A graphite electrode that breaks mid-heat stops the furnace, puts crews to work near a hot shell and wastes graphite that was already paid for. Most breaks are not random. They follow additions, heavy charges and columns that have been vibrating harder than usual for several heats. EAF electrode breakage prediction reads those signals and warns the melt shop in time to act at the next tap. See it on furnace data in a short melt shop demo.
Hear About an Electrode Break Before It Stops the Heat
iFactory models arc stability, regulation behaviour and column vibration for each phase. When breakage risk rises, the voice agent speaks to melt shop headsets and logs a check task for the next tap.
Small Share of Graphite, Large Share of Trouble
In a well-run shop, most electrode consumption is continuous. The tip is used up by the arc and the sides oxidise in the furnace atmosphere. A widely cited industry estimate puts that at about 95% of consumption. The other 5% or so is discontinuous: breaks at the socket, pin or body, butt losses and spalling. That share is small in tonnes of graphite, but each break also costs furnace time, crane time and a risky job near the shell. That is why breakage is worth predicting on its own. Our melt shop team can help you separate the two in your own records.
Six Causes, and What Each Looks Like in the Data
Industry experience points to two leading causes: the way electrodes are added, and the way scrap is mixed and loaded, including nonconductive material in the charge. Studies of column joints add thermal stress, over-tightening, vibration and side impact from the charge or furnace tilting. Each cause leaves a different trace in the data, which is what makes prediction possible. Map these causes to your furnace in a breakage review session.
Seven Data Streams That Warn of a Break
Most of these signals already exist in a melt shop. The power system measures arc current, the regulator records every electrode move, and the mast cylinders carry pressure transducers. What is usually missing is the step that puts them together per phase and per heat, and compares them with the history of past breaks. Vibration sensors on the arms are the one item some shops need to add. Check what your furnace already records with our support team.
When in the Heat a Break Is Most Likely
Breakage risk is not spread evenly across a heat. It peaks when the electrodes bore down into a fresh charge, with scrap shifting and the arc unstable. It falls once a flat bath forms and the arc is covered by foamy slag. The pattern below is typical for a scrap-based EAF, and the model learns the actual shape for each furnace. Compare it with your own heat logs in a heat profile review.
Warnings That Reach the Pulpit, Not a Report
A risk score in a dashboard is no use during bore-down. iFactory speaks the alert to melt shop headsets, names the phase and the driver, and lets the melter ask questions without taking their eyes off the furnace.
One Heat, One Warning, No Break
Here is an illustrative heat on a two-bucket furnace. The times and scores are examples, but the sequence is a common one: a column that was joined recently, a heavy second charge and vibration that builds during bore-down. The point is that the warning arrives while there is still a choice, not after the break. Ask our support team to replay a heat from your own history.
An Electrode Event Log Worth Keeping
A model learns from past breaks, so the quality of your break records decides how good the predictions can be. Many shops log a break as a single line in the shift report. The fields below turn each break, and each near miss, into a record the model and your engineers can use. Start logging them now, even before any AI is in place. Get a ready-made version in a setup call.
How to Size the Value of Fewer Breaks
The cost of a break has four parts: lost furnace time, the graphite left in the bath or scrapped, crane and crew time to recover the column, and the safety exposure of working near the shell. Furnace time is usually the largest, and only you know what a minute of your furnace is worth. The example below uses assumed inputs to show the arithmetic. Replace each one with your own figures, and our team can help you complete the case.
Multiply the recovered time by the value of a furnace minute in your shop, then add the graphite and recovery labour saved. Leave the safety benefit out of the sum, but keep it in the case.
Prediction Works Best Alongside Good Practice
A warning helps most when the basics are in place. These habits reduce breaks on their own and make the remaining ones easier to predict, because the data is cleaner and the causes are fewer. If you want a second opinion on your practices, talk to our melt shop specialists.
- Clean threads and pins before every joint
- Tighten to the supplier's torque, not more
- Record who joined each section and when
- Track breaks by supplier, grade and batch
- Keep heavy scrap away from the electrode path
- Keep nonconductive material out of the buckets
- Layer buckets the same way every time
- Record anything unusual in each charge
- Keep regulation tuned and check for hunting
- Check column vibration on tall shells
- Aim oxygen away from the electrodes
- Inspect arms, clamps and holders on schedule
Models, Voice Alerts and AI Server in One Package
iFactory delivers electrode breakage prediction as a turnkey bundle. A pre-configured NVIDIA AI server ships racked and ready with the models, furnace data connectors and the voice agent pre-loaded, so high-speed furnace data is processed on site. You rack it, plug in power and Ethernet, and the AI is live. To scope it for your melt shop, request a turnkey quote.
- Pre-configured NVIDIA AI edge server, racked and ready
- Connectors for power metering, regulator and furnace PLC
- Arm and mast vibration sensors where needed
- Breakage risk models trained on your heat history
- Voice alerts to melt shop headsets and the pulpit
- Check tasks logged to the CMMS for the next tap
- 24×7 remote monitoring of models and hardware
Server installed, furnace signals connected, past breaks and additions loaded.
Models trained per furnace and run silently, with risk scores checked against real events.
Voice alerts switched on, thresholds agreed, melters and crews trained.
Breakage Prediction Built for the Melt Shop
iFactory's Predictive AI and voice agent are designed for the pace of an EAF, where decisions happen in minutes and nobody has time to read a screen. Our support team can walk you through a pilot plan.
What Plant Managers Ask About EAF Electrode Breakage Prediction
Fewer Broken Columns, Fewer Lost Heats
Connect your furnace signals, load your break history and let iFactory warn the melt shop while there is still time to act. The full system is delivered turnkey and goes live in 6 to 12 weeks.
