WHAT WENT WRONG BETTER PILOT

Autopilots Are Not Automatic The importance of understanding how to operate avionics BY J. MAC MCCLELLAN

TAKEOFF INTO LOW CLOUDS or visibility demands careful planning and a thorough check to be sure all avionics and systems are set properly. The pilot of a Cirrus SR22 appeared to have done just that before departing into a 300-foot overcast at Cuyahoga County Airport in Cleveland, Ohio, bound for Buffalo, New York. But shortly after liftoff everything went wrong. The instrument-rated private pilot had logged more than 1,300 hours’ total time. He reported 400 hours of instrument flight experience and had completed a transition course into the Cirrus. The pilot’s logbooks were not available to the NTSB, so it could not determine how many hours he had logged in the SR22 he had purchased about eight months before the accident. The pilot had flown the Cirrus from his home base at Buffalo just a few hours before the accident flight. Weather conditions for his arrival were similar to the low clouds and light rain that were over the airport for departure. The SR22 was equipped with a flat-glass primary flight display (PFD) and multifunction display (MFD) that were linked to the autopilot. The avionics system included a recorder that noted airplane flight tracking, including attitude, altitude, airspeed, vertical speed, and also recorded selected autopilot modes. The pilot’s clearance was to fly the Runway 6 heading up to 3,000 feet. Before takeoff the pilot set the 3,000-foot altitude assignment into the altitude “bug” window. He also selected runway heading on the heading bug and entered a target vertical speed of 850 fpm. That is exactly the standard procedure that should be used for any IFR takeoff, particularly one into low clouds or low visibility. The avionics system also had a flight director that could show the pilot what pitch and bank attitude to maintain to satisfy the heading and climb targets he had entered into the system. Or the autopilot could be engaged and it would fly the airplane to the preselected targets.

82 Sport Aviation November 2011

This pilot clearly had confidence in his avionics/autopilot system, or perhaps not a great deal of confidence in his instrument flying abilities, because just five seconds after the flight-tracking data recorded liftoff he engaged the autopilot at an altitude of 61 feet above the runway. The tower controller reported that the takeoff and initial climb looked normal, but the Cirrus was quickly out of sight in the overcast. The pilot initially engaged the autopilot in heading hold mode so the system would maintain the runway heading selected by the heading bug on the PFD. One second later the pilot engaged the autopilot vertical speed mode. Pressing the vertical speed mode button on the autopilot control panel causes the autopilot to capture and attempt to maintain whatever vertical speed is set in the socalled “bug” window. Pressing the altitude button on the autopilot causes the system to capture the present altitude, whatever it may be. Pressing both the altitude mode and vertical speed mode buttons at the same time will “arm” the altitude hold mode so the autopilot will automatically capture the target altitude set in the altitude “bug.” For whatever reason, just one-half second after the pilot engaged vertical speed mode the autopilot went into altitude hold

mode and began to level the Cirrus at the current altitude of 940 feet MSL, just 61 feet above airport elevation. The target altitude “bug” reset to 940 feet because the altitude hold mode overrode the 3,000-foot target that had been preset before takeoff. About 12 seconds later with the airplane still at a dangerously low altitude the autopilot vertical speed target was reset to 700 fpm climb, but the recording showed the autopilot remained in altitude hold mode. Five seconds later vertical speed mode was engaged, but the selected rate was 400 fpm down. The recorded altitude target remained at 940 feet. At this point it appears that the pilot intervened because the Cirrus began a right turn from a heading of about 60 degrees. The turn continued for about one minute and 20 seconds, with altitude varying between 1,275 feet and 1,820 feet during the turn. Toward the end of the turn the bank angle reached an alarming 70 degrees right wing down and airspeed increased to 172 knots. During the extended right turn, autopilot modes jumped all over the place with both climbs and descents entered into the system, and heading and vertical speed modes being turned off and on. The airplane momentarily rolled wings level at 1,200 feet, but the nose was down 9 degrees. The pitch attitude began increasing until reaching 54 degrees nose up with the left wing down 29 degrees and airspeed decaying through 65 knots. This entire pull-up took about 19 seconds, with airspeed down to about 50 knots when it began to increase again. Altitude topped at 2,700 feet. Pitch attitude went from 50 degrees nose up to 60 degrees down. During this period the autopilot and flight director modes were all recorded as off. The airplane made another turning zoom that took it from 150 knots to 50 knots of airspeed in a left turn that reached 76 degrees left wing down. The autopilot heading mode was recorded on, but 10 seconds later turned off. The recorded altitude target was 12,000 feet, and the vertical speed was “bugged” at 1,700 fpm up. One more apparently out-of-control zoom was recorded taking the Cirrus up to 3,200 feet, with an airspeed of about 50 knots and a nose-up pitch of 50 degrees before it rolled through level into a right

turn. The recording showed the roll continued beyond the vertical to 120 degrees right wing down. The airplane crashed in a wooded area not far from the airport, killing both people aboard. One witness reported hearing an airplane that sounded like “it was doing tricks in the air,” but he couldn’t see it because of the clouds. Another witness driving in his car reported he saw the airplane at about 150 feet above the ground just before impact. During the entire episode the pilot made only two transmissions, both on the tower frequency, not the departure control he had been instructed to switch to. About two minutes after takeoff the pilot transmitted, “We’re having trouble getting...” About two minutes after that he transmitted, “Having trouble,” again. He did not respond to transmissions from the tower after acknowledging the instruction to switch to departure control shortly after takeoff.

During the extended right turn, autopilot modes jumped all over the place with both climbs and descents entered into the system, and heading and vertical speed modes being turned off and on. The flight recording in the avionics system also was programmed to record any fault annunciations for the PFD display or the autopilot, and there were none. The NTSB did not determine if the autopilot remained engaged and functioning at all, or most, times during the flight. However, the board did include a statement from the SR22 pilot’s operating handbook that says “it is possible to easily override full trim or autopilot inputs by using normal control inputs.” In other words, the NTSB is reminding us that even if the autopilot were doing its best to level the wings, the human pilot could easily overpower the automatic system. For a probable cause the NTSB comes to the incredibly obvious conclusion that the “pilot failed to maintain control of the

airplane while operating in instrument meteorological conditions due to spatial disorientation.” The more meaningful “contributing cause” from the board is “the pilot’s inattention to basic aircraft control while attempting to program the autopilot system.” But as a hint that it was left wondering about the accident’s pilot’s IFR flying abilities and knowledge of how to use the autopilot system, the NTSB included some “additional information” at the end of the narrative report. It turns out that on arrival at Cuyahoga County just a few hours before the accident flight, the Cirrus pilot needed four attempts to fly a successful ILS approach. The weather conditions were right at minimums, with the ceiling varying from 200 to 300 feet. Decision height on the ILS is at 200 feet. On the first approach, the airplane never captured the glide slope and continued on at the initial altitude of 3,000 feet. On the second try, the airplane drifted right of the localizer and high on the glide slope. The pilot emerged from the clouds but elected to miss the approach instead of turn back to the runway. On the third ILS approach, the Cirrus remained high on the glide slope all the way down and never reached decision height. On the fourth ILS, the airplane tracked both the localizer and glide slope accurately, and the airplane landed. Autopilots can add a great deal of safety potential to any IFR flight, particularly for the single pilot. The autopilot can do the routine tasks of holding heading and altitude, and tracking course guidance, freeing up the human pilot to monitor the overall situation. But if the human pilot doesn’t fully understand how his autopilot operates, confusion can be the result. This article is based solely on the official final NTSB report of the accident and is intended to bring readers’ attention to the issues raised in the report. It is not intended to judge or reach any definitive conclusions about the ability or capacity of any person, living or dead, or any aircraft or accessory.

J. Mac McClellan, EAA 747337, has been a pilot for more than 40 years, holds an ATP certificate, and owns a Beechcraft Baron. To contact Mac, e-mail [email protected].

www.eaa.org 83