A team of researchers from Mannheim expands the fundamental concept of digital transformation
How do I turn my physical product into a smart product of the future? This is the question many producers are forced to ask themselves when they think about the complex challenge of digitalizing their businesses. What’s happening in manufacturing is only symbolic for what’s happening in almost all areas of industry. A common approach now being adopted is not to make “smartness” a directly embedded element of physical objects, but to create digital twins to which intelligence can be outsourced. The CeMOS research center at Mannheim University of Applied Sciences has been working with the Steinbeis Transfer Center for Smart Industrial Solutions on ways to move digital twins on from their purely passive roles and place them into the roles of autonomous entities. To do this, the team has developed a prototype infrastructure.
Digital twins are a core concept of digital transformation in business and society in general. Based on the foundation of technology laid by the internet of things, they provide digital representations of a wide variety of objects in the physical world. The term “representation” has a broad definition, such that a digital twin could merely be a representation of data and the states of physical objects, like an “example of an object” in the main memory of a computer – or it could constitute a visual depiction by using corresponding virtual technology, augmented reality technology, or complex simulation models.
Every change of state in the physical object is reflected in the digital twin, but in parallel to this changes of state are reproduced in the digital twin and the physical object. This makes it possible to set up complex, cyberphysical systems with digital representations that cannot be meaningfully divorced from physical systems.
The advantage of designing such systems is that their functionality can be extended. Whereas it can take a great deal of time and effort to modify or expand machines or process-based equipment once they are set up, with some functions – such as analyzing sensor data, calibrating settings to match the context, or applying machine learning methods – there is the flexibility of adding them to digital twins. This makes it simple and straightforward to add smart features to physical devices in the digital space, without having to adapt the physical objects. Communication between physical devices and their digital twins is established and maintained via so-called digital threads.
The goal: autonomous digital twins
Until now, however, beyond the de facto standards of the internet of things, there is no real standardization in this area and this is currently hindered by the high levels of heterogeneity of physical objects, as well as by requirements affecting individual use cases when it comes to equipment timing, applied protocols, and data exchange formats. Also, digital twins are still currently perceived as playing a more passive role, driven by external events; indeed, in tangible terms they do not actually play any role in the field of autonomous, active, agent-like systems. This significantly limits the leverageable potential of digital twins with respect to value creation and productivity.
As part of the TWINEvent project, researchers at the Center for Mass Spectrometry and Optical Spectroscopy (CeMOS) at Mannheim University of Applied Sciences are developing an overall architecture that, in addition to providing standardized on-demand digital twins (digital twins as a service, or DTaaS), also provides a distributed and scalable runtime environment – i.e. a kind of habitat for digital twins.
ProxiCube: physical objects create their own digital twins
For a number of years, Proxivision, the Bensheim-based optoelectronics specialist, has been working on a project basis with the CeMOS team and experts at the Steinbeis Transfer Center for Smart Manufacturing Solutions. For their current project – ProxiCube, a kind of measurement cube – the experts have been using the overall architecture developed by CeMOS for advanced control and data processing. The cube measures all kinds of environmental data, assessing not just particulate matter – a particular highlight – but also liquid aerosols using a separate measuring channel.
This is especially important given the current pandemic. The large number of sensors that need to be delivered and rapid advancements in knowledge are making it important to gather extremely high volumes of location-specific data and carry out frequent software updates. These are precisely the kind of processes that can be automated with digital twins – remotely and extremely efficiently, without coming into physical contact with the measuring cubes.
The first step for the project team was to design a basic service that allows practically any physical object to generate a digital twin of itself based on its description in a standardized language. The only prerequisite for this is that the physical object is capable of communicating directly or indirectly with the DTaaS system, for example using gateway architecture or proxies. The standardized language for the description is not only used to describe the functions, capabilities, and properties of the physical object, it also makes it possible to place devices into semantic categories. This allows objects to be placed in their respective and currently valid context. The specific placement of digital twins is transparent and freely configurable. As a result, the entire infrastructure can be operated in the cloud, although it is also possible to set up on-premise solutions, operate systems on edge/fog servers, or even set up hybrid operations as required by the use case.
The basic service is responsible for providing a standardized programming interface for applications, thus automatically generating an abstraction layer between applications and the actual physical devices. In this way, an application developer can give one or more digital twins more advanced functions and capabilities. These can be parameter studies, simulations, or even optimization procedures, to name just a few, and the results and effects are – by definition – transferred to the physical object. The basic service thus automatically provides a generic instance of a digital twin, enriched with semantic properties, also facilitating bidirectional data exchange between the digital twin and the physical device, plus the automatic generation of a programming interface based on open standard protocols. This makes it possible for application developers to introduce smart capabilities to the twin (and thus to the device itself).
Autonomous communication between multiple digital twins
Beyond the basic service, the project team is working to develop an extension to the runtime environment that will allow the digital twins to operate autonomously and independently – and pursue predefined goals. To do this, methods used with multi-agent systems are being investigated and implemented as appropriate. The team believes this will enable complex processes to be mapped between multiple digital twins without the flow of control having to be managed by a central application. Especially in combination with semantic context classification, such a solution will enable communication and process mapping for digital twins that do not yet know each other. Here, too, there is huge application potential: Digital twins of workpieces can work independently to obtain offers for processing at different processing centers, or products already in distribution can try to activate the right levers to be delivered to the customer as quickly as possible within a specified budget. When digital twins solve problems, they are automatically solved in the physical world as well, and necessary measures can be introduced automatically.
The projects conducted by the researchers in Mannheim demonstrate the possibilities offered by the system in “smart anything” scenarios: Whether it is smart production, smart cities, or smart transportation, with the help of autonomous digital twins, distributed, scalable use cases can be automated on demand based on simple principles. The new system therefore provides an elementary building block for digital transformation in business and society as a whole, also contributing to the rapid and demand-oriented adaptation and implementation of complex technology.