Benchmark Report
Benchmark Report — Line Balancing (notional model)
Scope: quantify the value of takt-driven line balancing on the notional Starship Ship build model. Seam: the model is self-authored notional data (see proof/LIMITATIONS.md); this is a SIMULATED benchmark, not a real production-rate claim.
Method
- Subject: 16 build operations across IPV, PEZ, Raceway, Tank Cleaning, and
Megabay 2 Stacking, with finish-to-start precedence (src/vehicle.mjs).
- Takt: 60 min/vehicle (480 min/day ÷ 8 vehicles/day).
- Optimized: Ranked-Positional-Weight balance (
src/lean.mjs). - Baseline: one operation per workstation (no consolidation), scored at the
same 60-min cycle.
- Reproduce:
node verify.mjs→verification-report.json#/syntheticBenchmark.
Result
| Metric | Baseline | Balanced | Delta |
|---|---|---|---|
| Workstations | 16 | 11 | −31.25% |
| Line efficiency | 52.1% | 75.8% | +45.5% |
| Bottleneck | 60 min | 60 min | — |
| Throughput | — | 8 vehicles/day | — |
| Critical-path lead time | — | 56 h | — |
Interpretation
Consolidating 16 operations into 11 takt-bounded stations removes 5 stations of idle capacity and raises line efficiency by ~23 efficiency points, without violating any assembly precedence (verified, check 4) and without exceeding takt on any station. The theoretical minimum at this takt is 9 stations (check 2), so the RPW balance is within 2 stations of the lower bound.
Honesty note
This benchmark measures the algorithm on a model, not a factory. Real gains depend on real work content, real precedence, and real station constraints. The engines are data-agnostic — swap src/vehicle.mjs for real inputs (see run-deploy-instructions.md) to produce a real benchmark.