What Is a Manufacturing Knowledge Graph? Reimagining Data Infrastructure

A key component in a Manufacturing Data Hub that plays a pivotal role in reshaping the way manufacturers conceptualize their data infrastructure and MES is the Manufacturing Knowledge Graph (MKG.) But what is an MKG and what role does it play in manufacturing?

Many have hypothesized the idea of an MKG, but until recently (with the launch of the Rhize Manufacturing Data Hub), a full implementation hasn’t been available. This article seeks to demystify and set the record straight about the manufacturing knowledge graph.

To do that, we’ll cover the basic concept of the knowledge graph and how it can be applied to manufacturing data with the structure and standardization of the ISA-95 standard.

What is a Knowledge Graph?

Knowledge graphs are not a topic well known to those in manufacturing IT and MES roles.

A knowledge graph is a collection of interconnected descriptions of entities – objects, events, processes, or concepts. A knowledge graph models data as a collection of nodes (or entities) and edges (relationships between entities) in a network designed to connect and organize information that mirrors the real-world relationships and the semantics of the data.

To do that, we’ll cover the basic concept of the knowledge graph and how it can be applied to manufacturing data with the structure and standardization of the ISA-95 standard.

Knowledge Graphs & Manufacturing Data

To understand how knowledge graphs are being used in manufacturing, we must first address the current state of manufacturing data infrastructure, specifically its limitations.

The Limitations of Relational Data Structures

Defining a data structure to describe manufacturing entities like equipment, materials, and even bill of materials or recipes, and the relationships between them is not particularly difficult and has been done in relational structures for 30+ years, to varying degrees of success.

The underlying problems with relational databases arise in today’s global manufacturing environment. 

Unifying and utilizing an ever-expanding mountain of data created by machines, sensors, production lines, manufacturing sites and global supply chains is increasingly untenable using relational databases. Common limitations include:

  • Rigid data structure and relationships: Relational databases use a tabular structure where schema and relationships are predefined. As a manufacturing dataset evolves, grows and becomes more complex, those fixed structures can’t account for the introduction of new relationships. Once the schema is set, introducing new keys into the existing table structure can have undesirable effects on the structure and performance of the database.
  • Complex and slow querying: Relational databases use SQL, which can be cumbersome in complex, deep relationships. Query returns are also slow to navigate those relationships under a fixed data structure.
  • Demanding schema evolution: Modifying schema structure in a relational database is complex and time-consuming, and it could involve data migrations that disrupt operations.
  • Lack of semantic meaning: Relational databases don’t capture semantic meaning, making it challenging to infer insights, perform searches and integrate heterogeneous data without extract-transform-load (ETL) processes.

Knowledge Graph Applied to Manufacturing

The limitations of relational databases are no longer tenable in today’s manufacturing ecosystem. A knowledge graph structure, on the other hand, addresses these limitations by placing data in context by creating links and metadata, providing a structure for the integration and unification of the contained data.

For manufacturers embarking on their digital transformation or Industry 4.0 initiative, the benefits are clear.

  • Flexible data structure and relationships: The nodes and edges in the knowledge graph allow for a complete representation of the connection between manufacturing entities, making it easier to capture insights and drive improvements.
  • Efficient querying: When deploying an MKG to a GraphQL native graph database, semantic relationships allow for natural and more efficient data querying, meaning users get the information they need faster and with greater context.
  • Semantic meaning and understanding: All semantic and contextual information is captured, making it easier to enable advanced analytics and integrate heterogeneous data from sensors, databases and other sources without complex ETL processes.

It’s important to note that the major differentiation is not so much the entities as it is the descriptions of the entities and the descriptions of the relationships between them.

The key to flipping the MES paradigm of relational databases comes in changing the data structure so that it not only contains the entities and the relationships (in the form of primary and foreign keys) but contains the description of the relationship in a way that provides real context.

The Importance of the ISA-95 Standard

The ISA-95 standard was created to define the interactions between the ERP and MES layers. What it lacked (on purpose) was an implementation strategy. That has been its weak underbelly for nearly three decades. What the original creators of the standard could not possibly have known was that it would be nearly 30 years before an architecture and a set of technologies emerged that would allow for a truly successful implementation.

The standard is broad enough that it contains nearly every entity and relationship that exists within the space that it defines across substantially all manufacturing types. It is generally more difficult to find a manufacturing use that doesn’t fit than to model your use case using the standard. For the standard to achieve universal usefulness, its implementation must be extensible as well. That can only be efficiently achieved with an MKG.

The concept of an MKG provides a simplified way of transforming the standard into a schema that can both be used to model an enterprise but also to simplify the processes of extending the schema for the rare use cases that don’t fit. Because the MKG can be extended by simply adding an entity (a node) to the schema and defining the metadata that is its relationship to the existing MKG, we can achieve a complete and extensible implementation of the ISA-95 standard.

Manufacturing Knowledge Graph with Flexible Workflow and User Interfaces.

Current Data Infrastructure and MES offerings generally work to eliminate data silos and will often implement portions of the ISA-95 standard. However, most are implemented with user interfaces and business processes that are built on the opinion of the vendor, not the needs of the user. This makes customization and innovation difficult, especially for enterprise users.

At Rhize, we believe we’ve solved the problem by implementing the entire ISA-95 2018 standard into a knowledge graph, implemented in a GraphQL native Graph DB. This allows for a single standardized data structure that is equally responsive to both the ever-changing nature of your manufacturing data and the governance needs of the enterprise. By combining this with our Complex Rules Processing engine and drag-and-drop Workflow editor and engine, we’re getting the most out of MKG technology.