Does the hybrid twin actually help?

Controlled experiments on the reference plant simulator, where ground truth is known. Every number below is read from files written by the experiment scripts; nothing is typed in by hand. These are simulation results, not industrial validation.

Prediction accuracy · held-out data

Loading…

Closed-loop experiments · 24 h on the reference plant

Loading…

Early-warning lead time · no-action runs

Simulated

Pending experiment

Run: npx tsx scripts/run-experiments.ts

Reproduce these results

Generate synthetic data (seeded)

npx tsx scripts/generate-dataset.ts

Train and evaluate the models

python ml/train.py

Check TypeScript ↔ scikit-learn parity

npx tsx scripts/parity-test.ts

Run the closed-loop experiments

npx tsx scripts/run-experiments.ts