The University of Dayton Research Institute (UDRI), Dayton, Ohio, was contracted to conduct an independent analysis of the take-off performance of the aircraft using previously developed computer simulation techniques (DSS Contract No. 4M012-6-0005/0I-FR). Data used in the analysis included that derived from the FDR and aerodynamic performance information provided by Douglas Aircraft Co.

A two-pan study was conducted. A take-off sensitivity analysis was performed using a digital, fixed-stick simulation program to establish the relative performance degradation resulting from a variety of factors which were identified as having potential to adversely affect take-off performance. The second approach was to reconstruct the accident trajectory by solving the airplane equations of motion.

The take-off sensitivity analysis used a two-dimensional, three degrees of freedom digital take- off program to simulate various take-off scenarios. A normal take-off trajectory was simulated, and then various abnormal trajectories were generated under assumed conditions or events that might have degraded take-off performance.

For this analysis, a normal take-off consisted of initiating rotation one second after V_R and rotating to a pitch attitude of about 13 degrees at a rate of just under two degrees per second. This rate of rotation took into account the geometry-limited properties of the aircraft by ensuring that the aircraft was airborne before a pitch angle of 8.6 degrees was achieved. This resulted in the aircraft rotating to a pitch attitude of 12.6 degrees at a rotation rate slightly less than two degrees per second. An airspeed of V₂ + 10 was achieved at 35 feet above ground level and then maintained during the climb-out. The take-off weight used for the normal take-off simulation was 344,500 pounds. The corresponding take-off reference speeds were V_R - 154 KIAS; V₂ - 166 KIAS. Ground effect was considered in the simulation.

The abnormal conditions and events evaluated in the sensitivity analysis were early rotation; reduced thrust in one engine; failure of one engine; failure of two engines; and ice-contaminated wings. The individual effect of each factor on a normal take-off, as well as the combined effect of several factors, was evaluated.

The sensitivity analysis concluded that, of all the factors and events considered, the lift and drag penalties associated with ice-contaminated wings were necessary to result in a Qight profile that resembled the accident trajectory.

The reconstruction of the accident trajectory used a technique developed by UDRI for an analysis of a previous take-off accident involving a Boeing 727. Several changes to the original equations were made as well as slight modifications to the solution method. Aerodynamic and thrust data provided by Douglas Aircraft Co. were curve fitted and interpolated for insertion into the computer program. A moment equation was incorporated into the program in order to calculate elevator deflections. An algorithm was used to calculate pitch rate and rotational acceleration from the assumed pitch history profile. Terrain elevation was explicitly included in the take-off run and used for the calculation of the ground effect during the airbome segment of the accident flight.

The equations of motion were solved iteratively with known conditions as constraints. The solution of the equations of motion of the aircraft determines the lift coefficient (CL). The derived lift coefficient is largely insensitive to the assumed pitch history. The drag coefficient (CD) derived from solving the equations of motion is not as accurate as the CL, since, in this case, it was dependent upon assumed thrust. For example, loss of thrust from one engine cannot be distinguished from a 0.05 increase in CD. The UDRI accident reconstruction study concluded that the only acceptable solutions to the aircraft equation of motion required a significant loss of lift and a significant increase in drag. The calculated reduction in lift was approximately 30 per cent, the increase in drag at least 100 per cent.

At the request of the Board, a second independent series of computer performance simulations was performed. This work was conducted by a flight dynamics specialist of the Depanment of National Defence (DND). The primary purpose of these simulations was to funkier analyse the performance of the aircraft with varying amounts of thrust and of ice contamination on the wings.

Perfommance estimations were made using lift and drag data for the DC-8-63 provided by Douglas Aircraft in conjunction with the following additional assumptions: aircraft weight - 344,500 pounds; pitch angle at lift-off - 8 degrees; field ambient conditions - temperature -4.2 degrees C, altitude 425 feet asl. For reference purposes, lift-off was considered to occur 8,000 feet after the start of the take-off roll. Ground effect increments to lift and drag coefficients were eliminated if the aircraft height above ground was greater than that of the aircraft wing span or if the distance from start of take-off was greater than 10,500 feet (this distance corresponds to the location where the ground slopes steeply away at the end of the runway). Douglas Aircraft lift and drag data were extrapolated from 16 to 18 degrees angle of attack to obtain data over the angle of attack range O to 18 degrees.

The computer program written for the simulation calculated the aircraft post-lift-off perfommance at the desired time increments for a total of 20 seconds. The perfommance calculations were based on accelerated climbing/descending equations of motion.

The performance calculations were separated into two groups: cases with surface contamination and four engines operating nominally; and, cases with surface contamination and a single engine failure at a specified time. In addition, several cases without surface contamination were run as test cases to validate the program and provide a basis for comparison. These test cases demonstrated that the program accurately estimated the nommal climb performance of the DC-8-63 predicted by Douglas Aircraft.

Program options for individual simulation runs were as follows:

1. Time increment
2. Equivalent surface roughness - 0, 0.02, or 0.04 inch elements
3. Engine failure - failure time and degree of thrust loss
4. Acceleration - sets initial acceleration
5. Climb-out speed - target speed for steady climb-out
6. Maximum pitch attitude
7. Overrotate - allows rotation to pitch attitude higher than optimum for the degraded aerodynamic properties
8. Pitch rate start - specifies time for overrotation if allowed
9. Pitch-up - forces angle of attack to 18 degrees if speed decreases to stall in order to simulate high drag associated with the stall

The program was limited in that it could not duplicate the dynamic control inputs of the pilot at the controls, and, since an accurate pitch history was not available, it is unlikely that any predicted performance would exactly match the complete night trajectory of the accident take-off. However, assuming that the aircraft was flown using normal procedures (ie. nommal pitch limits, pitch rates and airspeeds), the first portion of the trajectory could be estimated with reasonable accuracy. Once the aircraft perfommance began to degrade below normal, the predictions become less accurate because pilot inputs have a significant influence.

The cases of surface contamination with four engines operating showed that the aircraft was capable of a safe climb-out with either contamination with surface roughness elements of 0.02 or 0.04 inches if the aircraft was flown at the optimum pitch attitude (which was lower than the nommal pitch attitude for climb-out with a clean airfoil). Small changes in pitch angles or airspeed had a significant effect on aircraft performance for both 0.02 and 0.04 inch surface roughness elements. Increasing pitch angle from 12.5 to 15 degrees with 0.02 inch contamination elements was sufficient to degrade the climb performance so that a successful climb-out was not possible.

With surface roughness contamination elements of 0.04 inches, the aircraft was more sensitive to pitch angle increases. A successful take-off with four engines operating and with surface roughness contamination elements of 0.04 inches was only possible if the pitch angle was maintained at or below 11.7 degrees.

For those cases which included an engine failure along with contamination of the wing surface, even lower pitch angles were necessary to ensure a successful take-off. When contamination with surface roughness elements of 0.02 inches was combined with an engine failure, a successful take-off was possible only if pitch angle was maintained at or below 11 degrees; with surface roughness elements of 0.04 inches and an engine failure, a successful take-off was not possible.

In all, 44 different simulations were conducted (including those which were conducted to validate the program). The simulations which resulted in the best "match" to the accident flight were those with lift and drag penalties associated with wing contamination with the equivalent of full surface roughness elements of 0.02 inches and the loss of thrust from one engine, and those with lift and drag penalties associated with wing contamination with the equivalent of full surface roughness elements of 0.04 inches, with or without a loss of thrust from one engine.

As part of its investigation, the Board conducted a series of simulator tests using a DC-8-63 training simulator located at the Sterling Airways Flight Training Centre, Kastrup, Denmark. Represen- tatives of Arrow Air, Douglas Aircraft Co., Pratt & Whitney, the FAA, and the NTSB were Dresent and observed the tests.

The aim of the simulator testing was to duplicate the situation faced by the crew in Gander on 12 December 1985. Various scenarios generated by technical concerns arising from the investigation, perfommance predictions by the aircraft manufacturer, the computer simulations performed by UDRI, and the Board's own perfommance analysis were "flown" by pilots from Arrow Air, Douglas Aircraft Co., and Sterling Airways.

The simulator was manufactured by Canadian Aviation Electronics (CAE) of Montreal, Quebec. For the purposes of the tests, it was programmed to reflect, as closely as possible, the ambient conditions at Gander at the time of the accident.

Prior to conducting the test "llights", the fidelity of the simulator was checked, both quantitatively and qualitatively. It was concluded by all who attended the tests that the simulator had reasonable lift, drag, and thrust fidelity in the night regime of interest. Handling qualities were determined to be acceptable.

The test scenarios included the use of low take-off reference speeds, failure of the number four engine, extension of the pitch trim compensator (PTC), deployment of the number four engine thrust reverser, and ice-contaminated wings. The scenarios were "flown" individually and in various combinations. All tests were flown by the pilot in the right seat. Various recovery techniques were utilized by the pilot when abnormalities occurred.

For those tests which simulated ice contamination of the wing, the simulator computer was reprogrammed with modified coefficient of lift and drag values derived from data provided by Douglas Aircraft Co. and UDRI. The changes in CL and CD were consistent with that occurring with upper wing surface contamination with roughness elements of 0.04 inches. The CL maximum value was achieved at 10 degrees angle of attack, which conformed to the value predicted in the UDRI perfommance study. The 10-degree value also provided a compromise fit for lift coefficient value imposed by the existing software description of the coeffficient of lift curve. Reprogramming was perfommed by the engineering staff of the Sterling Airways Training Centre.

The only scenarios flown which came close to duplicating the actual performance of the aircraft during the accident take-off were those that included the altered coefficients of lift and drag. Any attempt to ny the simulator at nommal climb-out angles with these CL and CD values resulted in a stall just after passing the runway end. The stall occurred at 168 KIAS at a pitch angle of about 12 degrees.

Rotating at a higher airspeed, reducing the pitch angle used to angles below the normal climb- out angle. and using full power after lift-off enabled a successful take-off to be conducted. It should be noted, however, that the detrimental effects on pitch stability associated with ice contamination could not be simulated.

After the stall occurred, the simulator "nose" would drop, and post-stall angles of attack could not be achieved. Because the simulator stall and post-stall qualities did not accurately reflect the manner in which the aircraft would respond during and after a stall, the drag values had limited value in the simulation with regard to trajectory prediction.

The tests also demonstrated that it was possible to maintain aircraft control with an outboard engine in idle reverse.

Additional simulator testing was conducted using a DC-8-63 training simulator located at the Flying Tigers Training Centre, Los Angeles, California. Representatives of Arrow Air, Douglas Aircraft Co., Pratt & Whitney, and the FAA were again present and observed the tests. As in the previous testing, the simulator was programmed to reflect as closely as possible the ambient conditions at Gander at the time of the accident, and the fidelity of the simulator was verified, both quantitatively and qualitatively.

The purpose of this second series of tests was to examine several accident scenarios which in- volved aircraft system malfunctions deemed by the Board to require funkier scrutiny. The scenarios examined included PTC runaway, full reverse thrust on the number four engine, in-flight deploy- ment of the ground spoilers, asymmetric flap conditions, take-off with closed slots, jamming of the elevator. and a comDIete hydraulic system failure.

Similarities with the accident night profile were observed in several of the scenarios tested. Application of full reverse thrust on the number four engine, a jammed elevator, severe flap asymmetry and attempting take-off with flaps retracted all resulted in marginal aircraft control. Although in some instances the pilot was able to complete a take-off successfully, the margin of control was such that, under actual night conditions and without any prior warning, successful completion of a take-off would be doubtful.

Airspeed and altitude values similar to those of the accident take-off were observed in the test runs that simulated application of full reverse thrust at or shortly after lift-off, take-off with asymmetric flaps (O and 18 degrees) and slots closed, and take-off with naps retracted and slots closed. Attempting take-off with either the left or right wing flaps in the retracted (O degrees) position resulted in the sounding of the take-off warning horn.

In each case that simulated a jammed elevator, the pitch angles that were achieved prior to lift-off would have resulted in a tail strike.

Test scenarios which simulated extension of the PTC, complete hydraulic failure, and take-off with the slots closed all resulted in successful take-offs. In each case, the take-off was completed without significant difficulties being experienced by the pilot at the controls.

It proved impossible to simulate in-flight deployment of the ground spoilers. Although cockpit indications of spoiler deployment were obtained (illumination of spoiler deployed light), no change in aircraft performance was observed.

It was also noted that, when faced with a situation involving degraded climb performance or control difficulties, a gear-up selection was rarely completed.

Considerable research has been conducted in the past two decades concerning the subject of flight crew fatigue. As a result of this research, fatigue-inducing factors and the consequences of fatigue on human performance have been identified.

Fatigue can be described as either "acute" or "chronic." The former refers to fatigue of short- term origin usually brought on by intensive and repeated activities and is often influenced by a short-term irregular sleep pattern; the latter refers to fatigue of long-term origin, is usually characterized by extended accumulation of flight and/or duty time, and sometimes may be accompanied bv long-term sleep degradation.

Key-fatigue producing elements have been identified as extended accumulation of flight and/or duty time; inadequate rest prior to night; multiple time-zone travel; flights which span the normal sleep period; short layovers; flights in an easterly direction; seven-day-plus flight panerns; and exposure to noise, vibration, and the aircraft microclimate which produces low humidity and cabin altitudes as high as 7,000 feet.

The effects of fatigue have been identified as judgement deterioration, alertness deterioration, an increase in error rate, irritability increase, and the development of sleep hunger, all of which have detrimental effects on the performance of flight crews.

Recent research has concentrated on quantifying fatigue-producing work patterns so that the likelihood of fatigue can be predicted. Several fatigue-rating indexes have been developed to be applied in the analysis of flight crew schedules.

Dr. Stanley Mohler, the Director of Aerospace Medicine, Wright State University School of Medicine, testified at the Board's public inquiry, regarding a fatigue-rating index he had developed in conjunction with other aerospace medicine researchers.

The index scores each of the flight segments of a schedule in accordance with a number of known fatigue-inducing factors. The cumulative fatigue potential of a given schedule is then calculated and compared against a Physiologic Fatigue Index which reflects a range of physiologic demands.

At the request of the Board, Dr. Mohler applied his fatigue-rating index to the December schedule of the accident flight crew up to the time of the accident. The physiological index for each flight segment was determined to fall into the category of "may dangerously deplete physiological reserves."