Summary of Findings

Baking does not reverse. The structural changes in electrode paste materials are irreversible, driven by graphite crystallite growth (White et al., Metall. Mater. Trans. B, 2020), so an outage does not return the column to a uniform state. It leaves a permanently baked lower section, a still-green upper section, and between them a transition that has been through part of the transformation and stopped.

Those two materials do not behave alike on cooling or on reheating. In the baking-zone temperature range, pre-baked material conducts heat an order of magnitude better than green paste undergoing carbonisation (White et al., 2020). Worse for a restart, green paste conductivity falls as it warms, reaching approximately 0.5 W·m⁻¹·K⁻¹ by about 400 °C, before rising steeply above 600 °C. Heat entering the column on restart therefore moves readily through the baked section and stalls precisely at the material that needs it. That asymmetry, not the outage duration alone, is what makes the restart period the exposed one.

Key Findings

• The column is not reset by an outage. Baking is irreversible, so the lower section stays baked and the boundary between baked and green material persists through the shutdown.
• The two materials differ by an order of magnitude in conductivity in the baking-zone temperature range (White et al., 2020) — 3.86 W·m⁻¹·K⁻¹ at 22 °C rising to 7.97 W·m⁻¹·K⁻¹ at 800 °C for pre-baked material, against a green-paste minimum near 0.5 W·m⁻¹·K⁻¹.
• Green paste conducts worse as it warms, up to about 400 °C, which is the temperature range a restart has to drive it through. The transition region is thermally the hardest part of the column to reheat.
• The baking zone has to be re-established, not resumed. It disappears as the column cools and reforms as load is restored, and its position while reforming is governed by current in the same way as in normal operation (Larsen, Platinum 2012).
• No restart curve is given here. Ramp rates, hold points and cadence during restart are furnace-specific and belong to the plant’s operating standard and the furnace supplier’s instructions.

Contents

1. What does an outage leave in the column?
2. Why is the restart period the exposed one?
3. What should be set up before a planned shutdown?
4. What is worth recording during the outage?
5. What governs the restart?
6. How does an unplanned shutdown differ?
7. Frequently asked questions
8. Related reading

1. What Does an Outage Leave in the Column?

Three sections in different states. Their boundaries reflect conditions at shutdown, but residual heat can continue to redistribute during cooling.

The baked section, below where the baking zone stood when load was removed. This material has completed the transformation and will not revert: the structural changes in electrode paste materials are irreversible, driven by graphite crystallite growth (White et al., 2020). It cools, but it cools as baked carbon.

The green section, above the transition. This is paste that never reached the transformation temperature. On cooling it stiffens as the binder passes back below its softening point, but it is chemically unchanged and will bake normally when it eventually reaches temperature.

The transition, between them. This is the section that was mid-transformation when load was removed — partly devolatilised, partly carbonised, and stopped part-way. It is the least predictable of the three, and it is the one that will be carrying load and current soonest after restart.

The practical consequence is that the column’s condition at restart is a function of where the baking zone stood at shutdown, not of how long the outage lasted. Two furnaces down for the same period, one with the zone sitting high and one with it sitting low, restart with different amounts of unbaked material above the load-bearing region.

2. Why Is the Restart Period the Exposed One?

Because heat has to travel back into the column through the material least able to carry it.

The measured contrast is large. Pre-baked material rises from 3.86 W·m⁻¹·K⁻¹ at 295 K (22 °C) to 7.97 W·m⁻¹·K⁻¹ by 1073 K (800 °C), nearly linearly. Green paste behaves in the opposite direction over the range that matters: its thermal conductivity decreases with temperature up to about 673 K (400 °C), bottoming near 0.5 W·m⁻¹·K⁻¹, stays roughly flat between 673 K and 873 K (400–600 °C), and only then climbs steeply, reaching 2.75 W·m⁻¹·K⁻¹ by 1073 K (800 °C) (White et al., 2020).

Three consequences follow for a restart.

The transition region is the thermal bottleneck. The baking isotherm lies between 673 K and 773 K (400–500 °C) (White et al., 2020), which is the range where green paste conductivity is at its minimum. Heat arriving from below meets its greatest resistance exactly where the transformation needs to resume.

The baked section heats first and fastest. It conducts well and it is nearest the heat. This produces a temperature gradient concentrated across the transition rather than distributed along the column.

The zone has to re-form under current. In normal operation, electrode current is the most important parameter for baking zone position in high-current-density operations, while mattering less at low current density (Larsen, Platinum 2012). A restart runs through the low-current-density condition on the way to the normal one, so the zone re-forms low and rises as load is restored. Slipping cadence calibrated for normal operation does not suit that period, and material advanced during it arrives below the shoes under-baked.

3. What Should Be Set Up Before a Planned Shutdown?

A planned outage is the case where the restart can be made easier rather than merely survived. Four items, decided before load is removed:

1. Where the baking zone should be standing at shutdown. This is the single decision with the largest effect on the restart, because it fixes how much unbaked material sits above the load-bearing region for the whole outage. Establishing the zone position for the furnace is treated in the baking zone article.
2. Column length and paste level. Both are held through the outage and both constrain what can be done on restart. A column set up for normal cadence is not necessarily set up for a slow re-establishment.
3. Charging stopped at a defined point, so that the material above the melt is of known age and condition rather than whatever the last shift added.
4. The restart plan agreed in advance with the furnace supplier’s instructions and the plant’s operating standard, including who decides the ramp and against what indications.

Items 1 and 3 are the ones most often left to the shift that happens to be on when load comes off, and they are the two that a restart is most sensitive to.

4. What Is Worth Recording During the Outage?

The outage record is what makes the restart assessable and what makes a subsequent breakage attributable.

• Zone position and column length at the moment load was removed.
• Paste level at shutdown, and whether charging was stopped before or after the level was set.
• Casing surface temperature at the fixed monitoring points, on the plant’s normal schedule for as long as monitoring continues.
• Outage duration and ambient conditions, since a long outage in cold ambient cools the column further than a short one in warm.
• The batch in service and its Certificate of Analysis (COA), because the partially baked material whose carbonisation will continue on restart is the material that was in the column at shutdown, not the next delivery.
• Any casing work carried out during the outage, which is both an opportunity and a variable — the outage is when casing condition can be examined, and a casing change alters the thermal path.

5. What Governs the Restart?

This article does not give a restart curve, and no general one should be taken from a published source. Ramp rate, hold points, load steps and cadence during re-establishment depend on furnace class, electrode diameter, cooling capacity, the zone position at shutdown and the outage duration. They are properly governed by the furnace supplier’s instructions and the plant’s operating standard.

What can be stated generally is the physics the restart plan has to respect, and the variables it turns on:

• Load ramp against zone re-establishment. The zone re-forms low and rises as current is restored. A ramp that outruns the re-establishment leaves the transition under-baked while load is already being carried; a ramp that is too slow leaves the transition under-baked for longer. Both directions are exposed, which is why the ramp is matched to indications rather than to a clock.
• Slipping cadence during the period. Cadence calibrated for normal operation advances material faster than it is baking while the zone is still low. Reducing cadence during re-establishment is the usual response, and the point at which it returns to normal is an indication-based decision.
• What the indications are. Casing surface temperature profile at the fixed points, electrical behaviour per phase, and off-gas character are the available signals; the level reading is one of them only if the column condition is known. These are the same indirect indicators used to infer zone position in normal operation.
• The first period after full load is reached is not the end of it. Material advanced during re-establishment reaches the region below the shoes later, so a restart-related weakness presents after the restart appears complete. This is the same delay that makes the breakage investigation examine the operating record of preceding days rather than only the moment of failure.

6. How Does an Unplanned Shutdown Differ?

In three ways, all of which make the restart harder rather than merely faster.

The zone position at shutdown is whatever it was. None of the preparation in Section 3 is available. The first task on restart is therefore to establish where the zone was standing, from the last available records rather than from a plan.

Charging state is unknown. Material may have been added shortly before load was lost, so the upper column may contain a recent charge that never began to melt — which is also a bridging precondition, treated in the bridging article.

Cooling may have been faster. A controlled shutdown allows load to be reduced in steps; a trip removes it at once. The resulting thermal gradient across the transition is steeper, and the transition is the section least able to accommodate it.

The response is the same in structure but slower in execution: establish the column condition from records before deciding the ramp, and treat the level reading as unverified until the column condition is known.

7. Frequently Asked Questions

Q1. Does a long outage require a different restart from a short one? Duration matters mainly through how far the column cooled, and the zone position at shutdown matters more. A short outage with the zone sitting low can leave more unbaked material above the load-bearing region than a long outage with the zone sitting high. Establish both before planning the ramp, rather than treating duration as the governing variable.

Q2. Should paste be charged during an outage? That depends on the plant’s standard and on the expected outage length, and it interacts with what happens on restart: material added during an outage is cold and unmelted when load returns, and it sits above a transition that is already the thermal bottleneck. Where charging during an outage is normal practice at the plant, the restart plan should account for the additional cold mass rather than treat the column as if it were unchanged.

Q3. We had a breakage two days after a restart. Is that related? Possibly, and the operating record of the restart period is the evidence. Material advanced while the zone was still re-establishing arrives below the shoes under-baked and reaches the load-bearing region later, so a restart-related weakness presents after the restart appears complete. Follow the breakage diagnostic path, and include the restart ramp and the cadence during re-establishment in the records reviewed.

Q4. Can the paste specification be changed to make restarts easier? Not usefully as a primary measure. The restart difficulty is a thermal and cadence problem rather than a composition problem, and a specification change addresses it only indirectly. Where restarts are frequent and problematic, the more productive review is of the zone position at shutdown, the ramp, and the cadence during re-establishment.

Q5. Is casing inspection during an outage worth the time? It is the only opportunity to examine casing condition at the heights that matter without the furnace in service, and casing defects connect to two separate failure routes — gas paths that can discharge in service, and, where the material adjacent is unbaked, paste escape. The flare-up article treats both. Scheduling the inspection into the outage rather than into a later opportunity is generally the more efficient sequence.

8. Related Reading

• How the Baking Zone Forms and Migrates in a Self-Baking Electrode (companion article) — the zone whose position at shutdown governs the restart
• Paste Column Bridging and Uneven Sintering — the condition that a charge added shortly before an outage can create
• Flare-Up and Abnormal Tar in Furnace Off-Gas — the casing findings an outage inspection can produce
• Söderberg Electrode Hard Breakage vs. Soft Breakage — how a restart-related weakness presents days later
• The Current Path in a Self-Baking Electrode — why the electrical length of the column changes as the zone re-forms
• What Volatile Matter Actually Does Inside a Söderberg Electrode — the transformation that stops part-way at shutdown

Author and Review

Written by Sterling Yang, Senior Vice President at Jiyang Carbon.

This article was reviewed by Mr. Yang, Technical Director at Jiyang Carbon, drawing on 40+ years of hands-on experience in carbon processing.

Thermal conductivity values, the baking isotherm range and the irreversibility of the structural change are taken from the published source cited in this article. No restart curve, ramp rate, hold point or cooling rate is given, because these are furnace-specific and are properly governed by the furnace supplier’s instructions and the plant’s operating standard. The article sets out the physics the restart plan has to respect and the variables it turns on.

Technical Consultation

For assistance in reviewing a restart record against a subsequent breakage, or in assessing column condition before a planned outage, please contact the technical team.

Get a Technical Consultation →