A functional digital twin is built from your diagrams and your documentation, with no sensors and no IoT project. Predictive maintenance, by contrast, relies on sensors fitted to the machine that take continuous measurements to anticipate wear. Both serve reliability, by two different routes.
The market often pairs the digital twin with sensors, as though one could not exist without the other. Platforms such as Mimorian show the opposite. Mimorian models equipment, structures failure diagnosis and captures the know-how of maintenance teams through a multi-agent AI architecture, working from the diagrams and manuals already in place. This article compares the two approaches, explains what can be done without sensors, and shows why they complement each other.
What is the difference between a functional digital twin and predictive maintenance?
The difference lies in the type of reasoning. Sensor-based predictive maintenance reads signals, vibration, temperature and current, and looks within them for the signs of an approaching failure. It is a statistical reading of measurements. The functional digital twin reasons about the structure of the machine: it knows the components, their links and their failure modes, so it works back from a symptom towards its possible causes. It is causal reasoning.
This distinction changes the use case. Sensors monitor a condition and raise an alert when it drifts. The functional twin guides a diagnosis once the failure is present, and keeps a record of the reasoning for next time. Standards frameworks, for their part, separate these two worlds: condition monitoring falls under ISO 17359, the digital twin for manufacturing under ISO 23247.
Can you build a digital twin without sensors?
Yes. A functional digital twin does not need sensors to exist. Its raw material is the electrical, pneumatic and hydraulic diagrams, the machine documentation and the history of interventions. The AI breaks down these diagrams and rebuilds the network of components, without fitting anything to the machine.
This point changes the start-up. No sensors to fit, no network to run, no IoT project to launch before the first result. The pilot runs in a closed loop, alongside the information system: an export is enough, the IT connection comes later, at your own pace. The first value lands within a few weeks, on real failures, because the modelling is done in advance by the vendor.
Why does sensor-based predictive maintenance deliver on its promises only with caution?
Sensor-based predictive maintenance has genuine value on critical rotating equipment, where a drift in vibration or temperature signals an imminent breakdown. The gain is real but measured: among the manufacturers that have deployed advanced predictive maintenance, 60% report better availability, for an average gain of at least 9% (PwC and Mainnovation, 2018). That result is paid for with a sensor and data project that is often long to install.
A more discreet limitation lies in the data. Good maintenance teams stop their machines before they break, and that is the mark of a mature team. Their history therefore contains few terminal failures, precisely the ones a statistical model would seek to learn from. The better the team, the less material the predictive model has. The functional twin works around this limitation by reasoning about the structure of the machine, not about the statistical history of failures.
| Criterion | Functional digital twin | Sensor-based predictive maintenance |
|---|---|---|
| Raw material | Diagrams, documentation, intervention histories | Fitted sensors, continuous measurements |
| Type of reasoning | Causal, on the structure of the machine | Statistical, on the measured signals |
| What it brings | Guided diagnosis and a memory of failures | An alert on a condition drift |
| Prerequisites to start | Your existing diagrams and manuals | Sensor fitting, network, data project |
| Time before the first result | A few weeks | The time needed to instrument and collect |
Do you have to choose between the two?
No, the two complement each other. Sensors keep their full place for monitoring the condition of critical rotating equipment. The functional twin brings the diagnostic reasoning and the memory of failures, across the whole fleet, without waiting for instrumentation. A sensor alert signals that a drift is coming, the functional twin then helps to understand why and what to check.
The simplest approach is often to start with the functional twin, because it begins with what you already have, then to add sensors where condition monitoring justifies it. The twin also complements the CMMS, which plans and records interventions, by adding the structure and the diagnosis.
Frequently asked questions
Does the functional digital twin predict failures?
No, and that is a deliberate choice. It does not guess at a failure within signals, it breaks it down from the structure of the machine. It works back from an observed symptom towards its possible causes, in order of probability, then keeps a record of the reasoning for next time. Sensor-based prediction remains the role of condition monitoring, which plugs in as a complement.
How many sensors do you need to start a functional twin?
None. The functional twin is built on your diagrams and your documentation. Sensors come later if needed, on the equipment where condition monitoring brings a real gain, without ever being a prerequisite to starting.
Conclusion
Three key points to remember.
- The functional twin reasons, predictive maintenance anticipates: one follows the structure of the machine to understand a failure, the other reads signals to spot a drift.
- You start without sensors: diagrams and documentation are enough, the first value lands within a few weeks, with no prior IoT project.
- The two complement each other: sensors for condition monitoring, the functional twin for diagnosis and the memory of failures, alongside the CMMS.
Next step for a maintenance manager: model two or three pieces of equipment from their diagrams, then decide, sensor by sensor, where condition monitoring genuinely adds value.
For the full picture, read our guide to the functional digital twin for maintenance, and on how it works alongside sensors, our page on sensors, IoT or SCADA compared with Mimorian.
📚 Sources :