One idea, applied to very different assets

A digital twin follows a real facility or system through live data, and lets you test changes on the model first. The asset differs from one industry to the next. The loop is the same.

  • The model starts from your drawings and records, then is corrected against what is physically there.
  • Live data keeps it in step with the asset, so it describes today and not the day it was built.
  • Changes and failures are tried on the model, and only the decision that holds up reaches the real asset.
The digital twin loop The physical asset sends live data to the twin model. The twin is used to simulate changes. The chosen decision is applied back to the physical asset. Physical asset Plant or building Twin model Kept live by data Simulate Test changes Decide Pick what holds up Live data Change applied
The loop is the same in every domain. Only the asset and the data change.

Pick an industry

For each one: what the twin represents, the data it uses, and the questions it can answer. You can link straight to one, for example digital-twin.html#manufacturing.

Data centres

The facility modelled rack by rack, with the power chain and airflow that serve each one.

What the twin represents

Halls, rows and racks; the power path from utility and generator through UPS and PDU to each rack; cooling units, containment and airflow.

Live data it uses

  • UPS, PDU and generator readings
  • Cooling unit and chiller status
  • Rack-inlet temperature and humidity
  • BMS and EPMS points over Modbus, SNMP or BACnet

Questions you can ask it

  • Can Hall B take 20 more racks without breaching its power or cooling limit?
  • What happens to inlet temperatures if one cooling unit fails?
  • Which row is the best home for a new high-density cluster?

What it improves. Capacity goes where power and cooling already exist, and failures are rehearsed on the model, not on the live floor.

This is what our DCIM already does. See the DCIM product.

Manufacturing

A plant modelled as machines, buffers and flows, so a change can be tried before the line is stopped.

What the twin represents

Production lines, individual machines, conveyors and buffers, and the way material and people move between them.

Live data it uses

  • PLC and SCADA tags
  • Vibration, temperature and current on key machines
  • Cycle-time and OEE counters
  • Inspection and quality results

Questions you can ask it

  • What does a new product mix do to line 2 throughput?
  • If this machine goes down for maintenance, where does the queue build?
  • Does adding a buffer here remove the bottleneck, or just move it?

What it improves. Maintenance is planned around production, and layout changes are tested against the real constraint rather than the assumed one.

Buildings and campuses

Floors, plant rooms and systems modelled together, so comfort and energy use can be traded off with evidence.

What the twin represents

Floor plans, HVAC zones and plant, lighting, lifts, and how many people use each space and when.

Live data it uses

  • BMS points and energy meters
  • Occupancy and air-quality sensors
  • Access-control counts
  • Weather feed

Questions you can ask it

  • What do new setpoints and schedules do to energy use and comfort?
  • How does a change in occupancy affect cooling load on each floor?
  • What does a chiller replacement save over its first year?

What it improves. Energy savings are checked on the model before any setpoint changes, and retrofit cases rest on measured baselines.

Energy and utilities

Network assets modelled with their loading history, so growth and faults can be planned for.

What the twin represents

Substations, transformers and feeders; solar plants, inverters and battery storage; and the loads they serve.

Live data it uses

  • SCADA and RTU telemetry
  • Smart-meter data
  • Inverter and battery status
  • Transformer temperature and loading

Questions you can ask it

  • Can this feeder take the projected load growth, and when does it stop being able to?
  • How do we re-route supply if a section is lost?
  • When should the battery charge and discharge to cut peak demand?

What it improves. Ageing assets are replaced in order of real risk, and renewable capacity is added where the network can absorb it.

Smart cities

Shared infrastructure modelled across departments, so one change is seen against the whole system.

What the twin represents

Road junctions and traffic flow, water and drainage networks, street lighting, and public facilities.

Live data it uses

  • Traffic sensor and camera counts
  • Flow, pressure and level meters
  • Street-light controller status
  • Event and incident feeds

Questions you can ask it

  • What do new signal timings do to queues on the main corridor?
  • Which valves isolate a pipe burst with the fewest customers affected?
  • How does a large public event change load on transport and water?

What it improves. Operators see the effect of a change across systems before it reaches residents.

Healthcare facilities

The hospital building and its critical services modelled so they can be tested without touching patient care.

What the twin represents

Wards and theatres, power and medical gas, pressure-controlled air zones, and the movement of equipment and beds.

Live data it uses

  • BMS and power-system status
  • Theatre and ward environmental sensors
  • Equipment location tags
  • Alarm logs

Questions you can ask it

  • What is the impact on one wing if a power feed fails?
  • How does a surge in admissions load each ward and its supporting plant?
  • Which equipment is idle, and where is it?

What it improves. Critical systems are rehearsed on the model. The twin models the facility, not individual patients, so patient records stay out of it.

Logistics and warehousing

A warehouse modelled as racking, routes and handling equipment, so throughput changes can be tried first.

What the twin represents

Racking and slotting, docks, conveyors, vehicles and robots, and temperature-controlled zones.

Live data it uses

  • WMS transactions
  • Scanner and RFID reads
  • Vehicle telematics
  • Temperature loggers

Questions you can ask it

  • What does a new slotting plan do to pick times?
  • Can the site handle peak-season volume with the current docks?
  • Where does a cold-room failure put stock at risk?

What it improves. Layout and staffing decisions are tested against real volumes, and cold-chain exposure is found early.

Telecom and edge sites

Distributed sites modelled as a fleet, so a fault at one is seen in the context of all the others.

What the twin represents

Towers and shelters, power systems with batteries and generators, cooling, and the links between sites.

Live data it uses

  • Network-management alarms
  • Rectifier, battery and generator status
  • Shelter temperature and door sensors
  • Fuel levels

Questions you can ask it

  • How long can this site run on batteries if the grid drops?
  • Which sites need a capacity upgrade first?
  • What is affected downstream if this link fails?

What it improves. Site visits and fuel runs go where they are needed, and weak sites are fixed before they cause an outage.

Edge sites can be built as container data centres.

How we build one

The same five steps in any domain. Each one ends with something you can check.

Survey and scope

Agree the decisions the twin must support, then walk the asset and its records to see what we have to work with. A twin built to answer a named question stays small enough to finish.

  • Decisions and questions to support
  • Drawings, as-built records and asset lists
  • Existing control and monitoring systems
  • Gaps in measurement

Model

Build the structure of the asset and the rules that govern how it behaves, such as electrical limits, thermal behaviour or material flow.

  • Asset hierarchy and layout
  • Connections and dependencies
  • Operating limits and constraints
  • Corrections to the as-built records

Connect

Bring in live data from the systems you already run, and add sensors only where something important is not being measured.

  • Protocol and API integration
  • Point mapping and naming
  • Added sensors and gateways
  • Data quality checks

Validate

Compare what the model predicts with what the asset actually does, and keep adjusting until the two agree closely enough to trust.

  • Predicted against measured values
  • Tolerance agreed with your team
  • Known limits written down

Operate and extend

Hand the twin to the people who will use it, keep it accurate as the asset changes, and extend it to the next area once the first one has proved itself.

  • Training and handover
  • Change control for the model
  • Extension to further areas

What decides whether a twin is useful

Accurate records

A model built on drawings that no longer match the site will give confident wrong answers. Reconciling the records with what is physically there comes first.

Enough measurement

The twin can only follow what is measured. Coverage of the points that matter for your questions counts for more than the total number of sensors.

A question it must answer

The twins that get used were built for a specific decision. The ones built to show everything usually end up as screensavers.

Questions we get asked

How is a digital twin different from a 3D model or a dashboard?

A 3D model shows what something looks like, and a dashboard shows what is happening now. A twin links the two: a model that follows the real asset through live data, and that you can change to see what would happen. The ability to ask what-if questions is what separates it from both.

Do we have to replace our existing systems?

No. A twin sits on top of what you already run. It reads from your existing control, monitoring and records systems, and where something is not measured we recommend sensors and tell you the cost first.

How long does it take?

It depends on how much of the asset is already instrumented and how accurate your drawings and records are. For a data centre with existing instrumentation, a single site is usually live in six to ten weeks. For other domains we scope the timeline after a survey rather than quoting one in advance.

Can we start small?

Yes, and we usually suggest it. Pick one asset class or one area, such as one hall, one line or one feeder, prove that the model matches reality, then extend it.

Where does the data live?

That is a decision to make before choosing a platform. Our default is to run inside your own environment so operational data stays within your perimeter, and a private cloud deployment is possible where your policy allows it.

Tell us what you want to test

Describe the asset and the decision you are trying to make. We reply with what a twin would need to model, which data it would use, and whether it is worth building.