Overall timeline
T = start of AA 587 takeoff roll.
T-105 seconds (approx.): The JAL 747-400 begins its takeoff roll. Throughout the flight of AA 587, the horizontal distance between the two planes was somewhere between 4 and 4.3 miles. The JAL flight was about 800 feet higher than the AA flight generally speaking.
T-0: AA 587 begins its takeoff roll.
T+96 seconds: The flight data recorder indicates first encounter with wake turbulence. “Airplane attitudes did not change much.” Widened wake-funnel has significant cross-track vector components (think of as a sudden burst of cross-wind)
T+107: An airframe rattling noise is heard on the cockpit voice recorder. the faulty (repaired) attachment point has taken the brunt of the first (likely athwartwise) wake encounter and sheared. Structural integrity of the tail has now been lost (see later) and it is vulnerable to any further such side-blows.
T+114: The captain makes a comment about wake encounter.
T+116: Flight data recorder begins to show aircraft movement “consistent with wake encounter.” It is “similar in intensity” to the first wake encounter. The plane "responds normally" at first. Then come the three side acceleration excursions.
T+121: A second airframe rattling sound is heard on the CVR. (Tail has torn off?) No, loosened fin is now into a lengthy L/R rocking across its attachment base, having suffered an initial fracture at one or more of its attachment point composite upper lugs (on the gustwards side). This rocking process is a destructive flutter-failure mode. Its death-rattle is transmitted via the "sounding board" of the fuselage.
T+121.5: Rudder position data becomes “unreliable.” (Rudder has begun to depart along with tail?) AGREE, in that the rudder is beginning to be activated [by the flight control system] to counter L/R yawing motions induced because the base of the fin has become semi-detached on one side and is "rocking quickly left and right). At about this point, the plane begins to undergo the “side acceleration excursions.” Caused by the rocking (tip of fin moving through as much as a metre) First there’s a .3 G excursion in one direction, then a .4 G excursion in the same direction, then a .3 G excursion in the opposite direction. This is “coincident with rudder deflections.” (Tail has ripped off but the pilots don't know it?) DISAGREE, Tail is shaking lose - vertical fin is pivoting crooss-ship to and fro (i.e. L/R relative to the direction of flight) around its base due to one side's composite upper brackets being mostly sheared. "coincident with rudder deflections" simply means that the flight control system is feeding in compensatory rudder inputs which are unfortunately out-of-phase with fin oscillation - and tend to accelerate the fin's breakaway. ) BREAKAWAY itself would be clean and indicated by a cessation of the drumming, suddenly ineffectual rudder pedals and a CofG change nose-drop. Over the next couple of seconds, the plane undergoes dramatic course changes: Its heading changes at a rate of about 10 deg/second, the bank angle increases through 25 degrees with the left wing down. Pitch drops to 30 deg nose down. Standard spiral dive entry due to loss of directional stability - as a function of the yaw resulting from the differential thrust outputs of the engines at "max power" Vertical force increases to 2 G’s. N.B.: NTSB description of eyewitness reports: “The general observation that the witness seemed to -- remarkably similar observations have been that they saw the aircraft "wobble," is the word used. "Wobble" aptly describes the L/R yaw/roll oscillations (fin movements and with rudder countering, yet reinforcing the destructive process) over the few seconds that it took for the fin to wholly detach - after being compromised on one side's brackets. And they saw pieces come from it. (fin and fairings/fillets, later the engines) And then it went into a steep spiraling dive into the ground.”
T+124: The FDR stops recording. (Electrical system goes out?) The plane is at an altitude of about 2,900 feet. If the FDR is in the tail area, possibility is that the fin detachment caused some disruption to the electrics thereabouts because of cables being pulled out/through and the resulting generator trips. The 2g max recorded would be insufficient to detach the engines - so they would have started detaching at about the centripetal forces equivalent to 5g (T+137 say) once the thrusted spiral dive had almost reached its terminal velocity (about 1000ft agl).
T+125: The copilot calls for max power. [N.B: CVR continues to work because it is powered by batteries.] Still think this "max power" call might have been precipitated by the nose-drop resulting from the sudden C of G change as the fin and rudder departed.
T+127: There are several comments suggesting loss of control of the aircraft. By this stage the aircraft would have entered the spiral dive (about quarter-way into first turn)
T+144: The cockpit voice recording ends.
Layout of the wreckage:
"If you look at the locations of the fin and the rudder, first in line in direction the airplane was traveling, the fin was first. Then at about -- then came the rudder about 200 yards later (rudder/fin flailing after detachment would have quickly caused hinge-line fracture) . And then all of this is within a -- one-half nautical mile of the crash site. So it's a fairly tight cluster, relatively speaking, of the wreckage. Spiral would have been tight - about a half-mile radius event . Engine detachments may have been almost coincident (pylon attachments being similarly stressed by the same centripetal forces)
Power plants, it is the left engine, the number one engine, in case you're interested, that is at the gas station. That's on 129th Street. That's about 700 feet from the Crater. The right engine, number two, is next to the boat behind the house. It's on 128th Street. And that's about 800 feet from the crater. But again, if you center all of this wreckage, it's more or less in a line.
UNIDENTIFIED MALE: How far from the crater?
BLACK: Number one engine 700 feet. Number two engine 800 feet. And these are rough measurements."
The lesson is obviously this: Composite structures (unlike fairings, fillets and other non load-bearing components) get their integrity and strength by being born of a singular process. A mixture of resins during in situ repairs just create weak-points that are always going to be vulnerabilities (as in this case). A glue sitting atop a glue will never make a good bond. When you incorporate a doubler and make that attachment point dissimilar and perhaps even sitting proud by a few mm, the likelihood is that it will be taking a disproportionately higher load (as well as being inherently weaker). That is not good. Thereafter whacking a few rivets through (as a salve to safety and QC) is really just a concession to your realisation and admission that it is otherwise a weaker proposition than the other attachment points. The six attachment points have to be both equally strong and share the loads equally, otherwise you have created a path for progressive failure. Structural composites have to be 100% as born from their process - or ditched and replaced. Repairs, in situ or otherwise, remain a bad idea. One has to wonder how many other "repairs" are out there.