3. • Introduction and manufacturing context
• Methods for extracting insights from the graph
• Use case evolution
• Platform maturity
• Questions
Outline
5. What caused a failure?
• Vertically integrated
• Batch processing
• Multiple teams
• Nonstandard analysis
methods
• Multiple data sources
Business Problem & Use Case
+
11. Date
Score
• Every batch (node) gets a “score”
• Scores can be analyzed in a
number of ways
Extract and Analyze Graph Data
Score
Process Data
Product
Qty: 100
Part A
Part A
Failure
12. Adding new nodes/relationships
• Products
• Processes
• In process testing
• Clinical feedback
– Initially built from single database
What next? Evolving the data model
13. Connecting Products
• Apply findings from existing products to new ones
• Alert other users of suspicious batches/materials more
quickly
• Improve sensitivity to weak signals
14. • Inline manufacturing feedback
– Sometimes, upstream testing can be a critical input to finished device
yield
– Predictions can lead to prescriptions
Expanding arenas
Input
param:
70%
50%
Yield
Input
param:
60%
95%
90%
Yield
60%
97%
95%
Yield
70%
Random Batch Matching