I was asked by Bob Tilley to be a member of an accident evaluation team for Rotor-EZ N2992 which crashed in May, fatally injuring the pilot. The evaluation team consisted of Bob Tilley, Steve Wright, myself, and one other engineer who is also well respected member of the canard community and prefers to remain anonymous.
This analysis is based on some first hand visual inspection, visual inspection of photographs of the crash site and wreckage, verbal reports of eyewitnesses, and a small amount of testing of crash aircraft hardware and equivalent hardware. Although this report was written by Marc J. Zeitlin, it is the forensic engineering judgment of the whole team, and this analysis and the resulting conclusion is the consensus opinion of all four people contributors on the investigation team. Many possible scenarios were examined other than the ones presented here - these are the most likely scenarios, in our collective judgment.
I will continue discussions of this report on the canard-aviator's mailing list.
The first question is: "what caused the pilot to attempt a turn back to the airport"?
Background: The eyewitness reports indicated a partial loss of power, a steep turn and nose drop, followed by a total loss of power, then a leveling out. While eyewitness reports (especially from non-pilots) are notoriously inaccurate, the pilot's actions in attempting a return to the field make these reports convincing.
a) First Possible Scenario:
One member of the evaluation team informed us that when the aircraft was found (and as can be seen in supplied photos "100_3049.jpg" and "Fuel Selector2.jpg")
the Andair "LEFT"/"BOTH"/"RIGHT"/"OFF" FS20X4 Fuel Selector Valve was in an intermediate position, between "BOTH" and "LEFT". He stated that the NTSB/FBO removed the valve and attempted to test flow at this intermediate position, and it was marginal at best.
This valve has detents for valve position - a positive lock both to get into and out of the "OFF" position, and light detents for the "RIGHT" and "LEFT" positions. After some investigation and playing with an identical valve, we have determined that any light impact or jarring could relatively easily move the valve out of the "LEFT" position and into the position it was found, so it is by no means certain that the valve was in that position when the aircraft took off - in fact, it is likely that it was not.
Although a discussion with a COZY builder (who has the identical valve in his COZY MKIV) revealed (through his discussions with the Andair factory engineers) that the valve is designed in such a way so that proportional flow should be supplied from each of the "LEFT" and "RIGHT" ports when the handle is in an intermediate position (that's how they get "BOTH" - it's 1/2 from each) so that FULL FLOW is always available to the outflow port, no matter what the position of the valve is as long as it's between "LEFT" and "RIGHT", this builder's testing of his identical valve showed the following flow characteristics:
Off = 6 o'clock
Left = 9 o'clock
Both = 12 o'clock
Right = 3 o'clock
Began flowing 7:30 position.
Full flow at 9 o'clock.
Gradual restriction of flow to about 25% when at 10:30.
Gradual increase to about 50% flow at 12:00
Flow gradually reduced to, and stopped at 1:30 position
Began flowing 4:30 position.
Full flow at 3 o'clock.
Gradual restriction of flow to about 25% when at 1:30.
Gradual increase to about 50% flow at 12:00
Flow gradually reduced to, and stopped at 10:30 position.
In other words, full flow occurs at 9, 12, and 3, (50% each at the 12 position, for a total of 100%) and a minimum of 25% flow at the 10:30 and 1:30 positions, with NO flow from the opposite side.
This indicates that the valve would have experienced a 75% decrease in MAXIMUM flow rate with the handle in the position in which it was found. The Andair web page doesn't state what the maximum flow rate of the valve is - if it's 30 gal/hr, then the 25% flow rate would be 7.5 gal/hr, which should be more than adequate to keep flying, if not have a stellar climb rate. If, as is more likely, it's OVER 30 gal/hr, then the reduced fuel flow rate really shouldn't be a huge problem, _IF_ the throttle was reduced to ensure the correct A/F ratio.
In our judgment, due to the ease of moving the valve between settings other than "OFF" and the probable maximum fuel flow rate, the fuel valve position is probably a red herring - it's not a major contributor to the engine power loss.
b) Second Possible Scenario:
One member of the evaluation team performed a flow test on the two fuel injectors that the pilot was using on his Rotary 13B engine. The pilot was using one injector per rotor. Here is that member's report:
"I was flow testing the injectors today. The first one I hooked up, turned on the EFI pump, allowed all air in the line to purge, then put power to the injector. Allowed time to get good flow thru it, shut it off, put it in a bucket to measure the volume, and watched the flow pattern. Everything looked good. Switched over to the second injector. the flow rate was pretty much the same but the pattern looked wrong, not an even spray more clumped like something was in the injector. So we tried it again, and the longer we had it on the less flow we got out, we double checked to insure the hose was in the gas etc. and all was fine. The flow eventually went to about 20% of full flow. The pilot's setup has only one injector per rotor. So essentially he was running on one rotor. For you nonrotary fans that's 3 cyl instead of 6."
Running on 1/2 of the cylinder equivalents would give far less than 1/2 power to the propeller. The pilot had stated that he believed that his engine was producing somewhere in the 150 - 160 HP region at full throttle at Sea Level - if it was now producing less than 75 HP, there's no question that a climb would be essentially impossible. The low power output also matches the eyewitness reports.
If only 20% fuel flow was available from one injector, and less than 1/2 power was available, we would think that this would be noticeable during the run up prior to flight or as a lack of RPM during the takeoff run. Otherwise, it would just be incredibly bad timing to have the nozzle clog just after rotation and liftoff.
Given the rubber fuel lines just upstream of the injectors, we believe it possible that a small bit of rubber got into the fuel line during assembly or modifications. There was also a small bit of foam/composite dust in the gascolator screen, so that is also a possible cause of the clogging.
Primary Cause Conclusion:
In our judgment, the scenario of clogged injector is the most likely one for a loss of power incident. It seems to fit the facts, symptoms, and observations. We'd give it a probability of approximately 90% as the primary cause of the accident, with a small additional possibility of reduced fuel flow due to valve position.
<Editorial Comment: given the semi-common contamination of fuel lines in composite aircraft due to construction methodology, for redundancy's sake using good fuel filters and two injectors per rotor in a rotary engine would be a good idea.>
The next question is "what did the pilot do when the engine lost power"?
Background: With his engine losing power, The pilot then apparently attempted a turn-back from low altitude - possibly somewhere in between 100 ft. and 300 ft. AGL.
a) First Possible Scenario:
Having successfully performed a turn back maneuver once before in this airplane, The pilot possibly had a false sense of confidence that he could do it again, from an even lower altitude. Ignoring, forgetting, or disagreeing with the recent discussion regarding optimum bank angles in a return-to-airport situation that had occurred on the canard-aviators list not a month or two prior to his accident, and engendered by his FIRST turn-back episode, he attempted a VERY highly banked (eyewitnesses said about 90 degrees of bank) turn at low speed. He did this to minimize the turning radius, which it does, but at the expense of excess altitude loss. The optimum bank angle (for ANY aircraft) is almost exactly 45 degrees - this will minimize altitude loss for a give teardrop turn. 90 degrees is FAR from optimal.
As anyone who has flown one of these canard aircraft is aware, a highly banked turn at low speeds will create an almost immediate nose-down attitude. This again matches up with the eyewitness accounts. The pilot was then at low altitude, picking up speed fast from a nose down attitude, in a steep bank which was almost certainly uncoordinated. This then caused the unporting the fuel tank feed line and a complete lack of fuel to the engine (also corroborated by eyewitness observations - they claimed that the engine died completely during/after the steep turn).
Due to the low altitude and extreme nose-down attitude, the pilot did not have enough time to level the wings and pull back (without stalling) to a level flight attitude whereby he could glide to the ground. He then impacted the ground in a nose low attitude, still banked to the left an indeterminate amount, and probably with the canard stalled.
b) Second Possible Scenario:
The pilot's Weight and Balance calculations come into play here. After examining the SQ2000's W&B spreadsheet, it seems as though the designers believed that the CG range of the SQ2K was 5.5" from forward to aft maxima (113" to 118.5"), rather than the 4.5" (97.5" to 102") of the COZY MKIV, even though they do not have lower winglets (which according to Nat Puffer's testing, add about 1/2" of rear CG range to the aircraft). If we assume that the rear CG is sacrosanct, then we have a 15" - 16.5" transformation from SQ2K to MKIV coordinates, depending upon how we interpret things.
With respect to this SQ2K in particular, the W&B calculations indicate that with the pilot at 220 lb, about 20 gallons of gas, and 10 lb of ballast in the nose (verified by one member of the evaluation team), this aircraft's CG would have been at about 120.4" in SQ2K coordinates - this is almost 2" past the SQ2K published rearmost CG position. We do NOT have any evidence that SQ/KLS ever did any stall/deep stall tests on the SQ2K aircraft to verify the rear CG position and margin for deep stall, and we know that without the lower winglets, there is reason to believe that the rear CG position SHOULD be farther forward than with a COZY MKIV. Given the SQ2K - MKIV transformation, this aircraft's CG would be somewhere between 1.9" and 3.4" REAR of the rearmost CG position on a COZY MKIV - well into instability and deep stall territory.
Why discuss W&B and deep stall? As the engine falters and speed decreases, the pilot begins a left turn back to the airport. According to COZY Newsletter #44:
".... and at a c.g. of 103.2, when the main wing stalled, the aircraft entered a spiral turn (not a spin) and Jim lost about 3,500 ft. before he was able to effect recovery."
A CG of 103.2 is only 1.2" past the rearmost CG position, albeit with the long canard. Even with the short canard, Nat conjectured that the deep stall margin was about 1" with the CG at 102". Since this aircraft was most certainly WELL aft of this equivalent position, a deep stall during his slow turn back to the airport, with a concurrent high descent rate spiral dive at a relatively flat angle becomes a VERY likely scenario.
A deep stall impact involves a much higher descent rate than a gliding descent entails, along with an inability to control the impact point, due to lack of directional and/or pitch control. Flying a canard aircraft with a too far rearward CG has long been known to be an extremely risky endeavor. The pilot's actions in not ensuring that he was flying within the published CG range exhibit very poor judgment on his part.
In fact, the two scenarios are not separate or mutually exclusive - if aft CG, a steep turn could induce a deep stall, even if one had what would otherwise be a safe forward speed.
The Pilot's Response - Conclusions:
In our judgment, both of these scenarios are possible - they both fit the facts, symptoms, and observations. We give the second "deep stall" scenario a probability of 75% as the cause of the TYPE of crash that occurred, and the first "steep turn" scenario a probability of 25%. In either case, the result is/would be the same.
Obviously, the pilot was going to hit the ground somewhere unless he got the engine running again, and fast, but the actions the pilot took, either in using way too steep a turn angle or in failing to ensure a correct CG position are what led to this particular set of circumstances.
Once in the alleged deep stall, it didn't matter if he got his engine running again or not.
The next question is "what happened when the aircraft hit the ground"?
It seems clear from photos "100_3044.jpg" and "100_3046.jpg" that the aircraft impacted the ground nose first on the left side, as the nose of the plane is sheared off to the right. we would estimate a nose down attitude of 20 - 40 degrees, with a horizontal velocity of 50 mph - 70 mph. The vertical velocity would be 30 mph to 50 mph - also a large range. Those are clearly pretty large ranges for attitude and speeds.
In photo "100_3045.jpg", we can see the left wheel buried a foot into the dirt, indicating an EXTREMELY hard impact of the gear on the ground. The lack of gouge either behind the wheel, or in front of it, indicates a lack of substantial forward speed as the plane hit the ground. The main gear is to the rear of the buried wheel, and the main fuselage is broken (see photo "100_3047.jpg") just behind the front seats, under the strakes, as one would expect with the main gear under extreme force.
The right main wheel is just to the right of the passenger seat, as seen in photo "100_3046.jpg", above.
I believe that after the nose and then the main gear hit the ground, the gear leg (a COZY MKIV strut) bent outward, forward and up (possibly almost to the strake). After the wheel buried and tore off of the leg, the strut rebounded, pushing the aircraft backward and to the right - one can only imagine the energy stored in the strut when bent almost flat - it would be more than enough to toss the aircraft backward a few feet. Since it hit first on the left side, the plane would also have been thrown a bit to the right on the rebound.
The strut did, in fact, tear loose from the mounting points on the fuselage, but it stayed apparently intact (it did not break in the middle) and stayed approximately in position in the fuselage. Due to its position, it would be possible for it to impart a rebound force to the aircraft even though it was torn from its mounting points.
Secondary to the fuselage impact, the canard broke free, the right winglet delaminated from the wingtip, and the instrument panel/nose section collapsed downward.
Crash Event - Conclusions:
In our judgment, this scenario seems to fit the observations. We'd give it a probability of approximately 85% as a crash description.
The last question is "why did the pilot die in this accident - could it have been survivable"?
During the first impact, as the fuselage contacted the ground, the apparently extremely flimsy folding seat that the pilot was sitting on collapsed flat. It seems that these seats have no true structure to them, and are closer to lawn chairs in their design and structural integrity. Judging from photo "100_3046.jpg", above, the pilot had a single shoulder belt on each side, with inertia reel retractors. These seats and harnesses would do absolutely nothing to protect a passenger during a high descent rate vertical impact.
A member of the evaluation team informed us that the pilot died of a broken neck and/or internal injuries, and was found in the back seat of the aircraft. This is consistent with impact damage, and the "rebound" from the landing gear that threw the aircraft rearward could have easily thrown the pilot rearward as well, especially since his seat and shoulder harness/seatbelt were not restraining him in any way. His neck could have been broken in the initial impact when the seat collapsed, or it could have occurred in the rebound event. Severe internal injuries could have occurred at either time as well.
Survivability - Conclusions:
We believe that the seating structure and restraining structure in this particular aircraft (and which seem to be consistent with the SQ2K design) were totally inadequate to prevent or mitigate injury during a high descent rate vertical impact event - we can see that the seat broke and came apart, and the single shoulder harness didn't/couldn't provide any restraint at all, once the seat broke (if it even could have before).
We believe that this MIGHT have been a survivable impact in a different aircraft - one with a solid seatback and headrest, with robust seatbelt and shoulder harness restraints. There is obviously no way to know for sure that a more robust structure would have saved him, but it is clear that this aircraft did not have such a robust restraint structure. The lack of this robust structure is clearly _A_ (not clearly _THE_) proximal cause of the pilot's death.
Probable substantial power loss due to clogged injector nozzle. This led to inadequate power upon takeoff and climb out. Subsequent poor decision making in attempting a turn back to the airport, rather than choosing to continue straight ahead or attempt a shallow turn to a less wooded area is one possible scenario. A deep stall during the turn, also leading to a very high descent rate, is a second, more probable scenario.
Basically, if this analysis is correct, poor preflight systems checking and/or poor luck with the clogged nozzle began the accident, and poor judgment either in flight and/or in W&B calculations caused this particular crash to possibly be much worse than it otherwise might have been.
There but for fate go we all - the pilot was a human being, and it caught up to him. I would need both hands to count the number of times I have been in situations that could have killed me, but I survived - this is NOT a comment on the pilot's capabilities or what he was as a human being. I merely mean to point out that we can never have too much information or be too careful. Pre-flight engine checks, W&B checks (ESPECIALLY IN 4-SEAT CANARD AIRCRAFT) and emergency procedure practice may have substantially mitigated, if not avoided, this accident.
Broken neck and/or internal injuries due to vertical impact and/or rebound G loading. Possibly/probably due to inadequate structural integrity of seats/harnesses.
Return to: Cozy MKIV Information
Copyright © 2005, All Rights Reserved, Marc