Cycle-Time Drift Detector
Detect gradual cycle-time drift against target standards
Reference implementation. Industrial use requires validation on plant-specific data. Metrics are from open evaluation splits, not customer plants.
Intended use
Screening for cycle-time drift on repetitive lines and assets, to prioritize where standards should be re-validated. Decision support, not automatic adjustment.
Inputs & outputs
Inputs
Cycle timestamps, target standards, product and shift context
Output
Drift episodes with magnitude, onset, and affected products
Training data & evaluation
Training-data classification
Synthetic cycle-time series with injected drift, plus public benchmark segments.
Evaluation approach
Detection F1 and drift-onset lead time on a held-out synthetic split.
F1 0.88· open split
Industrial relevance
Applies across FMCG / CPG, Metals. The value is highest where the affected asset is a constraint and additional output can be monetized.
Limitations
- Needs reliable per-cycle timestamps; sparse or aggregated data reduces sensitivity.
- Standards must be reasonably current, or drift is measured against a stale baseline.