This is a guest blogpost by Jim Webber, Chief Scientist at graph database provider Neo4j. It discusses Knowledge Graph-based digital twin technology for better supply chain management.
What links predictive maintenance at a leading manufacturer of construction and mining equipment, car manufacturing, and the efficient running of intelligent power plants?
They’re all examples of highly complex, interconnected systems where disruption is expensive and potentially hazardous. But what also connects them is the real-world use of digital twin technology. Digital twins are computer models of real business work, which help asset owners understand the details of what is happening, in real-time as well as retrospectively.
Modern digital twins are built atop knowledge graphs, a platform which can not only scale to the vast amounts of data accrued by assets and people, but also deal with the intricate structures and relationships between them. In turn, this deepens your visibility into key business processes.
For example, after Caterpillar digitised and ingested 27 million company manuals and documents into its knowledge graph, it was able to upgrade to a machine learning-powered predictive maintenance system as standard. The business benefits were tangible: reduced downtime, increased productivity, and reduced cost.
In a similar way, a major Japanese automaker, struggling with quality at the assembly line, built a knowledge graph that delivers metadata at scale and speed through which its factory teams get comprehensive lifecycle data support. The result is a welcome 5% increase in production efficiency and a 20% reduction in the time needed to get these products to the showroom.
In a third example, Tata Consultancy Services’ TCS IP2 SaaS service for power plant management has been enhanced by adoption of a digital twin service powered by a knowledge graph. The technology helps customers achieve a 9% reduction in emissions, lower fuel use, and $6m annual operational savings.
The reason behind the success of modern digital twins is the convergence of analytics, data science, machine learning, and AI (Artificial Intelligence). Knowledge graphs are being deployed because they make data smarter, providing a superior underlay for those techniques than other ways of organising data.
Millions of connected data elements that require near-real-time connectivity
A knowledge graph is an interconnected dataset enriched with business semantics. A knowledge graph allows you to reason about the underlying data, and use that data for complex decision-making at scale. That’s not something traditional data management systems (like relational databases) can offer.
That’s because when data gets ingested into a graph database, relationships are stored as first-class citizens, not as some afterthought to be expensively computed on the fly. Business assets in the knowledge graph are natively connected to their neighbours, and their neighbours, and so on. As the knowledge graph grows and gets richer, it becomes more useful.
Moreover, the rich network of data in your knowledge graph becomes very useful to your data science team. The topology of the knowledge graph provides opportunities for powerful graph analytics and graph machine learning–tools which are only available to graph users.
Adding onto that graph visualisation means that users who adopt a graph-based digital twin get an immediate head-start in understanding and managing their physical business, as well as being able to respond quickly to events and look for future problems.
The success of Caterpillar and other knowledge graph users shows that if you’re a modern enterprise looking to build a digital twin, you should model it as a knowledge graph.
With millions of connected data elements that require near-real-time connectivity, the only practical solution is a graph architecture.
And as knowledge graph users find, connecting your digital twin with external data from assets, sensors, markets, and even weather forecasts opens up many use cases. Knowledge graphs make this practical, ultimately providing you with even more business value.
The author is Chief Scientist at Neo4j Neo4j