Every tire mixed, built, cured and graded on one autonomous loop
Nine specialist agents perceive, decide and act across mixing, building, curing and inspection — inside safety envelopes your engineers signed, on the lines you already own.
Built for tire OEMs and rubber-goods manufacturers
The plant is paced by craft that is retiring faster than it is replaced
Tire manufacturing is safety-critical, tolerance-bound and rework-heavy — and still run on manual building, fixed cure recipes and spot inspection.
Market figures from Treadum market analysis; pilot ranges are targets, validated per site before autonomy is activated.
A tire is the only thing between the vehicle and the road — and it is still built by hand
Mix to consistent rheology. Extrude tread and sidewall to exact gauge. Calender steel-cord plies. Place a dozen components and splice them soundly. Cure to exact cure state. Then prove none of it drifted.
Drift is invisible until it is scrap — or a field escape
A compound drift, a mis-splice, a misplaced component, an under- or over-cure, a bubble or a force-variation failure each drive costly scrap, rejects or a safety-critical escape. Plant managers run blind on real-time compound quality, splice integrity, cure state, defects, takt and tire genealogy.
- Manual, craft-bound building and mixing that cap rate and consistency
- Splice, placement, bubble, separation and bare-spot defects driving scrap
- Uniformity and balance failures causing rejects and rework
- Batch-to-batch compound drift degrading quality and safety margin
- A retiring tire-builder and process-engineer workforce that cannot meet rate
Perceive, plan, act, verify — closed at the edge
Treadum runs the same loop on every process step, from the Banbury mixer to final grading.
- Perceive: Fuse RGB and line-scan vision, thermal, X-ray, shearography, uniformity, PLC, encoder and historian streams into one aligned view of the compound, component and tire.
- Plan: Simulate the mix, extrusion, build, splice and cure against the as-built twin to pick setpoints that hit uniformity, quality and takt.
- Act: Write bounded setpoints and motion plans back through the control envelope your engineers signed — or hold for approval.
- Verify: Predict cure state, defects and force variation per tire, then confirm against downstream inspection and grading.
- Learn: Every accepted, modified and rejected action feeds the site model, gated by golden datasets and rollback rules.
Nine agents, one factory orchestrator
Each agent owns a process step and a hard metric. They share one twin, one genealogy record and one audit log.
Mix-and-Compound
Banbury and downstream mixing control — batch rheology, dispersion, temperature and energy per kilogram, tuned against the compound spec instead of a fixed recipe clock.
Owns rheology · dispersion · kWh/kg
Extrude-and-Calender
Tread and sidewall extrusion plus fabric and steel-cord calendering — gauge, width, profile and tension held to spec with closed-loop die and roll adjustment.
Owns gauge · width · tension
Build-and-Splice
Green-tire building — component placement, splice overlap and open-splice risk, bead seating and drum tension, verified per component instead of per audit sample.
Owns splice overlap · bead seat
Cure-and-Mold
Press and mold control — cure-state estimation per tire, adaptive dwell, mold allocation and preventable over-cure energy removed from the cycle.
Owns cure state · dwell · energy
Defect-and-Inspect
Vision, X-ray and shearography fusion for bubbles, separations, bare spots, cord shift and splice defects, with a hold decision made at the station.
Owns hold decision at station
Uniformity-and-Balance
Radial and lateral force variation, conicity and balance predicted upstream and graded downstream, so rejects are prevented rather than sorted.
Owns RFV · LFV · conicity
Plus Yield-and-Takt, Robot-and-Handling and Quality-and-Conformance, coordinated by the factory orchestrator.
See all nine agentsNot a dashboard. A loop that changes what the line does.
Treadum closes from perception to line-and-robot control, so the plant gets rate and yield instead of another report.
Simulate the run before you make it
The Omniverse-based tire-line twin models builders, presses, molds, conveyors, robots and WIP buffers, and predicts uniformity, quality and takt before a construction or recipe change reaches the floor.
<100ms
Decisions at the station
Jetson-class edge inference targets sub-100 ms hold and flag decisions so a defective tire is caught at the drum, not four hours downstream. [ASPIRATIONAL]
Batch to grade, linked
Every cured tire carries a linked record: compound batch, component rolls, builder, drum, press, mold, cure trace, inspection images and final grade.
- batch 4468
- drum 2
- press 14
- grade A
Trained on the failures you almost never see
Splice-open, bare spot, cord shift, bead mis-seat and mold contamination are rare by definition. Cosmos and Omniverse Replicator synthesize labeled variants, validated against real inspection signatures before any training use.
- splice-open
- bare spot
- cord shift
- bead mis-seat
- mold contamination
Mold, press and route solved together
cuOpt optimizes mold allocation, cure sequencing, WIP movement, AGV routing and line balancing under construction, due-date, capacity, quality and energy constraints.
Every action is evidence
Immutable, assurance-grade logs record what each agent saw, proposed, did and why — with citations to the recipe, spec or standard that justified it.
11:02:10 cure-and-mold · press 07 dwell −9 s · cites recipe C-42 rev 7
Land on one wedge. Own the loop.
Each entry point has a hard, measurable baseline and a clear path to expansion.
Catch the splice at the drum
The build-and-splice agent measures overlap, gap and open-splice risk on every component instead of on an audit sample, and corrects drum tension and placement before the green tire leaves the station.
- Per-component splice measurement, not sampled
- Open-splice risk scored before cure
- Drum tension and placement corrected in the loop
Cure to state, not to a clock
The cure-and-mold agent estimates cure state per tire from press telemetry, mold history and compound batch, then adapts dwell instead of running a fixed recipe with a safety pad.
- Per-tire cure-state estimation
- Adaptive dwell inside a signed envelope
- Over-cure energy removed from the cycle
Fuse the inspection stack
The defect-and-inspect agent aligns X-ray, shearography, thermal and camera streams on one timeline so a separation signature is localized to a component and a build station, not just to a tire.
- Bubbles, separations, bare spots, cord shift, splice defects
- Localized to component, builder and batch
- Hold decision made at the station
Predict force variation before the machine measures it
The uniformity-and-balance agent predicts radial and lateral force variation and conicity from build and cure evidence, so grading confirms a prediction instead of discovering a surprise.
- RFV, LFV and conicity predicted upstream
- Grade confirmed downstream, not discovered
- Root cause traced to component and station
Treadum speaks the plant’s protocols on day one
We integrate with the mixing, building, curing, inspection and MES systems already installed, under your IT/OT segmentation rules.
Read broadly. Write narrowly.
Treadum ingests from PLCs, drives, presses, inspection systems and historians, and writes back only through the setpoints your controls team exposes. Nothing is written outside the envelope, and every write is logged with the evidence that justified it.
- OPC UA, MQTT Sparkplug B, Modbus TCP and EtherNet/IP
- Historian connectors for OSIsoft PI, Aveva and InfluxDB
- MES and QMS integration for genealogy and non-conformance
- X-ray, shearography and uniformity machine APIs
- ROS 2 and Isaac bridges for robots, AGVs and press handling
# factory-edge connector manifest
site: "plant-nl-02"
sources:
- kind: opcua
endpoint: "opc.tcp://curing-gw:4840"
nodes: ["press.*.temp", "press.*.dwell"]
- kind: xray
vendor: "station-xray-2"
stream: "grpc://xray-2:7443"
actuation:
envelope: "signed/plant-nl-02-v7.json"
mode: shadow # shadow | assist | boundedWhat plants say after the first shadow-mode week
“We had three people watching X-ray images and still shipped uniformity rejects. Treadum flagged the cord shift at the building drum, not four hours later at final inspection.”
“The cure agent found 40 seconds of margin on a construction we had run the same way for eleven years. It proved it in shadow mode before it touched a press.”
“Genealogy is the part I did not expect to care about. Every tire now has a linked record from batch to grade, and audit prep went from weeks to an afternoon.”
[PLACEHOLDER] Design-partner quotes are illustrative until pilot references are published.
Built for the constraints a tire group actually has
Recipe IP, network segmentation, audit evidence and multi-site rollout are first-class, not an afterthought.
Recipe protection
Compound, construction and tolerance data stays in your tenant. On-premises deployment available for sites that cannot send plant data outbound.
Security and trustSegmented by design
Factory-edge inference runs inside the OT network. Cloud is used for training and fleet management, never as a dependency for line decisions.
EnterpriseAudit-grade evidence
Immutable logs, citations and genealogy wired to IATF 16949, ISO 9001, DOT and UNECE R30 and R117 evidence needs.
ComplianceMulti-site scale
One control plane for a group: shared models where you want them, site-isolated where you do not, with a managed factory edge fleet.
Scale a groupThe three questions every plant asks first
No. Treadum runs on top of the equipment you already own. We read from your PLCs, drives, inspection systems and historian over OPC UA, MQTT and vendor APIs, and write back only through the setpoints and interlocks your controls engineers explicitly expose.
Every action is bounded by a control envelope with hard limits your process engineers sign off. Out-of-envelope proposals are held for human approval, and every decision — accepted, modified or rejected — lands in an immutable audit log that feeds the next training cycle.
You do. Recipe, construction and tolerance data stays in your tenant, models trained on it are yours, and nothing crosses tenant boundaries. Sites that require it can run entirely on-premises with no outbound plant data.
Put one cell on autonomy in 90 days
Pick one wedge — splice inspection, X-ray defect detection, cure-state optimization or uniformity prediction. We baseline it, run shadow mode, then graduate to bounded autonomy under a signed control envelope.
Pilots start in shadow mode. No line changes until accuracy and safety gates pass.