Episode Details

UNLOCKING FACTORY AI: WHY YOUR PLANT NEEDS PRODUCT-PROCESS-RESOURCE (PPR) KNOWLEDGE GRAPHSFactory artificial intelligence is often capable of reading massive amounts of data, yet it frequently lacks a fundamental understanding of the actual plant environment. Enterprise Resource Planning (ERP) systems know the orders, Manufacturing Execution Systems (MES) know the operations, and historians know the signals. However, none of these isolated records automatically explain how a product, process, and resource (PPR) came together at any given moment. This missing connection is precisely where promising artificial intelligence initiatives lose their footing. By implementing a PPR knowledge graph, manufacturing teams can transition from guessing based on raw values to reconstructing real operating situations.

To build an effective foundation for industrial artificial intelligence and digital transformation, plant engineers and data architects must focus on several core structural components:

• Moving from Columns to Relationships: Storing values is straightforward, but data becomes truly useful only when a model can clearly state what it describes, which source supplied it, and which operating situation it belongs to.
  
• Separating Plan from Actuals: Strict separation between planned routing definitions and actual execution records prevents models from incorrectly explaining yesterday's production issues using today's routing tables.
  
• Leveraging Industry Standards: Frameworks like ISO 95 provide vital terminology and backbones, while standards such as OPC-UA and Asset Administration Shells (AAS) help describe device-facing data and individual assets across the plant.
  
• Extending the Basic Triad: Modern research literature expands the standard product-process-resource triad to include undesired conditions, plausible causes, and mitigation knowledge to account for real-world failures and rework.
  

Integrating these semantic relationships enables powerful industrial use cases across the manufacturing lifecycle:

• Root Cause Analysis: Trace quality deviations backward and forward from the end of a production line through actual operations, specific tools, materials, and resource states.
  
• Context-Aware Prediction: Equip predictive models with proper operating context—such as product variants, recipes, and tool configurations—so identical sensor signals are interpreted accurately across different runs

Become a supporter of this podcast: https://www.spreaker.com/podcast/m365-fm-a-microsoft-mvp-podcast-by-mirko-peters--6704921/support.