Air Canada 646

Case Studies

Eddie sez:

We practice missed approaches to the point they are automatic and our muscle memory probably does make it automatic: power, pitch, flaps, gear. After that reflexive move, we look at how well the airplane is climbing and accelerating. And then we look at navigation. And that almost always works for a missed approach initiated at decision altitude where the engines are still spooled up and we are 100 or 200 feet in the air. The crew of his Canadair Regional Jet found out that things change when you get lower to the ground and especially once the engines are retarded to idle. When that happens, the automatic response can bite you.

The accident report is quite good and among other things, offers the first definition of a "balked landing" that I've seen from an official source. It is basically the termination of the approach after the decision to land has been made. Where it gets particularly dicey is when that happens after the engines are at idle. There are two key takeaways. First, engine spool up time could make it impossible to accelerate if you pitch up into your Takeoff / Go-Around flight director cues. You need to keep the airplane flying before you attempt to climb. Second, your aircraft stalls at a lower angle of attack (and lower airspeed) if you are in ground effect.

Everything here is from the references shown below, with a few comments in an alternate color.

Last revision:

2021-08-24

Accident Report

  • Date: 16 December 1997
  • Time: 23:48
  • Type: Canadair CL-600-2B19 Regional Jet CRJ-100ER
  • Operator: Air Canada
  • Registration: C-FSKI
  • Fatalities: 0 of 3 crew, 0 of 39 passengers
  • Aircraft Fate: Damaged beyond repair
  • Phase: Landing
  • Airport: (Departure) Toronto-Pearson International Airport, ON (CYYZ), Canada
  • Airport: (Destination) Fredericton Airport, NB (CYFC), Canada

Narrative

[Aviation Occurrence Report A97H0011, §1.1]

  • Air Canada Flight 646, a Canadair CL600-2B19 Regional Jet, departed Toronto-Lester B. Pearson International Airport, Ontario, at 2124 eastern standard time (EST) on a scheduled flight to Fredericton, New Brunswick. On board were two flight crew, one flight attendant, and 37 passengers plus two infants. The first officer, in the right-hand seat, had been assigned the pilot-flying (PF) responsibilities for this flight. The forecast and reported weather for the Fredericton Airport for the time of arrival was vertical visibility of 100 feet and horizontal visibility of one-eighth of a mile in fog. The runway visual range (RVR) was 1200 feet for the landing on runway 15 with the runway lights set to strength 5.
  • The flight was unremarkable until the aircraft was on final approach to the Fredericton airport. The autopilot was controlling the aircraft based on commands from the crew, the flight management system, and signals from the ground-based instrument landing system (ILS) for runway 15 at Fredericton. The aircraft’s landing lights were on for the approach and landing. The captain saw the glow from the runway approach lights through the fog at about 300 feet above ground level (agl), 100 feet above decision height for the approach. At decision height, 200 feet above the runway, the captain, the pilot-not-flying (PNF), called the lights in sight and the first officer responded that he was landing. The first officer disconnected the autopilot, at about 165 feet above ground, to hand fly the rest of the approach and landing.
  • After the autopilot was disconnected, the aircraft drifted above the glide path, and twice the captain coached the first officer to get the aircraft down to the glide path. The first officer reduced thrust in response to the captain’s first mention to get the aircraft down, and he reduced thrust to idle at about 80 feet agl. Moments later, the captain, aware that the aircraft was left of the centre line but not knowing the distance traveled down the runway, and not sure that a safe landing could be made, ordered a go-around, which the first officer acknowledged. The thrust levers were advanced, the first officer selected the go-around mode for the flight director, and he started to increase the pitch of the aircraft to the command bar indications, 10 degrees nose up. About one second after the first officer acknowledged the go-around, the stick shaker (stall warning) activated. As the aircraft reached 10 degrees nose up, about one and one-half seconds after the stick shaker activated, the captain called flaps and selected them to the go-around setting, the warbler tone associated with the stall protection system (SPS) sounded, and the aircraft stalled aerodynamically. The aircraft rolled right to approximately 55 degrees of bank, and the right wing tip contacted the runway about 2700 feet from the threshold and 45 feet left of the centre line, the right wing tip bending upwards about four feet from the tip. The aircraft rolled left toward wings level, then, about 260 feet further down the runway, struck it again, this time banked about 20 degrees to the right with the nose down about 12 degrees. The nose wheel assembly broke off, the right winglet broke off, the radome and underside of the nose cockpit area were heavily damaged, and electrical power, except for emergency lighting, was lost. The aircraft rolled left onto its main wheels and, with the engines now at full power, departed the right side of the runway just past the intersection with runway 09/27. The aircraft plowed through the snow, on its main wheels, until it struck a ditch parallel to and about 200 feet from the runway. The tracks in the snow past the ditch were much lighter than the tracks left by the main wheels. These marks were made by flap fairings and aircraft equipment dangling on wiring still attached to the aircraft. The marks show that the aircraft became airborne after striking the ditch, very low to the ground, and flew in an arc to next strike a sand hill about 1000 feet right of the runway. Ground marks made by the aircraft were largely obliterated by traffic during the rescue, and it could not be determined where the aircraft first hit the hill; however, there were pieces of the aircraft near the bottom of the hill. At the top of the hill, the aircraft slewed to the right, struck some trees, one approximately 22 inches in diameter, and came to rest. The aircraft stopped on a heading of 314° magnetic, about 1130 feet west of the runway and 2100 feet from the first impact point on the runway.

[Aviation Occurrence Report A97H0011, §1.6.2.2]

  • During the initial stages of an aerodynamic stall of the CL-65, the local airflow separation on the wing is minimal, and the level of buffet from the airflow separation cannot be considered a significant cue to stall warning. Because of its lack of normal stall warning, the aircraft was equipped with an SPS. The SPS comprises a stick shaker device to indicate that the stall speed is being approached and a stick pusher to cause the aircraft to pitch down, if necessary, to keep the aircraft from actually stalling.
  • AOA information is obtained from two AOA vanes, one on each side of the nose of the aircraft. The SPS computer, with two independent channels, uses this AOA information, combined with Mach, flap position, and lateral acceleration to signal the crew of impending stall and to prevent the aircraft from entering a stall. When the AOA is changing, the computer notes the rate of change and, if necessary, applies a correction to activate the protection system ahead of its normal trip points.
  • When the aircraft is airborne, the SPS computer continuously monitors inputs to determine the SPS AOA trip points. There are three trip points, each at a higher AOA, which initiate the following actions:
    • Auto-ignition—when either AOA vane reaches this trip point, continuous engine ignition is activated as a precaution against engine flame-out at high aircraft AOA.
    • Stick shaker—when either AOA vane reaches this trip point, the respective shaker is activated and the autopilot is disengaged. Because the control columns are interconnected, the shaking can be felt on both control columns.
    • Stick pusher—when either AOA vane reaches this trip point, the warbler sounds and the STALL/switch lights on the glareshield panel flash red. The pusher activates when both AOA vanes reach the pusher trip points.
    • Information from the flight recorders indicates the following events occurred during the attempted go-around: the stick shaker activated when the aircraft was at 129 knots as the pitch was being increased through 4 degrees; and the right roll and stall onset occurred and the warbler tone activated when the aircraft was at 124 knots and with the pitch at 9.7 degrees. At the time the warbler sounded, the left and right AOA vane readings were approximately 8.7 degrees and 9.4 degrees, respectively. The pusher did not activate because, while the right AOA reached its trip point, the left AOA did not.

[Aviation Occurrence Report A97H0011, §1.18.1.3] Many CL-65 pilots stated that on final approach the CL-65 is in a nose-low attitude because of its fairly high approach speed. It is recognized that the combination of low pitch attitude, high approach speed, darkness, and low visibility may result in a sensation that the aircraft is approaching the ground too fast, which would result in a tendency to raise the nose and to round out the aircraft earlier than required. This tendency is particularly pronounced when first transitioning to the CL-65 both from slower, more conventional aircraft and from larger aircraft with a higher pilot eye-reference point. In addition, because the engines are above the C of G, there is a tendency for the CL-65 aircraft to pitch up when thrust is reduced. These illusions and aerodynamic tendencies have been recognized and are discussed during pilot training. Both the captain and the first officer were aware of these characteristics.

[Aviation Occurrence Report A97H0011, §1.18.1.4] Go-around certification.

  • There are a number of terms used to describe the various phases of flight during the final approach and landing of transport category aircraft, and there is some inconsistency in how these terms are used. The term “go-around” is used in a variety of ways by regulators, manufacturers, and operators to describe a procedure where an aircraft discontinues either an approach, or a transition to visual flight for landing, or a landing, and then climbs away. The implications of this variation are significant, as these procedures are not identical. Below is a description of each phase and the associated major considerations for a go-around in each phase. The definitions here are to provide a standard for discussions within this report and are not necessarily related to aircraft certification.
    • Go-around—the act of terminating an approach to land, for whatever reason, and climbing away.
    • Missed Approach—termination of an instrument approach at or above the minimum descent altitude (MDA) or decision height. Normally, at the start of the missed approach, the aircraft would be on the desired flight path, configured with the landing gear down and the flaps as required for the type of approach, and with the power and speed stabilized. A missed approach could be required because of the inability of the crew to see enough of the runway environment to land, or because the aircraft was not in a position to land safely.
    • Rejected Landing—termination of the approach to land after the decision to land has been made by the crew. Normally, at the start of the rejected landing, the aircraft would be on the desired flight path, configured with the landing gear down and the flaps in the land position, and with the power and speed stabilized. A rejected landing could be required because the crew’s view of the runway environment was lost or there was some obstruction on the runway. The term rejected landing is not used in the context of aircraft certification, but it is a common term in the aviation community.
    • Rejected Landing with Power at Idle—a go-around from a missed approach or rejected landing is started with the engine(s) at approach power. However, there will be times when a go-around is required, or deemed to be required, after the power has been reduced to idle for landing. This is the area of the approach to land where the crew of Flight ACA646 found themselves. There are no Canadian or American certification requirements related to a rejected landing with the power at idle, Transport Canada does not require manufacturers or operators to discuss the subject in applicable manuals, and pilots are not required to train for such a manoeuvre.
    • Balked Landing—a balked landing is a type certification manoeuvre, and refers to the all-engine go-around from VREF in the landing configuration, as described above for a rejected landing. This term is rarely seen in AFMs or other manuals used by operators and pilots. For operator manuals, aircraft type ratings, and proficiency checks, the term rejected landing is normally used.
  • According to information provided by Transport Canada after the accident, a go-around or balked landing outside the demonstrated flight envelope is a high-risk manoeuvre. If a go-around is attempted from a low-energy state, such as after the thrust levers are reduced for landing, ground contact is likely, and any attempt to commence a climb before the engines have achieved go-around thrust could result in a stall. This is primarily because of the time required for the engines to spool up to go-around thrust—about eight seconds.
  • Significant aircraft energy-state conditions at the time that the go-around was initiated as compared to the conditions expected for the manoeuvre are as follows:
    1. The airspeed was six knots below the VREF speed of 139.
    2. The engine N1 speed was 29%, 39% below the nominal approach thrust setting of 68%.
    3. (The PF pulled them to idle at 80 feet.)

    4. The engine N1 speeds reached 68%, a nominal approach thrust setting, five seconds after the go-around acknowledgment.
    5. From the time of the go-around acknowledgment., the engines reached 92% N1 in 7.5 seconds; from a nominal approach thrust setting, go-around thrust would normally be reached in three seconds or less.
    6. The flaps started to rise three seconds after the go-around acknowledgment.; flaps are normally selected and moving within a second of the go-around call.
    7. The airspeed decreased at a rate of about three knots per second throughout the manoeuvre; airspeed would normally increase during the procedure [go-around].

[Aviation Occurrence Report A97H0011, ]

Analysis

[Aviation Occurrence Report A97H0011, 1.18.3.3] Appendix C is a plot of three curves (coefficient of lift versus AOA) for the last eight seconds of the flight. The solid line represents the expected CL–alpha curve based on certification flight test data in free air and at an entry rate of less than one knot/sec. The dotted line depicts the CL-alpha curve computed by Bombardier for the accident flight, and the line with the triangular points is the curve computed by the Institute of Aviation Research (IAR), a division of the National Research Council (NRC) of Canada. Also shown on the graph is the position of the natural aerodynamic stall for a clean (uncontaminated) wing. [ . . .] There are significant differences between the expected and computed curves as follows:

  • The computed curves are displaced below the expected curve such that for a given AOA the coefficient of lift is reduced except in the area approaching the stall where the computed curve approximates or is slightly higher than the expected curve.
  • The accident aircraft stalled at an AOA that was approximately 4.5 degrees lower than expected for the natural stall.
  • The maximum lift coefficient (CLmax) achieved was approximately 0.26 lower than expected for the natural stall.
images

Photo: Computed versus expected CL-alpha, Aviation Occurrence Report A97H0011 Appendix C
Click photo for a larger image

[Aviation Occurrence Report A97H0011, § 1.18.3.8] A review of literature regarding how ground effect alters aircraft stall characteristics was carried out. Although it is known that there is a definite reduction in stall AOA and CLmax due to ground effect, the review revealed that the available flight data were very limited and analysis techniques were not reliable. Notwithstanding these limitations, an attempt was made to quantify the impact of ground effect on the aircraft coefficient of lift and AOA for the accident flight, which indicated that the ground effect at the point of wing roll-off was likely to have caused a small lift increase (in the order of 2%) and a reduction in AOA of less than 0.3 degree for the same lift.

[Aviation Occurrence Report A97H0011, §2.3.4]

  • According to Transport Canada, when the go-around was initiated, the aircraft was outside of the flight envelope demonstrated during the certification process. Specifically, the retarding of the thrust levers to idle, and leaving them there until the aircraft was lower than 50 feet, was tantamount to a commitment for landing, and placed the aircraft into the low-energy regime from which a go-around could not be completed without contacting the ground. The reduction of thrust to idle was made in response to the captain’s coaching and because the aircraft was on speed and above the desired glide path. The significant difference between the occurrence go-around and go-arounds practised in training was the low-energy state of the aircraft, the most significant being the low engine thrust. Even though the thrust levers were advanced when the go-around was acknowledged by the first officer, it took the engines five seconds to reach the approach thrust level that would normally be the starting thrust level for the go-around. During these five seconds, the aircraft was not in a condition either to accelerate in level flight or to climb without losing airspeed.
  • The aircraft operating philosophy stressing that the flight director commands must be followed for proper flight control is valid for most anticipated flight conditions. Notwithstanding, not all commanded pitch attitudes are achievable or safe. In particular, following the command bars in go-around mode does not ensure that a safe flying speed will be maintained because, unlike in the windshear guidance mode, the positioning of the command bars does not take into consideration the airspeed, flap configuration, and the rate of change of the AOA—all factors to consider in achieving an adequate stall margin.

[Aviation Occurrence Report A97H0011, §2.4]

  • The investigation revealed that when the aircraft stalled, the aerodynamic performance of the wing was significantly degraded from the expected performance based on certification flight test data. The wing stalled at an AOA of about 9 degrees and a CLmax of 2.06, compared to expected values of 13.5 and 2.32. A number of factors that had the potential to contribute to the performance degradation were identified during the investigation and these will be discussed to determine the contribution of each to the performance of the wing.
  • The computed CL–alpha curves (shown at Appendix C) indicate degraded performance at the lower AOAs and show CL values above those expected just prior to the stall. Although the expected curve was calculated using the same method as was used for the Bombardier computed curve, the flight dynamics for the expected data were closely controlled and this fact would account for some minor variations between the expected and computed curves. The higher CL values of the computed curves approaching the stall are partly attributable to the fact that the expected curve is for free air (out of ground effect). Ground effect would shift the expected curve to the left and thus place the computed CL values below the expected curve. In addition, the situation approaching the stall is not quasi-static, and there may be some dynamic stall effect resulting in higher CL values just prior to the stall.
  • The performance reduction noted on the accident flight at the point that the aircraft stalled was approximately 4.5 degree in maximum fuselage AOA and 0.26 in CLmax. The study on ground effect estimated that at the minimum height above ground reached by the aircraft just prior to the stall (approximately 10 to 20 feet), the contribution of ground effect would have been approximately 0.3 degree of AOA. Given the considerable uncertainty associated with this estimate, the maximum reduction in AOA resulting from ground effect is considered to be in the order of 0.75±0.5 degrees. If this was the case, then the other factors which caused the reduction in performance accounted for approximately 3 to 4 degrees of AOA.

Cause

[Aviation Occurrence Report A97H0011, §3.1]

  • Disengagement of the autopilot at 165 feet rather than at the 80-foot minimum autopilot altitude resulted in an increased workload for the PF, allowed deviations from the glide path, and deprived the pilots of better visual cues for landing.
  • The go-around was attempted from a low-energy situation outside of the flight boundaries certified for the published go-around procedures; the aircraft’s low energy was primarily the result of the power being at idle.
  • The sequential nature of steps within the go-around procedures, in particular, in directing the pitch adjustment prior to noting the airspeed, the compelling nature of the command bars, and the high level of concentration required when initiating the go-around contributed to the first officer’s inadequate monitoring of the airspeed during the go-around attempt.
  • The published go-around procedure does not adequately reflect that once power is reduced to idle for landing, a go-around will probably not be completed without the aircraft contacting the runway (primarily because of the time required for the engines to spool up to go-around thrust).
  • The aircraft stalled at an angle of attack approximately 4.5 degrees lower, and at a CLmax 0.26 lower, than would be expected for the natural stall.
  • The maximum reduction in angle of attack resulting from ground effect is considered to be in the order of 0.75±0.5 degree: the aircraft angle of attack was influenced by ground effect during the go-around manoeuvre.

See Also:

Aviation Occurrence Report A97H0011, Loss of Control on Go-Around (Rejected Landing), Air Canada Canadair CL-600-2B19 C-FSKI, Frederiction Airport, New Brunswick, 16 December 1997, Transportation Safety Board of Canada