriched liquid along the centre and worsens segregation.

Between the rolls, sprays extract the bulk of the heat — usually air–mist nozzles, whose atomised droplets give heat-transfer coefficients of hundreds of W/m²K and turn down cleanly as speed changes. Cool too hard and the surface contracts against a hot interior, cracking it; cool too little and the shell is too weak. Steel also has a ductility trough between about 700 °C and 900 °C, where fine precipitates and thin ferrite films at austenite grain boundaries embrittle it — and straightening puts the outer-radius surface in tension, so the cooling schedule keeps that surface outside the trough when it reaches the straightener.

Steering the solidification: stirring, braking and soft reduction

Freezing from the outside in produces a problematic structure. A fine chill zone forms at the surface, then long columnar dendrites grow inward along the heat flow. Where they collide at the centre, the last liquid — enriched in carbon, phosphorus, sulphur and manganese, all rejected from the growing solid — is trapped as centreline segregation with shrinkage porosity, and plate rolled from such a slab can fail hydrogen-induced-cracking tests.

Two electromagnetic tools attack this. Electromagnetic stirring (EMS) — travelling or rotating fields at the mold, the strand or the final crater — swirls the liquid; the shear snaps dendrite tips off and disperses them as nuclei, forcing an early columnar-to-equiaxed transition so the core freezes as small random grains instead of colliding columns. Low superheat, near the bottom of the 15–40 °C range, helps for the same reason. Electromagnetic braking (EMBr) does the opposite job in slab molds: a static field decelerates the jets leaving the SEN, calming the meniscus so mold slag is not entrained as inclusions.

Soft reduction is the mechanical answer. Where the core's solid fraction is roughly 0.3–0.7 — mushy, not liquid — several roll pairs squeeze the strand with a taper of about 0.8–1.5 mm per metre, closing the shrinkage volume as it forms so no suction draws enriched liquid inward. That demands knowing where the mushy zone is, which is why every modern caster runs an online solidification model.

How it fails: breakouts and cracks

The characteristic catastrophe is a breakout: the shell ruptures below the mold and the liquid core pours into the machine, wrecking rolls and spray headers and costing many hours of production. The common variety is the sticker breakout — lubrication fails locally, the shell adheres to the copper, withdrawal tears it just below the sticking point, and the tear propagates downward behind a thin new skin until the weakened region clears the mold and lets go.

That mode is predictable because it has a thermal signature. Casters embed a grid of thermocouples in the mold copper, commonly two or more rows per face. A sticker appears as a hot spot at an upper thermocouple that reappears at the one below while lagging behind the casting speed, because the tear travels more slowly than the strand. Pattern-recognition detectors watch for that signature and automatically slow the machine so the tear heals inside the mold.

Cracking is the quieter failure. Longitudinal facial cracks plague peritectic grades (roughly 0.09–0.17 %C), where the δ-ferrite → austenite transformation adds an abrupt contraction that pulls the shell unevenly away from the mold and leaves thin hot spots; the remedy is a high-basicity flux that crystallises and deliberately reduces mold heat flux. Transverse cracks form at oscillation marks when straightening happens inside the ductility trough, and internal midway cracks come from tensile strain at the front due to bulging or over-aggressive sprays. Almost every casting defect is a stress applied to steel where it is weakest.

From Bessemer's rolls to endless strip

The idea is old. Henry Bessemer patented casting metal between two counter-rotating water-cooled rolls in 1857, but the steel of his day defeated him. Practical continuous casting arrived with Siegfried Junghans, whose oscillating mold — developed in Germany from the 1930s for brass and light alloys — solved the sticking problem, and with Irving Rossi, who licensed it into industry. Steel was first poured on such machines in the late 1940s at pilot plants in Britain and the United States; Rossi and Tadeusz Sendzimir founded Concast AG in Zurich in 1954. The curved-mold machine of the early 1960s was the commercial breakthrough, collapsing the building height from a tower into a manageable structure.

Adoption was then near-total — about 4% of world steel in 1970, ~30% by 1980, two-thirds by 1990, about 96% today — because the ingot route lost 10–15% of the steel to crops and scale and burned ~1 GJ per tonne reheating metal that had just been molten, against 95–98% yield for continuous casting.

Machines are built fully vertical, vertical-with-bending, curved-mold (the workhorse) or horizontal, and the frontier keeps pushing thickness down. Thin-slab casting arrived when Nucor started SMS's CSP caster at Crawfordsville, Indiana in 1989 — ~50 mm slabs from a funnel-shaped mold, fed hot straight into a rolling mill, the process that let mini-mills make flat-rolled sheet. Arvedi's ISP and later ESP lines fused caster and mill into a direct-linked and, with ESP, fully endless coil-to-coil process, and twin-roll strip casting (Castrip, Crawfordsville 2002) finally realised Bessemer's patent, solidifying ~1–2 mm strip in the bite between two rolls. The machines come mainly from SMS group, Primetals Technologies and Danieli — all still fighting the law Junghans fought: heat can only leave through the shell you have already made.

Continuous-casting product routes: thickness sets the speed, and thickness squared sets the freezing time

Route	As-cast section	Typical casting speed	Liquid-core (metallurgical) length	Downstream
Billet caster	~100–160 mm square	~2–6 m/min	~10–20 m	Bar, rod and section rolling
Bloom caster	~200–400 mm square/rectangular	~0.6–1.2 m/min	~15–30 m	Rails, beams, seamless tube, forging stock
Conventional slab caster	~200–250 mm × 900–2100 mm	~1–2 m/min	~20–40 m	Reheat furnace, then hot strip or plate mill
Thin-slab caster (CSP / ISP / ESP)	~50–90 mm thick	~4–6 m/min	~5–10 m	Direct-link tunnel furnace straight into the finishing mill
Twin-roll strip caster (Castrip)	~0.9–2 mm strip	~60–90 m/min	Solidification finishes inside the roll bite	Coiled almost as-cast; little or no hot rolling
Ingot casting (the route it replaced)	Discrete ingots, tonnes each	Batch — pour, strip, soak, roll	n/a	Reheat and bloom/slab the ingot before rolling