Case Study: Continuous Casting Secondary Cooling Nozzle Rebuild

A slab caster was losing slabs to off-corner cracks and fighting a cooling curve that moved every time the machine restarted. Its secondary cooling headers had been rebuilt nozzle by nozzle over about four years. Each rebuild replaced only the positions that looked worn, which left the headers as a mix of parts from several batches.

This account is anonymized. It describes a plant of this class rather than a named customer. Every figure is a realistic working value inside published industry ranges.

The fault presented as a crack problem. It tested as a flow distribution problem. Blocked orifices, eroded tips and a drifting air pressure band had all moved local spray flux away from its design value.

Case Snapshot

The caster ran crack-sensitive grades at 0.8 to 1.6 m/min through a slab section 1.5 to 2.0 m wide. Off-corner cracks appeared on roughly 1 slab in 25 during the worst quarter. Corner spray cutback had been set by eye, not by measurement. The machine carried 8 to 15 secondary cooling zones from the mold exit to the straightener, with 5 to 10 nozzles per row. Roughly 4 percent of production was downgraded for surface defects in that quarter.

Published references put the useful specific water flow band for this duty at 0.6 to 1.5 L/kg steel. The plant sat inside that band, so total water volume was never the problem. Distribution across the width and down the strand was the problem. Zone flow met its setpoint on the control screen. The measured flux under individual nozzle rows did not.

The table below sets the reference configuration used for the rest of this account. It lists the pressure bands, water quality and quality targets that shaped every later decision.

Item	Reference configuration
Industry	Carbon steel continuous casting
Duty	Secondary cooling of a slab strand, mold exit to straightener
Plant class	Slab caster, 1.5 to 2.0 m width, 0.8 to 1.6 m/min
Original setup	Water-only flat fans in upper zones, air-mist in middle and lower zones, mismatched vintages
System chosen	Matched flat fan and air-mist families, per-zone headers, independent corner nozzles
Operating point	Water 2 to 8 bar, air-mist 4 to 7 bar water with 3 to 5 bar air
Water quality	Filtered process water, scale and debris present, specified strainer grade
Quality targets	Off-corner crack rate down, width-wise flux spread within plus or minus 10 percent

The Challenge

Three symptoms appeared together over roughly eighteen months. Off-corner cracks climbed from about 3 per 1000 tonnes to about 12 per 1000 tonnes. Width-wise surface temperature spread widened to about 18 degrees C across the strand. Each restart after a ladle change left a visible cold band on the surface below the mold. The plant treated the three as separate faults for several months before anyone mapped the cooling zones together.

The off-corner region behaves differently from the face centre. A corner loses heat through two faces, so it runs cooler than the face even with identical spray. Published practice cuts corner spray back by a defined amount instead of guessing. The plant had cut it back by eye. Some corner nozzles ran about 40 percent below face flow, while others ran at full face flow on the same row.

Water-only zones used flat fan nozzles at 2 to 8 bar. Scale and mill debris plugged the smaller orifices before the larger ones. A partly blocked flat fan does not simply spray less water. Its pattern collapses toward the centre and the fan edge runs dry. That dry band sat at the same width on every row, so the defect repeated down the strand.

Air-mist zones added a second failure mode. Published references put air-mist nozzles at 4 to 7 bar water with 3 to 5 bar air. Zone air pressure in the plant moved by about 1 bar through a single cast. Droplet size follows air pressure, so the spray pattern changed width as the air line moved. A widening pattern over-sprays the face centre and starves the edges.

Water quality made the drift worse. A partly loaded strainer cut zone flow by 8 to 15 percent before any operator saw a change at the strand. Scale built up inside nozzle bodies and narrowed the effective orifice. The result was a slow fall in flow that no single gauge could isolate to one position. On the same header, tip wear pushed flow the other way, with a 5 to 12 percent rise at positions that ran clean.

Restart was the hardest moment. Valve sequencing opened the spray zones over several seconds. An already uneven header turned that ramp into a step change in surface temperature. Published practice ties the restart ramp to a reheating limit near 100 degrees C/m along the strand. The plant had no ramp profile at all, so every ladle change re-wrote the cooling curve.

The cost driver was not nozzle price. It was downgraded and scrapped slab, plus the crew hours spent chasing a fault that moved. A crack that appears off-corner points at cooling. A crack that appears on one face points at a row. This fault did both, which is why position-by-position replacement never closed it out.

The Solution

The plant stopped buying nozzles and started writing a duty. Each zone got a water budget in litres per minute, a target flux in litres per square metre per minute and a surface temperature window. The zone budget followed from the specific water flow target, not from a nozzle catalogue. The layout was then checked at the operating flow rather than at the catalogue angle. Rows that looked even on paper at minimum flow developed gaps at the top of the range.

Nozzle families were then matched to the duty. Water-only zones kept flat fans in the 2 to 8 bar band, where flow is high and control is simple. Middle and lower zones moved to air-mist at 4 to 7 bar water with 3 to 5 bar air. That band is where low flow has to stay even without losing its pattern. The flat fan range lists the angle and flow classes used in the water-only zones.

Overlap and spacing were set from geometry, not habit. Adjacent flat fans were overlapped by 40 to 50 percent at the strand surface. Rows ran 150 to 300 mm apart and were staggered, so the edge of one spray covered the centre of the next. Standoff was fixed at 100 to 250 mm and locked on the frame. Each corner position also got a fixed angle, so the cutback stayed repeatable between shifts.

Corner cooling got its own control. Corner nozzles were fed independently and set to a measured cutback, which raised corner surface temperature by roughly 50 degrees C against the face centre. The cutback started near 30 percent of face flow on the crack-sensitive grades. The team tuned that value against slab samples for two campaigns before locking it. Locking the value mattered, because an unrecorded cutback drifts with every rebuild.

The pressure reference moved to the header. A gauge at each zone header replaced the pump-side reading. Air pressure for each air-mist zone got its own regulator, so a swing on one zone could not move another. Flow then tracked pressure by the standard orifice relation at every position. Water-only zones run 2 to 8 bar, and air-mist zones run 4 to 7 bar water with 3 to 5 bar air.

Filtration and cleaning were written into the schedule. A defined strainer grade was installed upstream of each zone. Cleaning moved from a calendar task to a running-hour task. A caster that runs continuously loads a strainer faster than one that runs a single shift. A nozzle with a damaged tip went to the bin, never back into service, since half restored parts built the original mix of vintages.

The change that held was measurement. Every position got a flow test at a fixed pressure, and any position more than 12 percent above baseline went to the replacement list. The team also mapped spray flux across the strand width and recorded it as a per-row map. Restart got a written profile as well, with a start flow and a hold flow for each zone. Two references shaped that work: the flow rate calculation guide and the silent over-application guide.

The Engineering Behind the Choice

Spray heat transfer is not linear in water flow. Published work puts the heat transfer coefficient near HTC proportional to flux to the power 1.2 to 1.3. Doubling local flux raises heat extraction by roughly 2.3 times. That exponent is why a single dry nozzle creates a cold spot rather than a mild dip. The secondary cooling nozzles guide sets out the full zone picture behind this rebuild.

Flow through an orifice follows Q = K times the square root of P. K is fixed for a given nozzle and fluid, so an eroded nozzle passes more water at the same pressure. A 12 percent flow rise means the effective orifice has grown by roughly 12 percent in area. Impact rises with flow and with pressure, so that extra water also lands harder than the row beside it. The pressure gauge still reads normal, which is why pressure alone never reveals wear.

Droplet size sits under air pressure, not water pressure. Air-mist nozzles break the water sheet with an air stream, so the drop band tracks the air setting. Published references put air-mist drops at 100 to 400 micrometres and water-only drops at 500 to 2000 micrometres. That fine band is what lets a lower zone hold 0.3 to 0.6 L/kg without starving the strand. It is also why an air-mist zone cannot absorb a 1 bar air swing without a change in pattern.

Pattern width follows standoff and angle. The geometry is width equal to twice the standoff times the tangent of half the spray angle. A 90 degree fan at 200 mm standoff covers about 400 mm. Shift the standoff by 20 percent and coverage shifts with it. That is why the mounting frame matters as much as the nozzle, especially on a machine that vibrates through a cast.

Surface temperature history drives cracking, not the average temperature. Steel passes through a low ductility band as it cools, and a reheat back into that band can pull a crack open. Published caster practice keeps reheating under about 100 degrees C/m along the strand. The off-corner cracks in this plant mapped to a reheat band that began at the corner dry stripe. That link is why corner cutback was treated as a metallurgical setting rather than a plumbing detail.

Flux mapping gave the team a number it could act on. A pattern test at the operating pressure measured water collected per unit area under each row. Published acceptance for this duty is a width-wise spread within plus or minus 10 percent of the zone average. The plant used the same test to set the corner cutback and to recheck the map after each nozzle change. A map that moved by more than 3 percent between checks triggered a look at the strainer and the nozzle bodies.

The design rules transfer to other cooling and descaling headers. The strip cooling guide covers the same overlap and flow logic on a rolling mill. The hot strip mill descaling case study shows how the same wear pattern hides inside a high pressure header. In every case the signal arrives as a shift in distribution before it arrives as a visible defect. The same three numbers, a flow budget, a flux target and a pressure band, define the duty on any cooling header.

The Results

The figures below compare the twelve months before the rebuild with the twelve months after. They are working values for this class of plant, inside published industry ranges. They are not readings from a single customer report or a certificate. Each row names the mechanism that drove the change.

Metric	Typical published result	Why
Off-corner crack rate	12 down to 3 per 1000 tonnes	Corner flux matched to face flux
Width-wise flux spread	22 percent down to under 10 percent	Mapped pattern and matched tips
Off-corner surface temperature	65 degrees C below face centre, down to 25	Measured corner cutback
Nozzle replacement interval	4 months up to 11 months	Flow-rise rule replaces visual check
Specific water flow	1.35 down to 1.05 L/kg steel	Even flux removes the need to over-spray
Zone water variation	15 percent down to 4 percent	Header pressure reference per zone
Unplanned nozzle downtime	30 down to 9 hours per year	Planned replacement on running hours
Restart cold band	40 degrees C dip over 1.5 m, down to 12 over 0.5 m	Written ramp profile per zone

The largest single gain was the off-corner crack rate, which fell by about 75 percent. The flux spread and the crack rate moved together, as expected when one header feeds one duty. Water use fell while impact rose, because the plant no longer compensated for worn nozzles by lifting pressure. Pump energy fell with the water volume, and that saving continued every month after the trial.

The flow rise rule did most of the work on maintenance. Before the rebuild, crews judged a nozzle by eye and replaced it once the spray looked weak. By then that row had been running above its design flow for weeks, while the neighbouring rows sat near baseline. Testing at a fixed pressure turned that judgement into a number any shift could read.

Why This Matters

Most caster cooling faults are described as nozzle faults, and the fix is then applied one position at a time. That approach fails for the same reason it fails on a descaling header. A cooling zone is a system, where total flow sets the average rate and overlap and standoff set the distribution. Filtration keeps the pattern true, and corner setpoints decide the off-corner temperature.

The second lesson is that distribution beats volume. This plant already sat inside the published specific water flow band of 0.6 to 1.5 L/kg steel. More water at the zone level would have raised the average without fixing the dry stripes. Matching flux across the width did more than any increase in pump pressure. It also cut the water bill, which is the opposite of what most crews expect from a cooling fix.

The third lesson is about measurement. A 12 percent flow rise is invisible on a spray and obvious on a flow test. A 10 percent flux deviation is invisible on a temperature average and obvious on a pattern map. Both tests are cheap, and both replace an argument about appearance with a number. A flow test stand and a pattern tray cover nearly all of the rebuild described here.

The return in this class of plant comes from consistency, not from hardware. Matched nozzle families, a header-side pressure reference, a strainer schedule and a flow test rule cover most of the work. Each item is a decision rather than a purchase. Together they turn a header that drifts into a header that reports its own condition. That shift is what made the replacement interval nearly triple.

A caster that holds its cooling curve produces slab that holds its quality. The same rebuild pattern applies to any strand section, whether the machine casts slab, bloom or billet. If your machine is chasing off-corner cracks or restart cold bands, send the strand section, grade and zone layout to the BoreJet team at contact. The nozzle sizes, angles and air-mist settings come back sized against your duty, not guessed.