Digital twins and knowledge graphs: a match made in data heaven
In the world of data, graphs (and in particular knowledge graphs) are powerful tools that connect and contextualise pieces of information, creating a cohesive and insightful view into what the data is actually telling us.
These dynamic structures map out relationships between data points, transforming isolated information into a rich, interconnected web of knowledge.
But how do they connect with digital twins? And what are the benefits in doing so? In this blog, we aim to outline what graphs are, we will touch upon how a graph can be augmented with semantics and explain why the technology fits so well with digital twins.
What is a Digital Twin?
Although there are many definitions of Digital Twins, we will use the following one: A Digital Twin is a digital representation of physical entities and processes. Unlike project-based data storage, where data is only required for the duration of the project, a digital twin serves as a lasting representation of the physical system and its processes.
Over time, the physical system evolves: new assets are added, old ones are removed, and new processes may be introduced. This, of course should be reflected in its digital counterpart. On the digital side, new use-cases may arise, or new digital capabilities become possible, requiring the digital twin to be augmented with additional data, new relationships, and possibly existing digital assets to be augmented with new properties.
Therefore, it is crucial that the data representation of digital twins is extremely flexible, ensuring it can adapt and grow alongside the physical system it mirrors and the use-cases it is designed for.
What is a Graph?
A graph is a way to represent data. A graph shows how different things are connected. It consists of:
- Nodes (or vertices): These represent entities or objects. For example, in a social network, each person would be a node.
- Edges: These represent the relationships or connections between the nodes. For example, in a social network, an edge might represent a friendship between two people.
Without delving into the details of the various graph types, in general you could state that in addition to the nodes (representation of entities) and edges (representation of their relationships) properties characterising those entities and relationships are part of the graph as well.
Example: Imagine a graph consisting of nodes representing individual persons and edges representing the fact that the two persons are friends. The person node could have properties providing the name and age of the person. The “friend” relationship, indicating that the two connected persons are friends, could provide information on the beginning and possibly ending of that friendship.
What makes a Graph a Knowledge Graph?
Graphs, as described above, simply are a way to represent (raw) data. To get from raw data to information and even further up the pyramid to knowledge and wisdom, we need additional information on the meaning of the data itself. This is often overseen by people, because of the seemingly available semantic information in the graph. We, as humans often quickly understand what information is encapsulated in a graph in which the nodes have properties like “name:Jack” and “age:22” and the edges are named “friendOf”.
Yet the semantic information is actually represented in the naming of the nodes, edges and properties, and requires knowledge of the natural language and the context in which these words are used (e.g. Jack could also be the name of a dog). So obviously we need a more formalised way of adding knowledge to the graph, so that the data can be interpreted in a unique way, not only by humans, but also by computers.
Without going into details on how that is done (ontology is the word), imagine the graph to be extended with standardised (official ontology language) nodes and relationships that have a unique and well-defined meaning provided by an official standard. Those nodes and edges add information to the data nodes they are connected to.
Example: the nodes mentioned in the previous example can be connected through a standardised relationship of name “is a” to a node Person that, in turn, is connected through a standardised relationship “is a” to a standardised node of name Class. The “is a” relationship between the Class node and the Person node indicates that the Person node represents a Class (“is a” Class), or aggregation node. The “is a” relationship between the node with the name Jack and the Person node/class means that Jack “is a member of” or “is an instance of” the Person Class.
Ontologies allow for much more than demonstrated in the example above but remember that ontologies add structure (through classes and subclasses), add data schema’s and much more.
Why Graphs and Knowledge Graphs are ideal for Digital Twins
Graphs and knowledge graphs are ideally suited for digital twins due to their natural representation of highly complex systems, their extensibility, ability to add semantics and more. To demonstrate, let’s explore the characteristics of graphs and their applicability to digital twins:
Natural Representation: Graphs provide a natural and intuitive way to represent the (complex) physical domain. They are both user and machine interpretable, making it easier for non-engineers (like consultants) to understand and gain insights from the data. The formal structure of graphs also allows computers to process and interpret the data effectively.
Extensibility: Graphs are inherently open-ended, allowing new entity types, relationship types, and instances to be added without disrupting the existing system. This flexibility means that the same digital twin can be used for multiple use-cases within the same domain, creating opportunities for upsell and continuous improvement.
Interconnected Systems: Graph technology enables the interrelation of multiple federated graphs. For example, vehicle instances in a sales process graph can be linked to vehicle instances in a traffic monitoring graph. This interconnectedness allows for comprehensive analysis and insights across different domains.
Path Descriptions: Graphs naturally describe paths of connected elements, making them ideal for modelling systems like piping networks (oil, gas, water, sewer) or road networks. Depending on the representation format and query language, these paths can also be easily queried for detailed analysis.
Semantic Information: Knowledge graphs provide semantic information that enables AI-powered reasoners to infer new knowledge from the graph. This capability allows for the discovery of hidden insights and the augmentation of the graph with new information, enhancing decision-making and problem-solving.
Continuous Improvement: The ability to extend existing graphs in operating environments is a key feature of a graph-based Digital Twin Architecture. New information can be added by AI reasoners, consultants, or external services, ensuring that the digital twin evolves and improves over time. For instance, a CO2 emission service can ingest vehicle data and calculate emissions, adding this new information to the graph for all users to access.
Graphs and knowledge graphs transform isolated data points into a rich tapestry of interconnected knowledge, paving the way for advanced reasoning, decision-making, and continuous improvement in digital twin applications.
Summary
Graphs provide a versatile, dynamic, and intuitive method for representing and evolving complex, interconnected data. They are especially effective for applications like digital twins, where the data model cannot be tailored to a particular use-case but must accommodate various scenarios and adapt to growth and change over time. This leads to more predictive insights, and smarter decision-making, ultimately enhancing efficiency and fostering innovation across different industries.
For an example of how knowledge graphs work, check out our download below, and for part 2 of the blog series, From Data to Widsom, click here.
.png?h=250&iar=0&w=250&hash=EEEE8856DF5A166AB003975569C260BA)
Do you want to know moreor have a question?
Contact our experts!