Hot Metal
Three things worth your attention this week.
In issue #1 we said electrode prices would start diverging by origin. They have. GrafTech is closing Monterrey, cutting nameplate from 178 to 127 thousand tonnes a year and blaming overcapacity out of China and India, while Chinese UHP 600 mm prices rose about 3% in August and are firming into September, and India's HEG runs above 90% utilization at a 29% EBITDA margin. Same product, three markets. Treat electrode supply as a strategic contract with an origin clause, not a spot purchase.
Electrode vendors keep offering consumption guarantees tied to premium grades, and the fine print deserves its own metallurgist. Guarantees are typically conditioned on current profiles, regulation performance, and breakage exclusions, which are the very factors that dominate consumption. That isn't a criticism; the vendors know exactly where kg/t is made. It's the buyer who forgets, signs for premium graphite, and keeps the regulation problem.
The US case is the one to watch this month. Commerce set preliminary countervailing duties on large-diameter electrodes in late July: 103.49% for Chinese producers that didn't answer the questionnaire, single digits for the Indian ones, with the anti-dumping preliminary due by 22 September and a final decision pencilled in for December. Brazil's parallel case, flagged in issue #1, is still running with no provisional duty yet. For a US shop on Chinese graphite, landed price is about to be a customs number, not a negotiation.
It was the regulation all along
The first issue of this newsletter argued that kg/t hides three mechanisms and that crew variance beats capital projects. Thirteen issues on, the version we'd defend across a table from a supplier is sharper: electrode consumption is an electrical-stability measurement that happens to be paid for out of the consumables budget. Get regulation and arc stability right and kg/t drifts toward the published floor for modern high-power shops, broadly 1-1.5 kg/t, against the 2-3 kg/t that market reports still average across everything with a roof. Get them wrong and no grade of graphite on any price list buys it back. Premium graphite in a badly regulated furnace loses to standard graphite in a well regulated one, every time we've seen the two compared.
Break consumption into its published mechanisms and the pareto explains itself:
Tip consumption. Sublimation and erosion at the arc, scaling with current squared times time. This is the physics share, and the one honest lever is energy efficiency: fewer power-on minutes per tonne is fewer electrode-minutes per tonne. Everything since the kWh breakdown in issue #8 that trims kWh/t trims this term with it.
Sidewall oxidation. Graphite burning off the column above the slag line, scaling with surface temperature, exposed area, and time in the furnace atmosphere. Foam quality shows up here again (a covered arc runs a cooler column), as do spray-cooling systems and, bluntly, power-off minutes, when electrodes sit hot in an oxidizing atmosphere earning nothing.
Breakage. The discrete losses, and where shops differ most. A break isn't a consumption event; it's an evidence event. More on that below.
Joints and stubs. Losses at nipples and discarded ends: the housekeeping share, controlled by torque discipline and stub management.
In shops we've compared, the continuous mechanisms carry roughly three quarters of consumption and breakage plus stub loss the rest. The vendor literature that puts sidewall oxidation near half of total loss is mostly describing bucket-charged furnaces with open roofs and long power-off; in a flat-bath shop running tight power-on, the tip term leads. Either way the variance between shops sits overwhelmingly in the breakage and regulation-driven terms. That's the pareto message: the baseline is physics, the difference between shops is discipline.
Which brings us to the DRI-shop advantage nobody invoices. Continuous flat-bath feeding removes the two most violent moments in an electrode's life: bore-in through a scrap bucket, and the cave-in, a hundred tonnes of scrap shifting against a graphite column. Scrap shops budget for breaks like weather. A well-run flat-bath shop treats every break as an anomaly with a name. If your shop feeds continuously and still breaks electrodes at scrap-shop rates, the furnace is telling you something specific: look at regulation response, hydraulic condition, and joint practice, in that order, because that's the order the evidence usually points.
The anatomy of a break repays the reading. Breaks at the joint say torque or nipple practice. Breaks below the holder after a current excursion say the regulation slammed the column into something the foam was hiding. Thermal-shock breaks after a long idle say water and heat met where they shouldn't. The protocol below turns each break from a budget entry into a diagnosis.
One honest caveat before the notebook: measure kg/t properly or the whole argument floats. Counting electrodes issued from stores and dividing by tonnes is bookkeeping, not measurement. Weigh columns, track additions per furnace, reconcile monthly. Half the "consumption improvements" we've seen presented were inventory-timing artifacts. The furnace deserves better arithmetic.
Operator's Notebook — six steps after every break
Run the full protocol every time, same order, no exceptions for "obvious" causes. Obvious causes are how patterns hide.
Freeze the evidence. Photograph the break face and the fragments before cleanup. Recover the broken piece if it can be done safely; the fracture surface separates mechanical overload from thermal shock from joint failure, and it's gone once it's in the scrap bin.
Pull the traces. Regulation and electrical records, five minutes either side: current excursions, arc voltage behavior, mast movement commands and actual response. Most break causes live in those ten minutes.
Inspect the joint history. Which nipple, who torqued it, to what value, with which wrench. Joint breaks cluster in torque practice, not in graphite quality.
Classify the location. At the joint, column below the holder, or tip region. Location plus fracture face plus traces triangulates the cause in most cases. Write the classification down even when it seems trivial.
Correlate the process moment. Feed-rate change, foam collapse, charge event, post-idle restart? A break has context; log it with the classification.
Register and review. Every break in one register: date, location class, assigned cause, trace snapshot reference. Review monthly for clustering by crew, by joint, by cause. Three breaks with one signature is a project, not bad luck.
Next week: water in the furnace, the one leak with no acceptable response time.
Written by active DRI-EAF operators. Anonymous by necessity, specific by design.