The analysis engines, and what they refuse to answer

Every hypothesis meets a measurement.

• — Energy balance, specific excess power, V-n envelope, stall margin, Mach and VMO exceedance.
• — Phase coverage from take-off roll through initial climb, cruise, approach, landing and ground roll — a landing overrun is analysed as a landing, not as an aborted cruise.
• — Thrust availability distinguishes total power loss from an asymmetric failure, and withholds a verdict when an oscillating flight path makes the energy apex unreadable.
• — Aircraft envelopes for common transport classes; an unlisted type is solved on a parametric synthetic envelope that is labelled as such on the result, never silently substituted.

Would the intervention have worked?

• — Thrust restored, drag cleaned, bank limited, speed traded and immediate escape, each flown from the recorded state.
• — Results are bounded by two physical ceilings: the class climb rate, and the energy the aeroplane actually held plus the energy the intervention added.
• — A survivable-window scan reports the last instant at which the intervention still avoided ground contact.
• — Interventions that do not apply to the recorded regime — an energy counterfactual against a pitch-driven dive, a low-energy trial on a high-energy event — are refused with the reason stated.

What is missing, and what it is worth.

• — Causal sequencing with antecedent vocabulary for flight control, trim, air data, ATC and TCAS events.
• — Evidence sufficiency scored per conclusion against ICAO Annex 13 reporting needs, with the specific holding that would move it most.
• — Analytical results are reported as analysis, never as evidence: a reconstructed chain never lifts a sufficiency score in place of the recorder it stands in for.
• — Gap radar ranks unexamined themes by the class of evidence that would close them.

The site as a measurement.

• — High-energy impact solving: TNT equivalence, crater scale, main-field and ejecta radii, fragmentation index and in-flight separation.
• — Structural break-up and wing or lifting-surface separation, with bounded descent and separation-height estimates rather than a ballistic point claim.
• — Ground-roll and structure-strike frames for overruns, embankment and building impacts, including ground casualties.
• — Survivability curves from deceleration, cabin and section integrity, exits, fire onset, rescue time and environment — reported as an uncertainty band, re-solved rather than widened.
• — No survivor is invented where no structural pocket exists, and none is rounded away where one does.

Two aircraft, one solution.

• — Closest point of approach, closure rate, separation-minima breach and warning time available.
• — Advisory compliance judged from the advisory instant where the time is known, and stated as a trailing-window inference where it is not.
• — Vertical and lateral avoidance counterfactuals, so a heading change can be tested and not only a climb or descent.

Rotary wing solved as rotary wing.

• — Weight-class rotorcraft envelopes — light single through heavy twin — with the fixed-wing stall and V-n tests gated off rather than misapplied.
• — Autorotation energy budget, flare height available, height–velocity exposure and vortex-ring onset judged against induced velocity.
• — Rotor integrity read from the wreckage: a separated head, mast, gearbox or tail boom is scored as a drive-train and rotor-system finding.
• — Drive train, anti-torque, degraded visual environment and rotary-wing recorder carriage carried as standing systemic asks on the gap radar.
• — Blind-tested against three public rotorcraft occurrences: a night terrain impact, an urban building strike and an offshore rotor separation.

A propeller is not a small fan.

• — Windmilling versus feathered disc drag, with the climb-gradient penalty an unfeathered propeller costs on the remaining engine.
• — Minimum control speed corrected for the bank actually held toward the live engine, and the margin the aircraft had against it.
• — Propeller overspeed and tip-Mach checks where blade speed is recorded, and governing left open where it is not.
• — Blade-signature forensics: bending and chordwise scoring read as power delivered at impact, with cause and consequence left undecided.
• — Reduction-gearbox continuity, autofeather, pitch-change and asymmetric-handling carried as standing asks on the gap radar.
• — Blind-tested against a public twin-turboprop asymmetry occurrence after lift-off; the frame is gated off entirely for turbofan aircraft.

Beyond fixed-wing energy.

• — Ditching: water-entry loads, hull integrity and flotation time.
• — In-flight fire: progression rate, containment and diversion time available.
• — Icing: accretion rate, performance penalty and certification-envelope exceedance.
• — Controlled flight into terrain and loss of visual reference solved as their own family, not as a control malfunction.
• — Fuel exhaustion or contamination, slow pressurisation loss, crew incapacitation, loading errors, ramp events, turbulence injury, post-crash fire and wake encounters each use their own governing tests.

A real hypothesis, never a shortcut to intent.

• — Tests sustained route deviation, communications loss, transponder state, navigation anomalies and subsequent flight-path consistency as one mechanism-specific frame.
• — Separates navigation interference or spoofing from a deliberate diversion where the evidence permits, while leaving motive and attribution outside the model.
• — Raises a deduplicated case alert when the saved worksheet crosses the stated threshold, so repeated analysis does not flood the alert rail.
• — An unexplained endpoint remains unexplained when the required operational, recorder or communications evidence is absent.