Digital Twin

What is a digital twin?

A digital twin is a virtual replica, or digital repre­sen­ta­tion, of a phys­ical object or process that is updated with data from the real world. It combines three elements: the phys­ical asset, its digital model and the data connec­tion between them. The model can be as simple as a set of struc­tured prop­er­ties or as complex as a physics-based simu­la­tion. What makes it a digital twin is that it reflects the current state of the real asset and changes with it.

Researchers often sepa­rate three levels. A digital model has no auto­matic data exchange with the real object. A digital shadow receives data from the asset in one direc­tion. A digital twin exchanges data in both direc­tions, so insights from the digital side can flow back and influ­ence the phys­ical one, for example by adjusting machine para­me­ters. In everyday indus­trial language, all three are often called digital twins, which explains why defi­n­i­tions vary so much.

The concept of digital twins is not new. NASA used phys­ical and later digital copies of space­craft to analyze prob­lems remotely, and the term digital twin became popular in manu­fac­turing in the 2010s together with Industry 4.0. What has changed is the cost of sensors, connec­tivity and computing, which makes digital twin tech­nolo­gies prac­tical for every manu­fac­turing facility, not only for aero­space projects, and a core building block of digital trans­for­ma­tion.

How digital twins work in manu­fac­turing

To create a digital twin, a manu­fac­turer collects data from the systems that already exist in the manu­fac­turing envi­ron­ment and orga­nizes it around the asset. Sensors deliver measure­ments such as temper­a­ture, vibra­tion or energy use. Controllers and SCADA systems provide machine states. MES and ERP systems contribute orders, batches and main­te­nance records. Engi­neering tools add CAD models, bills of mate­rial and docu­men­ta­tion.

This data needs a common struc­ture and a reli­able trans­port. On the shop floor, OPC UA and MQTT move live data. For the content of the twin, the Asset Admin­is­tra­tion Shell provides a stan­dard­ized format, so that the digital twin of a compo­nent from one supplier can be read by the soft­ware of another. Processing happens at the edge, close to the machine, or in the cloud, depending on latency and data volume. Analytics, simu­la­tion and machine learning then turn digital twin data into answers, and digital twins can predict what is likely to happen next: Is the bearing wearing out? Will the new product variant fit into the line? How much energy did this work­piece consume?

What are the four types of digital twins?

A common clas­si­fi­ca­tion sorts digital twins by the scope of what they repre­sent:

The levels build on each other. A process twin is only as good as the data from the system, asset and compo­nent twins below it, which is why stan­dard­ized data at the compo­nent level matters so much.

Appli­ca­tions of digital twins in manu­fac­turing

Digital twin tech­nology in manu­fac­turing covers the whole product life­cycle and every step of the manu­fac­turing process. These are the digital twin use cases used in manu­fac­turing today:

• Product devel­op­ment and virtual commis­sioning: engi­neers test a machine or line in the digital twin before it is built, which reduces commis­sioning time on site.
• Predic­tive main­te­nance: sensor data from the twin reveals wear before a failure occurs. See condi­tion moni­toring for the under­lying methods.
• Quality and process opti­miza­tion: the digital twin links process para­me­ters with quality results and shows which settings produce scrap.
• Energy and sustain­able manu­fac­turing: in the OI4 Energy Moni­toring project, members track energy and compressed air consump­tion per work­piece on two pilot machines, as a basis for CO₂ foot­prints and energy-aware pricing.
• Digital Product Pass­port: the twin of a product becomes the source for its Digital Product Pass­port, including mate­rials and carbon foot­print.
• Training: digital twins for training let oper­a­tors prac­tice changeovers and fault handling on a virtual line before they work on the real one.
• Commis­sioning and service: tech­ni­cians scan a QR code and see the docu­men­ta­tion, spare parts and history of the exact machine in front of them.

What are the main bene­fits of digital twins in manu­fac­turing?

The bene­fits of using digital twins depend on the use case, but they fall into a few groups. Digital twins help reduce unplanned down­time because prob­lems become visible early. Digital twins enable shorter engi­neering and commis­sioning because errors are found in the model, not on the shop floor. A digital twin provides one consis­tent view of an asset for everyone who works with it. They improve quality because process data and results are connected. And they create trans­parency across the supply chain, which regu­la­tion increas­ingly demands.

The economic value is harder to put into numbers than the tech­nology. The OI4 project Exploring the Economic Value of Digital Twins addresses exactly this gap. Its members built a library of economic poten­tials for inter­op­er­able, AAS-based digital twins, applied it to two use cases and devel­oped a calcu­la­tion method with KPIs that compa­nies can use in their own context.

Why inter­op­er­ability decides whether a digital twin scales

Many a digital twin solu­tion works well inside one vendor’s ecosystem and fails at its border. The inte­gra­tion of digital twins across compa­nies is where most projects get stuck. A machine builder who receives compo­nents from dozens of suppliers cannot main­tain a different inte­gra­tion for each of them. The same applies to oper­a­tors running machines from different builders. Without a shared struc­ture, every digital twin becomes an island, and the effort grows with every new partner. This is a classic case of vendor lock-in.

The Asset Admin­is­tra­tion Shell solves this by stan­dard­izing how a digital twin is struc­tured and accessed. At the “Follow the Twin” chal­lenge at Hannover Messe 2024, OI4 members provided product data in AAS format through two inde­pen­dent repos­i­to­ries, and visi­tors could retrieve it consis­tently, regard­less of vendor or backend. Members of the OI4 Compo­nent Supplier work­group also built Twin­sphere, an AAS-based plat­form that gives medium-sized manu­fac­turers access to digital twins, product carbon foot­prints and VDI 2770 docu­men­ta­tion.

The OI4 Digital Twin work­group works on the inte­gra­tion and manage­ment of inter­op­er­able digital twins, and since 2023 a liaison agree­ment connects OI4 with the Digital Twin Consor­tium to align tech­nology compo­nents across both commu­ni­ties.

How to imple­ment a digital twin in manu­fac­turing

1. Start with a ques­tion, not a plat­form. Choose a problem with a measur­able result, for example unplanned stops on one machine or manual data entry at goods-in.
2. Use the data you already have. Most facto­ries have more data than they use. Connect existing controllers and systems before adding new sensors.
3. Agree on a struc­ture early. Build your digital twins on the AAS and its submodel templates, so that the next use case and the next partner can reuse the data.
4. Plan for brown­field. Older machines often lack stan­dard inter­faces. Retrofit sensors and edge devices bridge the gap.
5. Scale step by step. Move from one asset to a line, then to the plant. Digital twin devel­op­ment and digital twin modeling with simu­la­tion or AI come once the data foun­da­tion is reli­able.

Chal­lenges when imple­menting digital twins

The OI4 energy moni­toring project lists the obsta­cles that many manu­fac­turing compa­nies face: machines from different vendors and gener­a­tions, missing stan­dard inter­faces on legacy equip­ment, limited protocol support, inte­gra­tion with existing IT systems, incom­pat­ible networks and the effort of retro­fitting sensors in a running produc­tion. Secu­rity is another factor. A digital twin that can change machine para­me­ters needs the same protec­tion as the machine itself, which links digital twins to OT secu­rity. Finally, a digital twin needs owners. Without clear respon­si­bility for data quality, the twin drifts away from reality and loses its value.

The future of digital twins

Three devel­op­ments shape the future of smart manu­fac­turing with digital twins and digital twin-based services. First, regu­la­tion: the Digital Product Pass­port and the EU Data Act require struc­tured, share­able product data, which turns digital twins from a nice-to-have into infra­struc­ture. Second, data spaces such as Manufacturing‑X let digital twins cross company borders under clear usage rules. Third, AI: agents that monitor a plant, order mate­rial or schedule main­te­nance need struc­tured, reli­able data to act on, and a stan­dard­ized digital twin is exactly that.

Frequently asked ques­tions