MaintenanceNewVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI reads motor current from the electrical cabinet to warn of failing conveyors, fans and pumps weeks or months ahead (ArcelorMittal)

Added 2026-10-29 · Last updated 2026-10-29

Manufacturers in this case: ArcelorMittal · all manufacturers using AI

Role

Maintenance & reliability (condition monitoring of motor-driven equipment) · Typical users: maintenance manager, reliability technician or planner, electrician, plant engineer

The problem

Many motor-driven machines are hard to monitor with handheld routes or vibration sensors: they are hot, dirty, moving, behind guards or simply too many to walk to. Faults in bearings, belts, couplings or rollers then show up as breakdowns or quality problems. At ArcelorMittal's hot strip mill in Ghent, Belgium, conveyors move plates of hot steel through the rolling process, and sensors mounted on the equipment cannot handle the temperatures. 'It's virtually impossible to use manual or vibration-based techniques to assess their condition,' says Andy Roegis, industrial digitalization manager for northern Europe at ArcelorMittal.

How AI helps

Instead of putting sensors on the machine, small high-frequency sensors are clipped onto the power cables inside the motor control cabinet (the electrical room where the motor starters and drives sit). They measure current and voltage 24/7. AI-based analytics first learn the normal electrical 'signature' of each motor and the machine it drives (about four weeks of automatic learning at ArcelorMittal), then watch for changes: a rising peak at the frequency of a belt, a bearing or the shaft rotation points to a specific developing fault. The system sends an alert with the suspected cause, maintenance inspects and repairs at a planned stop, and the data then confirms that the signature is back to normal.

Step by step

List the motor-driven machines whose failure stops production or hurts quality, and mark those that are hard or unsafe to reach for routes or vibration sensors (heat, height, guards, many identical units).
Pull 2 to 3 years of breakdowns and downtime hours for those machines from the CMMS, so you know which failures you want to catch and what an hour of stoppage costs you.
Check the motor control cabinets with an electrician: how each motor is started (direct, soft starter or variable speed drive), space for sensors and safe installation during a planned stop; ask the vendor which drive types and failure modes it supports.
Pilot on 5 to 10 machines with a known history, let the system learn for the period the vendor requires, and agree who receives alerts and how each one becomes an inspection work order.
Log every alert: confirmed or not, what was found, how far ahead, and what it would have cost if it had failed; review the log monthly with the vendor.
Extend to more machines only once the alerts have proved useful in your plant, and keep your safety inspections and routes for what the electrical signature cannot see.

Tools

Samotics SAM4 (sensors in the motor control cabinet, AI analysis of current and voltage): used at ArcelorMittal Ghent and at the wood-panel and food and beverage plants
Other tools to evaluate: other electrical signature analysis (ESA or MCSA) systems and portable motor testers; vibration-based monitoring where sensors can be mounted on the machine (see the related vibration case)
For the first step without new software: your asset list and CMMS failure history, plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below to choose the pilot assets

In the AI tools index: Samotics SAM4 · all AI tools used by manufacturers

Result

ArcelorMittal (steel, hot strip mill in Ghent, Belgium), Samotics SAM4: during a 12-month pilot on the motors driving the hot-strip conveyors, 12 developing faults were detected, all confirmed by the mill's maintenance engineers, some up to 4 months in advance ('That enabled us to schedule maintenance at the optimal time,' Roegis). ArcelorMittal then rolled the system out to 56 critical runout table rollers, where it detected 27 more failures, some up to 7 months in advance, which the mill estimates saved 31 hours of downtime (Samotics case study). Wood-based panel manufacturer (not named), SAM4 on eight exhaust fans of one production line: 6 developing faults detected over 15 months (fixed by replacing a clutch, impeller or bearings, cleaning an impeller, or welding holes in the fan housing), no unexpected fan failure in that period, up to 12 hours of unplanned downtime prevented at a cost of €8,000 per hour, or up to €96,000 (Samotics case study). Food and beverage manufacturer (not named): a slow rise at a cooling pump's rotation frequency led to an alert; three weeks later the trend spiked and the inspection found a coupling missing one of its bolts, fixed at the next planned stop: 'all we had to do was put in a new bolt' (Samotics case study).

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

All three stories are published by the vendor (Samotics). The ArcelorMittal page itself is inconsistent: its introduction credits the 31 hours of avoided downtime to the 12-month pilot, while the body credits them to the 27 failures detected after the rollout. The wood-panel and food plants are not named.
The downtime saved is the customer's own estimate; use your own downtime cost and failure history to judge the value, not the vendor's.
Installing sensors in a motor control cabinet is electrical work: only qualified electricians, with lockout and the plant's electrical safety rules.
Electrical signatures see the motor and what it drives, but not everything (leaks, guarding, structural cracks). Keep the safety inspections and rounds that the law or your insurer requires.
An alert is a reason to inspect, not a diagnosis: a person confirms the fault before parts are ordered or a line is stopped.

Try it yourself

You are a reliability analyst at a manufacturing plant.

I will paste: (1) our list of motor-driven machines [asset id, description, what it drives (conveyor, fan, pump, mixer...), motor power, starter type (direct, soft starter, variable speed drive), location and access (hot, high, guarded, outdoor...), whether a vibration route or sensor exists]; (2) the CMMS history for the last 24 to 36 months [asset id, date, failure description, downtime hours, repair notes]; (3) our cost of one hour of downtime per line, if known.

Do this:
1. Make a table of the machines with: number of failures, total downtime hours (copy the figures I gave, do not invent any), main failure types from the repair notes, access difficulty, and starter type.
2. Propose 5 to 10 machines for a pilot of motor current monitoring from the electrical cabinet, and explain each choice in one line (frequent failures, hard to reach, critical line). Flag machines on variable speed drives as 'CHECK WITH VENDOR'.
3. Write a one-page alert procedure: who receives an alert, how it becomes an inspection work order, what to record after the inspection (confirmed or not, what was found, how many days ahead).
4. List 10 questions to ask a vendor before a pilot (supported starters and drives, learning period, failure modes detected, installation during a stop, data access, price per motor).

Do not calculate savings or totals yourself: tell me which spreadsheet formulas to use. Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.

Want more like this? Join the newsletter

Related cases

All 12 Maintenance use cases →