ZenFactory · running factory · verifiable decisions
A factory that runs — and can back up every number
No slide and no mock-up: ZenFactory is a running industrial demo from the holding register of a 1998 machine to an auditable decision in the control loop. Industry 4.0 provides the data chain; Industry 5.0 shows how people, energy, resilience and governance turn it into controllable production.
loading plant data · 29 assets · OPC UA, Modbus, Sparkplug B · AAS · historian · Policy Gate · audit ledger
Human-centred decisions, sustainable energy flows, OT security and recovery behaviour are not merely claimed; they become visible from measured values, simulation, policies and user interactions.
Explore the demo
To an evidenced decision in five minutes
Order, tool wear, 70 kW limit, options, Policy Gate, human approval and measured effect, in one guided path.
Open short demo →Review the technology
Trace the data path, write path and evidence chain
OPC UA, Modbus, Sparkplug B, edge, historian, AAS, Policy Gate, canary and ledger, described as an auditable chain.
Open technology →Transfer
What of this is realistic in your production
Separated for operators and machine builders: legacy equipment, asset administration shell, engineering data and the limits of implementation.
Choose the right path →Short demo · 5 minutes
From problem to evidenced effect
01
Order under pressure
1,000 parts by 5 pm, ST020 wearing, 70 kW limit active.
02
Compare options
Reduce feed, shift the order, or change nothing.
03
Policy checks
Safety, quality, energy and write path are evaluated deterministically.
04
Human decides
No intervention without approval; autonomy stays off.
05
Evidence the effect
Canary, watchdog, expected-vs-actual and ledger show the result.
Experience Industry 5.0
A scenario that brings people, energy and resilience together
The Augsburg plant has to deliver 1000 parts by 17:00. ST020 shows tool degradation, while plant power is capped at 70 kW. ZenFactory shows several action options with trade-offs, lets the Policy Gate check them deterministically and records human approval before every controlled intervention.
Trigger
Tool RUL 47 ± 8 min · energy constraint active
Options
Reduce feed · move order share · no change
Governance
Policy Gate · Human Approval · Canary · Watchdog
Evidence
Expected vs Actual · Audit Ledger · rollback ready
I run a manufacturing operation
Your oldest machine has no interface. It can count. That is enough.
Almost every controller gives up counter readings, even one from 1998. From those comes a complete OEE per ISO 22400 — calculated at the edge, without touching the controller and without a new machine.
15 holding registers are enough. The band saw in this factory gets within half a percentage point of its machine’s true value.
Which problem do you have? →I build machines
Your customer demands an Asset Administration Shell. Your engineering data already has most of it.
Topology, interfaces and roles are in EPLAN or TIA. AutomationML lets you read them in, instead of reconstructing them in interviews with maintenance — the longest part of any integration.
134 property identifiers at property level, not at submodel level. Nameplate per IDTA 02006-2-0, read out rather than asserted.
What you have to deliver →Under laboratory conditions
± 1
cut prediction error of tool life, on average
With realistic noise
± 30
cuts — measured, not estimated
Both figures are on this page, with the script to recompute them. We publish the second one voluntarily because it would come out in operation anyway — the only question is whether it comes from the vendor or from a disappointed customer.
Intro call
90 minutes, and you leave with a list rather than an impression.
If you manufacture: which of your machines can deliver a
reliable KPI without modification — and which cannot.
If you build machines: what your engineering data
already yields for an Asset Administration Shell, and what is really missing.
You bring what you have: a machine list, a photo of the control cabinet, a screenshot of the controller, an EPLAN or TIA export. We go through it. At the end there is a list with three columns — works immediately, works with effort, doesn’t work — and the reason for every entry.
What it is not: a demo followed by a quotation. The demo is already here, around the clock, and you don’t need me for it.
Everywhere on this interface it says how reliable the number underneath is. What belongs to the next stage of expansion is drawn dashed. What an intervention achieved is recorded as a measurement in the evidence ledger, not as a promise. That is the test you should apply to every demo: whoever doesn’t state their limits hasn’t measured them.
Five plant types under one model — a chained production line, a parallel CNC cell, a high-speed filling line, a batch process plant and a machine from 1998 that only speaks Modbus. They share a nameplate and technical data and nothing else; a cell has no cycle, a reactor has no piece count, and the band saw doesn’t compute its own OEE — the connector at its side does. The cross-check is on Data paths, Technology and Model.