Definition
Correlation between AI-optimized shutdown duration compression and post-shutdown infant mortality rate.
The Measurement Gap
Major shutdowns are the most expensive maintenance events in heavy industry, and AI scheduling tools are increasingly used to compress their duration. Every hour saved in a shutdown translates directly to production revenue recovered, creating enormous pressure to accept the optimizer's compressed timeline. Legacy shutdown management metrics reinforce this pressure. Duration variance (actual versus planned) is the headline metric, and beating the plan is celebrated. Cost metrics track whether the shutdown came in under budget. Scope completion metrics confirm whether all planned tasks were finished. What none of these metrics capture is what happens after restart. Post-shutdown infant mortality, the spike in unplanned failures that occurs when equipment returns to service after maintenance, is tracked (if at all) as a separate operational problem unrelated to shutdown execution quality. Availability metrics in the 30 days after restart are reported in isolation, without linking back to the degree of shutdown compression that preceded them. QA hold point compliance during shutdown execution is typically a binary pass/fail that does not reveal whether hold times were shortened under schedule pressure. The net result is that shutdown duration compression appears to be pure upside, because the quality cost is measured in a different reporting period by different people using different metrics.
Decision Supported: Is the AI shutdown optimizer achieving duration reduction at the expense of post-restart reliability, and what is the net economic impact after accounting for post-shutdown failure costs?
Formula
Plain Language
For each shutdown, calculate how much the AI optimizer compressed the duration compared to the engineering estimate (duration compression). Then calculate the ratio of unplanned failure rate in the first 30 days after restart to the steady-state failure rate from the 6 months before the shutdown (infant mortality ratio). Compute the correlation between duration compression and infant mortality ratio across multiple shutdowns. A positive correlation means that shutdowns with more compression tend to produce more post-restart failures.
Notation
Duration_compression = 1 - (Actual_shutdown_hours / Estimated_shutdown_hours)
Infant_mortality_ratio = Failure_rate_30day_post_restart / Failure_rate_steady_state
SQDI = Correlation(Duration_compression, Infant_mortality_ratio) across multiple shutdowns
Variables
| Variable | Definition | Unit |
|---|---|---|
| Actual_shutdown_hours | Realized shutdown duration from asset lockout to restart clearance | Hours |
| Estimated_shutdown_hours | Engineering estimate of shutdown duration before AI optimization was applied | Hours |
| Duration_compression | Proportion of estimated duration that was eliminated by the optimizer | Ratio (0 to 1) |
| Failure_rate_30day_post_restart | Unplanned failures in the first 30 days after restart divided by operating hours in that period | Failures per operating hour |
| Failure_rate_steady_state | Average unplanned failure rate in the 6 months preceding the shutdown | Failures per operating hour |
| Infant_mortality_ratio | Post-restart failure rate relative to steady-state baseline | Ratio |
| SQDI | Pearson or Spearman correlation coefficient between duration compression and infant mortality ratio | Correlation (-1 to 1) |
Proposed formula, pending validation pilots.