Plant data is everywhere. Boiler chemistry sits in the BHEL HMI. Steam temperature lives in the Emerson Ovation DCS. Three years of vibration history archives in AVEVA PI. Work orders, spares, and unit availability are buried in SAP PM. Each system answers its own question well, and almost none of them talk to each other in a way an AI can use. iFactory ships an enterprise-grade AI server straight to your plant — pre-configured, pre-tested, racked and ready, with every connector, model, and dashboard pre-loaded. You power it on, connect Ethernet, and the AI starts reading every system within days. Get a quote — and a delivery date.
One AI Brain. Eight Plant Systems. 6–12 Weeks From Order to Live.
Order a fully integrated, AI-powered plant intelligence system — deployed on enterprise-grade AI hardware at your premises, pre-configured, pre-tested, and ready to run within days. The appliance lands racked and loaded with all eight integrations, models, and dashboards. You connect power and Ethernet. The AI starts reading your plant. Perpetual license. Full source access. Total data sovereignty on-premises.
Hardware + Software + Integrations — Shipped as One Appliance
This is a physical product, not a cloud subscription. iFactory delivers an enterprise-grade AI server, racked and ready, with the operating system hardened, all eight plant integrations pre-configured, the AI models pre-loaded against your tag dictionary, and the dashboards ready for your operators. You receive it, you power it, you connect it to your network — the AI starts reading your plant within days of unboxing.
Enterprise-grade AI server shipped to your plant, racked and pre-configured. NVIDIA GPU, hardened OS, all integrations loaded, models pre-trained. Power on, connect Ethernet, the AI starts reading your plant within days.
No monthly fees, no per-seat charges, no recurring license costs. Ever. Once purchased, the platform is yours. Future costs are entirely optional and at your discretion.
All data lives on your server, behind your firewall. No data leaves your network unless you choose. Compliant with regulator data-localisation rules and corporate IT policy.
Modify, extend, and customise the platform freely within your organisation. Your team adapts the system as the business evolves — with or without iFactory involvement.
Predictive maintenance, anomaly detection, NLP-driven work orders — built into the platform from day one. Not bolted on. Not an add-on tier.
Coal-Fired Unit 6 · 500 MW · BHEL Boiler · Emerson Ovation 4.0 · AVEVA PI
This is the actual configuration of a reference deployment. The architecture, integrations, and tag counts shown below are not theoretical — they reflect a live unit currently operating with iFactory in the loop.
From Field Sensor to ERP — All Eight Layers, One Integration
The integration package follows the ISA-95 automation pyramid, but reads laterally — every system is a peer in the AI's data graph, not a hierarchy in a corporate diagram. Each system is connected through its native protocol; nothing is bolted on with a screen-scraper.
Native Modbus TCP and OPC-UA from the BHEL boiler HMI. Steam temperature, drum level, mill status, ID/FD/PA fan currents, soot blower sequence, and burner tilt — every tag, every second.
OPC-UA secure subscription to the Ovation 4.0 native server. 18,400 I/O points across boiler, turbine, generator, BOP, emissions, and stack monitoring — read-only with full quality codes.
AVEVA Adapter for OPC-UA to PI Data Archive. 142,000 PI tags streamed in real-time plus 5 years of historical data for model training. PI Asset Framework hierarchy preserved for context.
Direct integration with the System 1 server. Turbine rotor dynamics, generator bearings, BFP, ID/FD fan vibration spectra. FFT envelope and waveform data available for the PdM model stack.
Six stack analysers — NOx, SOx, CO, CO₂, O₂, particulate. Modbus TCP and OPC-UA. Compliance reporting feeds and emissions optimisation models read from the same stream.
REST API and OData connectors. Work order generation, spare parts material master, functional locations, equipment hierarchy, and outage history — all bidirectional.
Coal calorific value, ash, moisture, sulphur from the laboratory information management system. Combustion optimisation models use lab results to predict mill load and ash handling.
Schedule of generation from the regional load dispatch centre. The AI uses the dispatch curve to optimise ramp rate, forecast fuel demand, and pre-position auxiliaries before each load change.
How 1.2 Million Events Per Hour Become One Decision
The integration is not a passive data lake. Each layer transforms the upstream data into something the next layer can act on, with sub-10ms latency end-to-end on the on-site appliance.
Eight native protocol adapters running in parallel. OPC-UA, Modbus TCP, IEC 60870, REST, OData, AVEVA PI Adapter, System 1 API, SAP RFC. Each adapter handles its own subscription, quality codes, and reconnection.
Tag identifiers from each upstream system are mapped to a unified asset model — your Asset Framework hierarchy plus engineering units, quality codes, and time-zone normalisation. 99.9% mapping accuracy after calibration on your tag dictionary.
On-appliance time-series database with edge retention policy. Hot tier stays in memory for sub-10ms inference. Warm tier on local NVMe for 90 days. Cold tier writes back to AVEVA PI for compliance — your historian remains the system of record.
LSTM degradation models, FFT envelope analysers, multi-sensor fusion, and an on-device LLM for plain-language summaries — all running on the local NVIDIA GPU. Inference happens against the unified data graph, not against any single system in isolation.
SAP PM work orders raised through OData. Setpoint candidates pushed back to Ovation on a write-confirm pattern — operator approves before any setpoint moves. Compliance reports exported to PI for audit. Every action logged with operator ID and timestamp.
An Hour in the Feed — What 1.2M Events Look Like in Motion
A representative sample of tags currently streaming from the reference plant into the AI brain. Tag names follow AVEVA PI conventions; values are illustrative. The point is the variety — the AI reads boiler chemistry, turbine vibration, ERP work orders, and dispatch schedule on the same time axis.
From Purchase Order to Live AI — 6 to 12 Weeks
Two timelines run in parallel. The full PO-to-go-live cycle is 6–12 weeks — engineering scope, hardware build, pre-test, ship, install, calibrate, train. The unbox-to-operational cycle, once the appliance lands at your plant, is days — because everything is pre-configured, pre-tested, and pre-loaded before it ships. Most of the 6–12 weeks happens before the appliance ever reaches your site.
Tag dictionary export from your PI System and Ovation. Functional location export from SAP. Network architecture review. Firewall rules drafted with your IT team. Signed scope document — no scope creep after this point.
Your appliance is built — NVIDIA GPU installed, OS hardened, all eight integration adapters loaded, models pre-trained on your historian export. Full integration test in our lab against a digital replica of your tag set before shipment.
Appliance shipped globally — racked, pre-configured, pre-tested. Field engineer dispatched for rack install, OPC-UA endpoint configuration, AVEVA PI Adapter setup, SAP connector authentication. The AI is reading your plant within days of the appliance arriving on site.
Tag mapping calibration to 99.9% accuracy. Models retrained on live data from your unit. Pilot detection and recommendation feed reviewed with your operators and reliability team. Validation mode — no setpoint writes yet.
Full production mode. Write-confirm to Ovation activated for approved categories. SAP PM work order generation live. Operator and engineering training delivered on-site. Source code, documentation, and admin handover complete. You own it.
The Math of Owning Your AI Stack
Cloud SaaS vendors price by the tag, by the user, by the asset, by the API call — and the price grows with your data. iFactory is a one-time capital purchase. Over five years, the difference is significant. Over ten, it is structural.
- Annual subscription tied to tag count — grows every year
- Per-user seat fees for engineers, operators, managers
- Data egress fees from your plant to the vendor cloud
- Loss of access if subscription lapses or vendor changes terms
- Your operational data sits on third-party infrastructure
- Customisation locked behind professional services billing
- One-time hardware + license purchase, no recurring fees
- Unlimited users, unlimited tags, unlimited assets
- All data on-premises — zero egress, zero exposure
- Platform stays operational regardless of vendor relationship
- Your data never leaves your firewall
- Your team modifies and extends within your organisation
What Plant IT & Automation Engineers Ask Before Sign-Off
Both — your choice per category. By default, all integrations are read-only on initial commissioning. Write-back is enabled only for explicitly approved categories (e.g. soot-blower scheduling, mill load setpoints) and always on a write-confirm pattern — the operator approves the setpoint candidate in the HMI before it executes.
Common scenario, fully supported. We use a protocol converter or SCADA-layer gateway to bridge legacy Modbus, DNP3, or proprietary serial protocols to OPC-UA at the plant edge. The AI brain only ever sees OPC-UA on its inbound side — protocol diversity is handled before it reaches the AI.
No, the PI System remains your historian and system of record. iFactory reads from PI for historical model training and writes back compliance-grade data to PI for audit. The AI's hot-tier in-memory cache exists for sub-10ms inference; PI is the cold-tier system of record.
Your platform keeps running. You have the perpetual license, the source code access, the trained models on disk, and the documentation for the integration adapters. Your team — or any qualified integrator — can continue to operate, modify, and extend the platform indefinitely. Vendor independence is the point.
Order Your AI Package. Take Delivery in 6–12 Weeks. Own It Forever.
Two ways to take the next step. Get a fixed-price quote — our integration architects map your specific DCS, historian, ERP, and control stack to the iFactory package, you leave with a quote, a delivery date, and a signed scope document. Or join our upcoming live event to walk through every AI server configuration we ship — see the exact hardware spec, the integration adapters, the pricing tier that matches your plant. Either path leads to the same place: PO signed, appliance shipped, you own it forever — perpetual license, full source access, all data on your premises.
