Digital twins — virtual representations of physical assets, processes, or systems that enable real-time monitoring, simulation, and optimisation — have reached enterprise production maturity in 2026. ISO 23247 established the first international standard for digital twin frameworks in manufacturing, providing a vendor-neutral reference architecture that enables enterprise digital twin programmes to be designed for interoperability rather than single-vendor lock-in. This guide covers the ISO 23247 framework, the technology components, and the enterprise deployment patterns that deliver measurable ROI.
ISO 23247: The Digital Twin Standard
ISO 23247 — Framework Overview
ISO 23247 is a four-part international standard for digital twin frameworks in manufacturing, published by ISO/IEC JTC1. It defines: Part 1 — Overview and general principles; Part 2 — Reference architecture; Part 3 — Digital representation of manufacturing elements; Part 4 — Information exchange. The standard establishes four functional entities in a digital twin system: the Observable Manufacturing Element (the physical asset), the Digital Twin (its virtual counterpart), the User Entity (humans and systems consuming twin data), and the Cross-cutting Functions (communication, security, data management) that connect them.
Digital Twin Architecture Layers
| Layer | Function | Technologies |
|---|
| Physical Asset Layer | The real-world equipment, process, or system being twinned | OPC-UA, MQTT sensors, PLC/SCADA data, CAD models |
| Data Acquisition Layer | Collects real-time data from physical asset via IoT and protocols | AWS IoT Greengrass, Azure IoT Hub, Siemens MindSphere connector |
| Digital Twin Core | The virtual model — 3D geometry, physics simulation, data bindings | NVIDIA Omniverse, Azure Digital Twins, Siemens Teamcenter |
| Analytics Layer | AI/ML models running on twin data — anomaly detection, predictive maintenance | Azure ML, AWS SageMaker, Python ML pipelines |
| Application Layer | User-facing interfaces — dashboards, VR walkthroughs, operator alerts | Power BI, custom React apps, Unity/Unreal for immersive |
Enterprise Use Cases with Proven ROI
25%
Reduction in unplanned equipment downtime from predictive maintenance enabled by digital twin anomaly detection — the highest-value single use case for manufacturing digital twins
15%
Energy efficiency improvement in facilities using building digital twins for HVAC and power optimisation — measurable in 6–12 months post-deployment
40%
Reduction in physical prototype iterations for product development teams using digital twins for simulation — replacing expensive physical tests with high-fidelity virtual equivalents
🏭
Predictive Maintenance
The highest-ROI digital twin use case. Mirror CNC machines, turbines, motors, or production lines in real time using OPC-UA/MQTT sensor data. Train anomaly detection models on historical failure signatures. Alert maintenance teams 12–72 hours before failure occurs. Average ROI: $3–5 saved for every $1 invested in predictive maintenance digital twins. Connect to your
ERP work order system for automated maintenance scheduling.
🏗️
Facility Digital Twin
3D model of a building or factory floor combined with real-time sensor data — HVAC telemetry, occupancy sensors, energy metering, security systems. Enables: remote facility management, energy optimisation via AI-driven HVAC control, space utilisation analytics, emergency simulation. Microsoft Azure Digital Twins is the most deployed enterprise facility twin platform in 2026.
🚢
Supply Chain Twin
Real-time mirror of goods in transit — location, condition (temperature, humidity), estimated arrival, customs status. Enables: exception management (alert when cold chain broken), delivery prediction, dynamic rerouting around disruptions. Significant ROI for pharmaceutical, food, and luxury goods companies where supply chain visibility directly impacts customer outcomes.
⚙️
Process Optimisation Twin
Simulate manufacturing process changes in the digital twin before implementing physically. Test: new tool paths, material changes, line speed adjustments, recipe modifications. Reduce changeover time and eliminate physical trial runs. NVIDIA Omniverse Isaac Sim enables high-fidelity physics simulation of manufacturing processes for design-of-experiment workflows without production disruption.
01
Phase 1
Single Asset Pilot
Start with one asset — the highest-value machine or most failure-prone equipment. Deploy IoT sensors if not already present (vibration, temperature, current draw). Connect to Azure IoT Hub or AWS IoT Core. Build a simple real-time analytics dashboard showing current asset state vs baseline. Run for 60 days, identify anomaly detection opportunities. This pilot proves value and builds the team capability for scale-out.
Single asset pilotIoT sensor deployment60-day baseline
02
Phase 2
Add Predictive ML and Scale
Train anomaly detection on 90+ days of baseline data. Deploy ML model (Isolation Forest, LSTM, or purpose-built tool like Azure Anomaly Detector) for real-time alerting. Roll out to 10–50 additional assets using the proven pattern. Connect alerts to your maintenance management system. At 50+ assets, the ROI calculation typically shows 18-month payback on the full digital twin programme investment.
Anomaly detection MLCMMS integrationScale-out pattern