Interrogating your twin: causal reasoning in manufacturing systems
summary
Uses a causal DAG plus observational data to estimate interventions on a 23‑machine line and produce actionable policies. Tests implied conditional independencies (one slightly violated) against roughly 1,500 shifts of six‑month data, finds ~2x day‑shift failure rate, and identifies medium machines as highest cost. Workflow uses BNLearn for structure search, Granger‑style deep models for time series discovery, then linear regression (and later boosted trees) to estimate direct and total effects and produce a shift‑assignment policy; LLMs generate DAG hypotheses with a human‑in‑the‑loop verifier.
AI-written summary · about
chapters (11)
- 0:00Causal inference in manufacturing
- 3:49Correlation vs causation
- 7:44Building causal graph
- 10:12Use case: 23 machines
- 13:19Artisanal workflow explained
- 17:08Testing and refining graph
- 21:22Causal ML with fake data
- 23:24Key takeaways
- 25:49Power of writing DAG
- 26:09LLMs for causal reasoning
- 28:41Q&A
related talks
33mApplying Causal Inference in Industry 4.0: A Case Study from Glasswool Production
11mCausal Inference For Socio-Economic & Engineering Systems
10mAI-Powered Digital Twins: Patterns for Smart Manufacturing
56mWhat can be passively learned about causality?
01h 01mThe Mathematics of Causal Inference, with Reflections on Machine Learning and the Logic of Science
29m🧠🧹 Causality - Mental Hygiene for Data Science
15mStabilizing Ops & Eliminating Variability on
43mBeyond Just Prediction: Causal Thinking in Machine Learning