SKIP TO CONTENT
minto.ai
PROOF · THE RECORD

The question you actually have is not “does your AI work.” It is “have you stood inside a plant like mine.”

READ THIS FIRST — THE ERA LINE

The five projects below were delivered with industrial sensing, signal processing and machine learning — before large language models existed. The agentic platform did not produce these outcomes: the system found the fault, and a human did the reasoning. The hard part was always the plant, not the model. Ten years of reading machines in environments that defeat conventional sensing is what the reasoning layer now sits on top of. New capability, same ground truth.

Rail-sleeper insert foundry

KALLAKAL · PATIL GROUP · SAND CASTING

A 6,000 tonne-a-year sand-casting foundry built around a 45 kW sand mixer, plus pumps, conveyors, sieves, vibro motors, exhaust fans, shot blasts and cranes. Airborne sand, heat, impact loading and electromagnetic interference defeat conventional sensing through sensor failure, cabling and poor machine access.

Mapping load paths and drive-train mechanics against fault history showed that instrumenting under half the machines would head off around 80% of unplanned stoppages.

OUTCOME — THE CUSTOMER'S OWN FIGURES · A quarterly 83 hours of lost production — about 57,000 inserts — cut by roughly 90%, with some months running downtime-free.

Blooming mill

JAMSHEDPUR · TATA STEEL · 10 MnTPA INTEGRATED WORKS

Scoped as predictive maintenance, then reframed around recurring hot-saw failures once the investigation went deeper. Radiant heat from red-hot steel, heavy torsional load during cuts and repeated shock loading produced unpredictable shaft breakages and erratic blade life — conditions that destroy mounted sensors and drown their signal.

OUTCOME — THE CUSTOMER'S OWN FIGURES · Four significant hot-saw faults caught across a February-to-May window — and the pattern behind them exposed a structural weakness that led to a full frame redesign.

The engagement started as monitoring and became a design fix. That is what reasoning about causes produces — and what alerting never does. Diagnosis, not monitoring.

Hot strip mill

BOKARO · SAIL · 4.5 MnTPA AUTOMATED MILL

Walking-beam reheat furnaces, six roughing stands, six finishing mills, and EOT cranes with dual-motor hoists driving one drum through coupling and gearbox. Constant load variation, rapid start-stop duty and a moving platform make sensor mounting impractical.

OUTCOME — THE CUSTOMER'S OWN FIGURES · Within three days of installation, motor winding problems and gearbox anomalies were identified — including intermediate gearbox issues and motor stalling on the dual-drive hoists.

Sinter plant

BHILAI · SAIL · THE BRIDGE PROJECT

Hammer crushers, rod mills, an extensive conveyor network and agitators under high torsional load, in a heavily dust-laden environment with HV equipment in hard-to-reach positions and legacy assets running past design life. The machine mix runs from HV hammer crushers on fluid couplings to LV nodulizer agitators on VFDs.

This is the bridge: it began in the same sensing era — and it is now the site where the daily equipment intelligence is delivered live, across 50 machines, before every shift. The turn is described below.

PVC compounding

THIMMAPUR · HIL LIMITED

Hot and cold mixers in tandem — the hot mixer running at 20 to 50 m/s circumferential speed and fluidising the blend to 130–140 °C, the water-cooled mixer bringing it back to 40 °C, both on VFD-controlled motors under thermal and torsional stress. Blade wear develops as subtle patterns, and VFD operation undermines conventional vibration analysis.

The commercial risk is downstream: bad compound reaching the extruders can seize them for two to three days.

OUTCOME — THE CUSTOMER'S OWN FIGURES · The system flags shifts in blade performance and mixing efficiency before batch quality is affected.

Reliability is not always about downtime. Here the loss is silent — in scrapped batches — and one prevented seizure carries the business case.
THE TURN

Same plant. Same sensing. A different kind of morning.

What changed between the two eras at the Bhilai sinter plant, in concrete terms:

BEFORE — THE SENSING ERA
The system found the fault on a machine.
A human did the reasoning: cause, connections, what to check, in what order.
Findings arrived per machine, when they fired.
Whether a repair worked lived in someone's memory.
NOW — THE LIVE DEPLOYMENT
The system reasons across the whole plant — 50 machines, 5 areas, shared panels, the water utilities included.
It ranks the findings, assigns them to named crews with decision rules, and opens with a production decision.
It arrives before every shift, covering everything — including the machines with no data.
It closes the loop with the field: prior actions carry outcomes, repairs carry standing verdicts, and it corrects its own calls in the open.
DEPLOYMENT FOOTPRINT
BHILAI
Sinter plant · live daily intelligence
BOKARO
Hot strip mill · crane hoists
JAMSHEDPUR
Blooming mill · hot saws
KALLAKAL
Foundry · sand casting
THIMMAPUR
PVC compounding · mixers
Client organisations include SAIL, TATA Steel Long Products, Patil Group and HIL Limited.
“The morning picture ended the argument about what to do first. That alone changed the shift.”
HEAD OF RELIABILITY · INTEGRATED STEEL PLANT · PLACEHOLDER
“It told us it had named the wrong machine, and then named the right one. That is when the team started trusting it.”
AREA MAINTENANCE HEAD · SINTER PLANT · PLACEHOLDER
“One device in the cabinet, and we see the whole drivetrain. Nothing bolted to the machine.”
ELECTRICAL SUPERINTENDENT · STEEL LONG PRODUCTS · PLACEHOLDER