Definition
The percentage of AI or digital twin generated maintenance recommendations that can be executed as specified, given actual parts availability, crew qualifications, tool availability, and outage window constraints.
The Measurement Gap
The digital twin recommends a maintenance scenario that is mathematically optimal but assumes parts availability, crew skills, and outage windows that do not exist. The "optimal" plan fails on execution, and the fallback plan is worse than what would have been chosen without the twin. No legacy metric measures recommendation executability. Maintenance cost per gross-ton-mile captures outcomes, not decision quality.
Decision Supported:
When the AI or digital twin says "do this maintenance," can we actually do it? Or are recommendations ignoring real-world constraints and setting up execution failures?
Formula
Plain language: For each AI/twin maintenance recommendation, check at planned execution time whether all resources are available: parts in stock, qualified crew available, tools available, outage window sufficient. Score as executable, partially executable, or non-executable. Compute percentage fully executable.
Notation:
RES = N_executable / N_recommendations
With partial credit: RES_weighted = (N_exec x 1.0 + N_partial x 0.5 + N_non x 0.0) / N_recommendations
Variables:
| Variable | Description | Unit |
|---|---|---|
| N_executable | Recommendations where all constraints met at planned execution time | Count |
| N_partial | Recommendations where 1-2 non-critical constraints require substitution | Count |
| N_non | Recommendations where a critical constraint cannot be met | Count |
| N_recommendations | Total AI/twin recommendations in evaluation period | Count |
Sign Convention: Higher is better. 1.0 = every recommendation was executable as specified.
Proposed formula, pending validation pilots.