The decision is not “digital twin or no digital twin”
A legacy circular knitting factory rarely starts with a clean data architecture. Its machines may come from different production years, use different controller generations, and expose different levels of information. Some units provide a usable speed or alarm signal; others require a sensor, gateway or manual reading before a production event can be recorded.
That is why a digital-twin business case for a legacy factory should not begin with a software demonstration. It should begin with a retrofit boundary. The buyer needs to know which machines, processes and decisions will be represented, what evidence will be collected, and which existing assets can remain in service.
A new smart machine may offer better connectivity from the first day, but it also creates a larger capital commitment and may disturb proven tooling, operators and recipes. Retrofitting can preserve the installed base, yet integration effort, sensor quality and data normalization become part of the project cost. The correct choice is therefore a site-specific comparison, not a generic claim that one architecture is always cheaper.
A practical starting point is the circular knitting machine predictive-maintenance ROI guide. The retrofit decision adds one question: can the existing machine produce sufficiently trustworthy data for a useful model?
Define the twin before pricing the platform
“Digital twin” can describe several different systems. A buyer should require the supplier to state which level is included:
- Asset record: machine identity, configuration, manuals, service history and component changes.
- Condition view: operating signals such as speed, temperature, vibration, pressure, current or alarm state.
- Process model: recipe, yarn, gauge, feeder condition, production order and quality result linked to the machine.
- Diagnostic model: rules or algorithms that identify abnormal behavior and recommend an inspection.
- Simulation model: a virtual representation used to test a change before it reaches the physical machine.
A legacy project may gain useful value from the first three levels without immediately purchasing a full simulation environment. The request for quotation should separate these levels. Otherwise, a low entry price may omit the gateway work and a high price may include capabilities the factory cannot yet use.
Audit the installed base first
Before comparing retrofit cost with new equipment, build a machine register. Capture the make, model, serial number, diameter, gauge, controller, drive, feeder arrangement, sensor interfaces, communication ports, operating recipes and current maintenance records. Record whether the machine is stable enough to serve as a reference asset; a twin built on unverified data will reproduce uncertainty rather than remove it.
The audit should also identify the business decision the twin must support. Examples include:
- Detecting a developing needle-track or cam-wear condition.
- Explaining periodic fabric defects by linking them to a feeder or cylinder position.
- Comparing output and energy performance between similar machines.
- Verifying that a retrofit changed downtime without shifting the defect burden elsewhere.
- Creating a service record that a supplier and factory can interpret in the same way.
If no decision is named, the project can become a dashboard exercise. A dashboard may display data while leaving maintenance and production teams without a defined action.
Retrofit cost categories for a legacy knitting factory
A reliable estimate separates one-time engineering from recurring operation. Use a worksheet like this rather than one blended platform quotation.
| Cost block | What the buyer should request | Evidence for comparison |
|---|---|---|
| Machine survey | Site visits, controller review and signal map | Machine-by-machine scope |
| Sensors | Type, mounting, calibration and replacement plan | Datasheets and sample readings |
| Edge gateway | Protocols, buffering, security and machine capacity | Network diagram and test log |
| Data model | Naming, timestamps, recipes and asset hierarchy | Field dictionary |
| Integration | MES, CMMS, quality and user access connections | Interface list |
| Twin software | Licences, users, models and update rules | Functional specification |
| Commissioning | Baseline run, alarm testing and acceptance | Signed test protocol |
| Training | Operator, maintenance and engineering workflows | Training plan |
| Ongoing cost | Hosting, support, model updates and sensors | Renewal schedule |
The supplier should state what happens when an old controller cannot expose a required signal. The alternatives might be an external sensor, a manual confirmation step, a gateway change or exclusion of that machine from the first phase. Each alternative changes the model’s confidence and the cost of ownership.
Compare retrofit with a new smart machine on the same basis
A new machine quotation should include more than the machine price. Ask for commissioning, spare parts, training, data access, software subscriptions, network requirements, recipe migration and the cost of moving production. A new asset can reduce integration friction, but it may also introduce a new learning curve and require validation of fabric behavior.
Compare both paths over the same planning horizon using a total-cost worksheet:
Total project cost = capital equipment + integration + commissioning + training + planned downtime + recurring software and support.
For the retrofit option, add sensor replacement, gateway maintenance and the cost of keeping legacy interfaces supported. For the new-machine option, add installation, tooling, operator qualification, recipe transfer and the opportunity cost of replacing a still-serviceable machine.
Do not insert a made-up payback period. Instead, model scenarios using the factory’s own records:
- hours of avoidable downtime;
- first-grade and downgraded output;
- maintenance response time;
- energy per production unit where measurement is reliable;
- cost of a missed delivery or delayed order;
- expected useful life of the existing machine;
- cost and availability of critical parts.
The result should show a conservative, expected and upside case. If a supplier presents one guaranteed saving percentage without showing the baseline, measurement method and exclusions, ask for a revised proposal.
Data quality is the hidden retrofit cost
Legacy machines often use different names for the same event. One controller may report a stop as “fault,” another as “alarm,” and a third may not report it at all. A digital twin needs a common event dictionary, time synchronization and clear ownership of corrections.
The factory should require:
- a unique asset ID for every machine;
- a documented timestamp source and timezone;
- a versioned recipe and yarn record;
- a rule for missing or implausible signals;
- a distinction between planned and unplanned stops;
- a method for linking a fabric roll or order to the machine state;
- an audit trail for model changes;
- a retention and access policy for production data.
If the platform cannot show how a recommendation was produced, maintenance teams may ignore it. The buyer should ask for examples of false alarms, missed events and model retraining rather than only a successful demonstration. A circular knitting machine condition-monitoring checklist can help define the machine-side signals before the digital-twin supplier maps them into a platform.
Pilot design and acceptance criteria
Start with a representative group, not only the newest or cleanest machine. Include at least one asset with the controller or sensor limitation that the full project must address. Define a baseline period and keep the production recipe as stable as practical during the comparison.
A pilot acceptance document can include:
- Signal availability and timestamp completeness.
- Correct mapping of machine, feeder, recipe and production order.
- Alarm-to-work-order traceability.
- Repeatability of the same measurement on the same condition.
- Operator and maintenance response workflow.
- Handling of missing data and communication outages.
- Data export and ownership rights.
- A documented decision about whether to scale, redesign or stop.
The circular knitting machine OEE downtime coding dashboard can help define the event vocabulary before the twin is commissioned. For older assets, the vocabulary is often more important than the visual dashboard.
When replacement is the more defensible choice
Replacement may be justified when a machine cannot be made safe, stable or supportable, when critical parts are unavailable, when the controller cannot provide the minimum signals, or when retrofit engineering costs approach the cost of a new asset with a clearly documented interface.
That decision should be evidence-based. A new smart machine is not automatically a digital-twin success: the buyer still needs data ownership, service documentation, network security, model acceptance and a process for updating the virtual representation after physical changes.
Conversely, a legacy machine is not automatically unsuitable. If its mechanical condition is known, production is stable and key signals can be added without compromising safety, a controlled retrofit may provide useful operational visibility while preserving proven tooling.
FAQ
Is a digital twin the same as adding sensors?
No. Sensors provide observations. A digital twin also requires an asset model, context, data rules and a defined decision or simulation use case.
Should every machine be connected in the first phase?
Usually not. A representative pilot can expose data and integration issues before the factory commits to a fleet-wide rollout.
How should a supplier prove ROI?
Require a baseline, a measurement period, an agreed formula, exclusions and access to the underlying event and production data. Avoid accepting a saving percentage without those items.
Can a legacy controller be included?
Often it can, but the required gateway or external sensors depend on the available signals, protocol, safety constraints and machine condition. The supplier should document the limitation rather than hide it in a general compatibility statement.
What should happen after the pilot?
The factory should issue a scale, redesign or stop decision based on data quality, user adoption, operational actionability and total cost—not on dashboard appearance alone.
References
- ISO 13374 — Condition Monitoring and Diagnostics of Machines, data processing and presentation principles. https://www.iso.org/standard/73486.html
- ISO 55001 — Asset management systems and lifecycle decision context. https://www.iso.org/standard/83054.html
- ITMA — Industry 5.0 and the new textile workforce, including connected machinery and operator capability. https://itma.com/insights/blog/blog-detail/itma-2027/2026/04/08/industry-5.0-and-the-new-textile-workforce–the-future-of-textile-manufacturing
- Textile World — Smart manufacturing and automation developments in textile production. https://www.textileworld.com/textile-world/features/2026/05/building-a-smarter-textile-enterprise-with-ai-and-automation/
- NIST — Cybersecurity Framework resources for manufacturing environments. https://www.nist.gov/cyberframework/manufacturing