Manufacturing: Unlocking product traceability using an industrial knowledge graph

  • 17 August 2022
  • 0 replies
  • 264 views

Userlevel 3
  • Seasoned Practitioner
  • 16 replies

We’re @Uzair Wali and @kelvin, Senior Data Scientist and Data Science Lead in Cognite’s Manufacturing delivery team. In this post we talk about the increasingly important ability to intuitively and flexibly query data from all steps of a product life cycle and across source systems, with examples we’ve implemented on Cognite Data Fusion together with our users.

The need for traceability

In many manufacturing industries, the ability to trace a product through its manufacturing life cycle, whether internal or supply chain is extremely important. It entails the collection and management of information regarding what has been done in manufacturing processes, from the raw materials and parts used to the shipment of finished products. An industrial knowledge graph that enables this traceability has the potential to not only let users speed up or automate existing use cases, it also opens up possibilities for considerable value addition.

Typical questions

  • A customer complains about the quality of a product. What happened in my production line at that time, and which other products were affected?

  • A fault happened in the production. Which customers received bad products?

  • Historically, which factors in the production influence the final product quality, and as such can be used to predict the quality of future cycles?

Example use cases

  • Food safety and accountability, in which internal and supply chain data must be auditable

  • Precise and efficient recalls after product issues are detected or reported

  • Digital product passports, where businesses or end consumer themselves can introspect the origins and process of the product they have bought, with batteries being the leading use case in EU

  • Performing analysis on historical data to improve process and quality

Modeling a product life cycle

To enable comprehensive traceability, a knowledge graph should facilitate the modeling of products, processes and their flows as nodes and directed relationships, contextualized with industrial data types allowing complex querying across sources and IT, OT and ET data intuitively.

The example below demonstrates modeling of a fish farming life cycle, from eggs to harvest, with each stage representing a population of fish in a particular physical cage, farmed for a given period, and each directed relationship representing a movement and its metadata.

gZQJQCJI5bEWIVxGZWKWt7B0nOPme7RKxgDad2bM6o4LjJqfBOHGfHtd476D8ugFd4Hnc4tvjsM885YC4g-_gQ8BlDjO2Vz3ntKi_oAM_NJD9SDVNRex5OrZnAvuDOODjVdYoKUTGkPgI1Cw9AQT0vY

Final remarks

Companies that comply with certain regulations and standards (e.g. ISO 9001 Quality Management System) are already obligated to collect certain data for auditing purposes. However, querying over these data — often stored in siloed systems — is a manual, laborious and error-prone process. In one case, a customer using Cognite Data Fusion to represent its manufacturing data reduced the time from receiving an auditing request from the authorities to returning the response “from 5 working days to minutes”.

We’re curious to hear your thoughts and experiences with product traceability!

 

References


0 replies

Be the first to reply!

Reply