Case Study: Descaling and Strip Cooling Headers in a Hot Strip Mill

A hot strip mill was losing yield to residual scale and fighting a widening spread in mechanical properties. The descaler had been rebuilt in stages over several years. Each rebuild replaced only the nozzles that looked worn. The result was one header assembled from several unmatched batches, each with a different flow.

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. No test report, certificate or customer document is represented here.

The fault looked like a nozzle problem. It behaved like a system problem. Flow, impact, angle, standoff, filtration and wear all moved together. Repairing one position at a time never repaired the header.

The plant called it scale carryover. Operators saw it as patches of dark oxide that survived the descaler. The quality team saw it as a slow drift in coil properties. Both groups were describing the same header.

Case Snapshot

Item	Reference configuration
Industry	Carbon steel hot rolling
Duty	Primary and secondary descaling, plus strip cooling
Line	Hot strip mill finishing train
Material	Low carbon and micro-alloyed steel grades
Strip temperature at the descaler	1,000 to 1,080 degrees C
Header type	Manifold with individually fed nozzle positions
Nozzle family	Flat fan, small spray angle
Spray angle band	22 to 30 degrees
Header pressure band	140 to 200 bar
Standoff to strip	90 to 130 mm
Cooling overlap	35 to 45 percent across the width
Target	Uniform scale removal and flat width-wise cooling

The Challenge

The finishing mill ran low carbon and micro-alloyed grades through a conventional hot strip line. Strip entered the descaler at roughly 1,000 to 1,080 degrees C. Primary scale had to come off before the first finishing stand. Secondary scale formed again between stands as the strip re-oxidised in air. When the headers were out of condition, both stages suffered.

Three symptoms appeared together. Patches of scale survived primary descaling. The temperature profile across the strip width lost its flatness. The spread in yield strength and elongation from coil to coil grew. None of the three could be traced to a single nozzle.

The maintenance history explained the drift

The team replaced nozzles when a spray looked weak. Wear in a flat fan nozzle is gradual. A worn orifice passes more flow at the same pressure. Published practice puts the point where an operator notices the change well after flow has already risen. A rise of about 10 percent often comes first, long before any visible defect.

So the header slowly became uneven. Some positions were new. Some were half worn. Some were fully worn.

Every position delivered a different flow at the same header pressure. Impact followed flow. Coverage followed angle and standoff. The plant had no record of which position was at which stage.

Pressure was set at the pump, not at the header. That single choice hid the problem. A pump discharge gauge reads total system pressure.

It does not show the pressure each nozzle position actually sees. Small losses along a long manifold differ between the two ends. Operators had no way to know that individual positions were running at different pressures.

Cooling carried its own version of the fault

The cooling headers used overlapping flat fans across the width. Overlap is meant to sit between 30 and 50 percent. As orifices wore, each fan widened and the overlap changed.

The cooling rate then varied across the width. The centre of the strip and its edges cooled at different rates. Shape defects followed.

The water bill crept upward as well. Crews raised pressure to compensate for worn nozzles. The pump pushed more water through every position, including the good ones. Actual impact at the strip rose and fell depending on which position was worn. Total volume rose regardless.

The quiet cost

Downtime was the visible cost. Nozzle work was unplanned. Crews pulled positions during short breaks and rebuilt what they could. Each intervention restored one position and disturbed the balance of the rest.

The hidden cost was larger. Uneven cooling changes the phase transformation across the width. That change shows up as a spread in properties. A plant can pass that spread downstream for months before anyone measures it properly.

Coil property spread also breaks downstream processes. A stamper setting a press for one hardness will scrap parts from a harder coil. A plant may blame its customer’s tooling while the cause sits in the mill’s cooling headers.

The Solution

The plant stopped treating the header as a list of parts. It rebuilt the header as one system with defined flow, angle, standoff and pressure. The rebuild had six parts, and they had to be done in order.

Define the duty first

The team wrote down what the header had to do before buying anything. That meant strip temperature at the descaler, strip speed, width range and the required impact at the surface. It also meant the cooling rate target for each cooling header. Duty sets flow. Buying nozzles first is how a header becomes mismatched.

The team also recorded which grades were most sensitive to cooling rate. Those grades set the tighter of the two cooling targets. Everything else in the rebuild followed from those numbers.

Match the nozzle family and angle

The plant standardised on flat fan nozzles across the descaler and the cooling headers. Flat fans give a thin, high impact sheet, which suits scale removal. Small spray angles of roughly 20 to 40 degrees keep the impact concentrated. Wider angles spread the same flow thin and lose impact at the surface.

The cooling headers used a matched flat fan with a deliberately chosen angle. That angle kept overlap between 30 and 50 percent across the full width.

Standardising the family mattered as much as choosing it. A header fed by several part families has no single flow baseline. A header fed by one family can be tested against one curve. That single decision made the flow test schedule possible later.

Set pressure at the header

The plant moved the pressure reference to the header inlet. A gauge at each header replaced the pump-side reading as the number operators trust. Flow then follows Q = K times the square root of P.

Double the pressure and flow rises by about 41 percent. Impact rises with both flow and pressure, so F scales with Q times the square root of P. Setting pressure at the header makes both predictable at every position.

The pressure band itself was set in the published descaler range of roughly 100 to 250 bar. The plant settled at 140 to 200 bar at the header for its duty. Higher pressure was available, but it bought little extra scale removal and cost pump energy.

Fix filtration and strainer discipline

Worn seals and debris plug small orifices. A partly blocked nozzle sprays off axis and cools the wrong place. The rebuild added a defined strainer grade and a scheduled cleaning step. It also set a rule that a nozzle with a damaged tip is replaced, never cleaned and returned. Half restored parts are how a header drifts.

Replace on measured flow, not appearance

This was the change that held. Every position got a flow test at a fixed pressure on a fixed schedule. A position that drifted more than 10 percent above its baseline went to the replacement list. Operators no longer judged by eye. The test made wear visible long before the spray looked wrong.

The schedule was tied to running hours, not to calendar weeks. A mill that runs continuously wears a header faster than one that runs one shift per day. Running hours gave a fair baseline across positions.

Document the header

The final step was a header map. Each position carried its nozzle family, angle, baseline flow and last test date. A rebuild plan became a document instead of a memory. When a position failed, the crew knew its history within seconds.

Together these steps turned a list of parts into a controlled system. The header now had one flow target, one pressure reference and one replacement rule. The team used two references most during the trial. One was the jet nozzle guide. The other was the flow rate calculation guide.

The Engineering Behind the Choice

The rebuild worked because the physics of a header is simple, and the plant respected it. Four relationships drove every decision.

Flow and pressure

Flow through an orifice follows Q = K times the square root of P. K is a constant for a given nozzle and fluid. This single relation explains most of the story.

Raise pressure by four times and flow doubles. A measured 10 percent flow rise means the effective orifice has grown. The gauge may read the same pressure, but the nozzle now delivers more water than the header was designed for.

This is also why flow, not pressure, is the wear signal. Pressure is the input. Flow is the output that actually touches the strip. A test stand that measures flow at a fixed pressure reads the true condition of the nozzle.

Impact is not just pressure

Impact follows the same logic in two steps. Impact force rises with flow and with pressure. Published practice writes it as F proportional to Q times the square root of P.

Both terms matter. A worn nozzle that passes 10 percent more flow also hits the strip harder. That sounds helpful. It is not, because the extra flow is not matched by the other positions.

Unmatched impact across a header is worse than uniformly low impact. Uniform conditions let the mill set one recipe. Uneven conditions force the mill to run hot enough for the worst position. That costs energy and still leaves scale at the edges.

Why small angles and close standoff

Descaling is a momentum problem, not a volume problem. Water has to strike the scale and break it away. A small spray angle concentrates the same flow into a narrower sheet. That raises impact per unit area at the surface.

A wide angle spreads the flow and lowers the impact. This is why descaler headers use flat fans at roughly 20 to 40 degrees. A wide angle cannot do the same job at the same flow.

Standoff sets the working distance. Descaler headers run close, often 90 to 130 mm from the strip. At that distance a small angle still covers the needed width while keeping impact high. Push the header further away and the sheet spreads and slows. Coverage may improve, but scale removal does not.

Overlap matters more on the cooling headers. Fans spaced so that neighbours overlap by 30 to 50 percent give even coverage. Below that band the edges run dry and cool slowly. Above it the centre can be over cooled and the edges starved as the pattern spreads.

Temperature and scale growth

Strip reaches the descaler at roughly 950 to 1,100 degrees C. Steel oxidises quickly at that temperature. Scale removed at the primary descaler starts to reform before the first finishing stand. Secondary scale grows again between stands. Each set of headers gets one short window to remove what has formed. Miss that window and the scale rolls into the surface.

Hot scale is also soft. Cold scale is brittle and hard to remove. This is why header impact matters more than total water volume. A low impact spray at high volume wets the strip without breaking the scale. A high impact spray at moderate volume does the work with less water.

Cooling is a rate, not a soak

Cooling headers have a different duty. They set the cooling rate after the last stand. That rate decides the phase transformation and therefore the properties. Total flow sets the average rate. Overlap sets the uniformity across the width. If overlap falls below 30 percent the edges cool slower. If it rises above 50 percent the centre can be over cooled.

The mill measured uniformity as the standard deviation of temperature across the width at a fixed point downstream. Before the rebuild that spread sat near 14 degrees C. After it sat near 5 degrees C.

Why wear hides

A nozzle does not fail suddenly. The orifice erodes or corrodes a little at a time. Flow rises, and impact rises with it. The spray pattern also widens slightly, so overlap changes. At 10 percent extra flow the change is measurable but not visible. On a mill that runs continuously, that 10 percent shows up first as uneven cooling and shape defects. By the time an operator sees a weak spray, the product has been out of specification for a while. The guide to silent over-application covers this pattern in detail.

The product page for flat fan nozzles lists the families used here. The cooling duty follows the principles in the strip cooling nozzles guide.

The Results

After the rebuild the header held one flow target and one pressure reference. The numbers below compare the twelve months before the rebuild with the twelve months after. They are working figures from this class of plant, not from a single test report.

Metric	Before	After	Unit
Scale remaining after primary descaler	100	42	relative index
Width-wise cooling spread	14	5	degrees C
Coil property spread, yield strength	28	11	MPa
Nozzle replacement interval	3	9	months
Water consumption	1.60	1.25	m3 per tonne
Unplanned downtime for nozzle work	22	7	hours per year

The largest single gain was the nozzle replacement interval, which roughly tripled. Scale remaining and cooling uniformity moved together, which is expected when one header feeds one duty. Water consumption fell even though impact rose, because the plant no longer compensated for worn nozzles by raising pressure.

Unplanned downtime fell by about two thirds, and the remaining hours were planned. That shift alone paid for the flow test stand within the first year.

Why This Case Matters

Most descaling faults are described as nozzle faults. The fix is then applied position by position. This case shows why that approach fails. A header is a system. Orifice size sets flow. Pressure sets flow and impact together. Angle and standoff set coverage and overlap. Filtration keeps the pattern true. Wear degrades all of these at once, slowly and invisibly.

The practical lessons transfer to any plant that runs a header. Write the duty down first. Set pressure at the header, not at the pump. Measure flow per position on a schedule. Replace at a 10 percent rise, not at the first visible defect. Keep a header map so the history survives a shift change.

None of these steps needs much new hardware. Matched nozzles, a header gauge and a simple flow test stand cover most of it. The return comes from consistency. A header that behaves as one system produces strip that behaves as one product.

The same reasoning applies beyond steel. Any cooling or coating header that runs continuously will drift the same way. The signal is always the same quiet rise in flow before anyone sees a problem.

If your mill is fighting scale carryover or uneven cooling, talk to us at contact BoreJet. We can review your duty, pressure band and nozzle family, then help you set a per-position flow test schedule.