Calibration relationships tighten the bound modestly

All 54 static checks pass. The complete eight-part orientation covers reduce the reference-acceleration allowance by 3.7–7.1% across the six cases. This is useful progress in the bounding method; it does not establish useful long-record prediction ranges.

These are generated software cases, not new aircraft analyses. The allowance describes how much the reference acceleration may change across the permitted calibration set. It is neither measured aircraft acceleration nor sensor noise.

All acceleration allowances are in m/s². Each cover value is the worst of all eight fixed yaw parts. No part was dropped. All 51 coordinates and all 30 original training inequalities were retained, with 48 checked coupling constraints added in every calculation. The physical bounds stayed fixed.

The comparison took 4.85 minutes on one numerical thread. It used 3,240 linear-program proposals and 9,720 exact support reservations; all indexed journals, source hashes and sealed results were verified. There were no trajectory integrations, aircraft reads or optimizer starts.

The next step is to trace the larger position-reconstruction allowance to its remaining terms before another expensive full-history calculation. Whole-domain numerical qualification, the complete prediction period and the aircraft error model remain separate requirements. The main 53-stretch findings are unchanged.

Method and finite limits · Original prediction failures · Shared implementation · Prediction requirements · Reproduce the evidence.