From data containers to semantic models – why digital twins need a foundation
Digital twins are widely regarded as a key enabler of connected industry. Yet many companies still treat them as little more than digital replicas of production data. Johannes Eckstein and Dr. Oliver Braun, Managing Directors of NuCOS GmbH in Stuttgart, argue that this view falls short. In their perspective, digital twins only unlock real value when they rest on a semantic foundation and are combined with machine intelligence. Only then can data be transformed into knowledge and digital representations into learning systems. Rather than serving as static mirrors of reality, digital twins should support understanding, decision-making and continuous improvement. Introducing them “for their own sake” is pointless, Eckstein and Braun emphasize. Their full potential emerges only when data, models and people interact based on a shared semantic understanding.
“The ideal digital twin understands the real-world context of a machine, material or process,” explains Johannes Eckstein. This requires terminology, parameters and states to be clearly defined and semantically linked. A requirement-optimized data model goes far beyond a structured database: it formalizes the language of the technical domain and makes explicit the knowledge that previously resided in documents, spreadsheets or individual experience.
Dr. Oliver Braun adds: “We see digital twins as dynamic systems that grow with the requirements of their environment. Without a semantic data model, they remain static representations. With such a model, they become an enabler of interoperability, automation and quality assurance.”
Machine Intelligence: A Tool, Not a Substitute
Johannes Eckstein deliberately avoids the term “artificial intelligence.” In his view, it is misleading. “There is nothing artificial about it,” he explains. “What we are dealing with is machine intelligence – trained and shaped by humans.”
Machine intelligence is derived from human expertise translated into models capable of recognizing patterns and formulating hypotheses. In the context of digital twins, it supports the analysis of production data, the early detection of anomalies and the optimization of process parameters. However, the prerequisites remain unchanged: high data quality and a clear understanding of context.
“Machine intelligence alone does not solve problems,” Eckstein emphasizes. “It only becomes effective when it operates on semantically clean and comprehensible data. That is when it turns into a partner rather than a black box.”
From Theory to Industrial Practice
NuCOS supports companies in implementing digital twins through modular, interoperable solutions rather than monolithic platforms. With software systems such as AddiPlan, AddiBase and AddiMap, the company covers a wide range of applications – from process planning and quality assurance to traceability and data-driven business models – all built on a shared semantic foundation.
Many clients begin with clearly defined, manageable use cases. Thanks to the structured data basis, their digital twins can evolve step by step. This approach lowers entry barriers and helps companies avoid technological lock-in.
User-Centricity as a Productivity Factor
User-centric design is another core principle of NuCOS. “Technology only creates value if it is actually used,” says Eckstein. In industrial environments, many technically sound systems fail because they are too complex or because users are not sufficiently involved.
For this reason, NuCOS integrates users early in development processes through workshops, prototypes and iterative testing. A well-designed user interface not only increases acceptance but also improves data quality – and clean data is the lifeblood of any digital twin.
Connected Twins and Data Spaces
Looking ahead, Eckstein and Braun expect digital twins to evolve into connected, cross-domain systems. These will link production facilities, laboratories, simulations and business processes via shared data spaces based on open standards and cloud-edge architectures.
For small and medium-sized enterprises in particular, this development represents an opportunity rather than a threat. “The key is to start with a solid semantic foundation,” Eckstein explains. “If you model your data properly today, you can scale tomorrow – without being tied to a specific platform or provider.”
Digital twins are more than digital replicas. They offer the opportunity to understand machines, processes and products through a common, machine-readable language. Semantic data models, machine intelligence and user-centric software form the foundation that makes this possible.
“START SMALL, BUT GET IT RIGHT”
An interview with Johannes Eckstein and Dr. Oliver Braun, who discuss data models, AI and the future of digital twins
Mr. Eckstein, Dr. Braun, what would you say makes an ideal digital twin – and what role do requirement-optimized data models play?
Johannes Eckstein:
Rather than just replicating data, the ideal digital twin is able to understand contexts. In other words, it must be able to semantically capture the real-world context of a machine, material or process. Data can only be properly interpreted, compared and reused in new applications if the terminology, parameters and conditions are clearly defined and linked to each other.
Dr. Oliver Braun:
This is exactly where requirement-optimized data models come in. We draw a distinction between simple information containers and semantic models that explicitly replicate technical requirements. These semantic models ensure that, rather than remaining static, a digital twin is able to grow along with the relevant development, production or quality control requirements. We see this as the key to enabling interoperability and automation.
What are the specific benefits of using AI in digital twins?
Johannes Eckstein:
Artificial intelligence only comes into its own when it has access to consistent, comprehensible data. In a digital twin, AI can detect anomalies, optimize process parameters and support decision-making processes. But it can only do so if the data is structured, contextualized and comprehensible.
Dr. Oliver Braun:
One example of how we use AI is to analyze process data from additive manufacturing systems in order to develop quality indicators. As well as faster detection of the causes of anomalies, this also enables predictive optimization of future component batches. It elevates AI from a mere analysis tool to an integral part of the digital twin – a learning, explainable system.
What services do you provide to help your clients implement digital twins in practice?
Dr. Oliver Braun:
Our clients often come to us with very specific requirements covering everything from manufacturing to research and product development We help them build the semantic data foundation, integrate existing systems via APIs and create interfaces with MES, ERP or laboratory environments.
Johannes Eckstein:
We also use these models to develop practical applications, for example for process planning, traceability, quality control or the data economy. One of our priorities is to ensure that the client isn’t tied to a monolithic system – they should be able to build and expand their system gradually. That’s why we use modular software like AddiPlan, AddiBase and AddiMap, which can be combined to create an open, interoperable ecosystem.
Why are user-friendliness and a positive user experience so important in your work?
Johannes Eckstein:
Because technology only adds value if it’s actually used. In industry, there are any number of good systems that nevertheless fail because they’re too complex to operate or because the users aren’t on board. For us, usability isn’t just a nice-to-have, it’s a productivity factor.
Dr. Oliver Braun:
We integrate user-centricity from the very beginning. The aim is to allow engineers, production workers or quality managers to focus on their work rather than on the tool. A good user interface lowers the entry barriers and improves data quality, both of which are key to digital twins’ success.
Which technological trends do you think will shape the future of digital twins? And how can small and medium-sized enterprises (SMEs) in particular keep up with these developments?
Dr. Oliver Braun:
We’re seeing a distinct trend towards connected, cross-domain digital twins that replicate entire value creation processes rather than just individual machines or products. This calls for standardized interfaces, cloud-edge solutions and interoperable data spaces.
Johannes Eckstein:
For SMEs, it’s not a question of having to implement everything at once. The key is to start with clearly defined, data-driven use cases such as traceability, process monitoring or AI-driven quality control. If the data models are built right, you can then scale up gradually. And that’s exactly how we approach things: start small, but get it right – with a semantic foundation that allows you to grow.
Contact
Dr. Oliver Braun (author, interviewee)
Steinbeis Entrepreneur
Steinbeis Consulting Center NuCOS (Stuttgart)
www.nucos.de
Managing Director
NuCOS GmbH (Stuttgart)
www.nucos.de
Johannes Eckstein (author, interviewee)
Steinbeis Entrepreneur
Steinbeis Consulting Center NuCOS (Stuttgart)
www.nucos.de
Managing Director
NuCOS GmbH (Stuttgart)
www.nucos.de