Graphs on the Ground Part III: Knowledge Graphs in Manufacturing

For hundreds of years, a single craftsman working in isolation could produce a product – a cobbler made shoes, a smith made tools, and a cartwright made wagons. Then industrialization happened and the factory sprang into being. By organizing the labor of many people around an assembly line, factories produced products faster than ever. At the same time, by componentizing the process, they could create more complex goods. Managing output in this new paradigm required different techniques than a simple cobbler in his shop, however.

Just as the shift from artisanal to industrial production required new approaches, so too does the shift from traditional to modern manufacturing. The factory of today is vastly more complex than the factories of decades past. Thanks to internet-of-things (IoT) enabled machinery, the globalization of supply lines, and the proliferation of technical standards, 21st century manufacturing requires 21st century techniques. Knowledge graphs are one such modern tool with broad application within manufacturing. This article will explore several of the key use cases for knowledge graphs within this sector.

Predictive Maintenance

Modern production lines consist of high-tech machines that contain hundreds, if not thousands of sensors and perform operations that jointly produce a specific product. Each of these machines is made up of components that wear out over time. When one piece breaks, the whole operation grinds to a halt. As a result, keeping conveyor belts moving requires careful attention not only to how the different machines interact with one another, but also to how the components of each machine interact internally. Knowledge graphs can help with both.

A knowledge graph can be used as a digital twin of a machine or even of an entire production facility. Using a graph, organizations can model each component of a machine, its parameters, relationships to other components, even alternative parts that could replace it. Each piece becomes a node in the graph with a semantically defined relationship to the others. Running an analysis of this digital model allows companies to identify what could go wrong and proactively intervene. (See figure 1.)

Figure 1. Vastly Simplified Model of a Machine as a Knowledge Graph

In the same way we can model the attributes of and relationships between machine components, we can model the interaction and properties of stages in production processes. Real time data from IoT sensors can facilitate constant monitoring of these systems and help forecast issues down the line before they happen. Because knowledge graphs reside in a graph database, they typically aren’t optimized to store this IoT data directly. Instead, the metadata for the sensors lives in the graph, which uses it to virtualize the IoT databases that hold the time series data.

Building Automation

A similar use case to predictive maintenance comes in the form of building automation. The advent of smart factories has turned entire buildings into something akin to a giant machine. Organizations can monitor and operate systems including air conditioning, fire safety, and physical security remotely. Using a knowledge graph, they can semantically define the relationships between these systems, even the physical spaces in which they’re located.

In fact, the popularity of this approach has resulted in the creation of an open-source ontology called Brick. Brick helps companies jumpstart the process of modeling their buildings in a knowledge graph by providing a dictionary of common terms and concepts along with pre-defined relationships that all use the same standards as RDF graph databases, like Ontotext’s GraphDB.

Figure 2. Example of Model Using Brick

These digital building models are especially useful for understanding power consumption because they show how different draws on power relate to one another. As smart energy grids begin to gain traction in the next two decades, graphs like these will be vital to understanding how facilities’ power use balances across the larger energy network.

Supply Chain Management

Manufacturing knowledge graphs don’t just have to model physical objects, however. They can also provide insight into more abstract concepts like supply chains. As we’ve all become acutely aware, thanks to the pandemic-related disruptions of the past two years, modern supply chains involve numerous parties, often scattered across multiple countries, if not continents. Storing information about the origins of product components in a knowledge graph allows for better analysis of these intricate relationships.

Subcontractors and the parts they contribute can become nodes in a supply chain graph. Then organizations can link additional information about their subcontractors to those nodes. This might include geographic data, news coverage, even health data. Then, when there’s a Covid outbreak, a canal blockage, or a human rights violation, companies know exactly which aspects of their supply chain will be impacted.

Tracking Standards

Finally, knowledge graphs can be useful for tracking ever-evolving sets of standards governing industrial activity. These standards cover everything from material strength to types of machinery to safety protocols. Since standards vary geographically and by manufacturing sector, often multiple sets of standards apply to the same facilities. Some standards overlap, some define the same concepts in different terms, some are voluntary, and some are legally mandated. Keeping up with the relationships between these standards can be a headache, especially when they’re routinely updated.

Using a knowledge graph, companies can relate different standards to one another and even break them down into semantically linked requirements. They can store both historic and current versions of the same standards, to track how they evolve over time. Once the information is in knowledge graph form, companies can query across the standards to identify how each interrelates, providing a clearer picture of the standards landscape and making it easier to stay in compliance.

Conclusion

Knowledge graphs can be a potent tool for modern manufacturing operations. As products, processes, and facilities have become more complicated, knowledge graphs provide a way to manage the complexity. They can represent both physical and abstract systems and allow companies to digitally explore and identify dependencies. Ultimately, building and analyzing knowledge graphs helps companies diagnose and remediate potential problems as or even before they occur in order to keep production lines running smoothly.

Do you want to learn how Ontotext’s knowledge graph-based technology can help in your particular use case?