JOURNEY IN AERONAUTICAL RESEARCH: A Career at NASA Langley Research Center

Monographs in Aerospace History, Number 12

 

 

CHAPTER 15

Accident Investigations

 

[137] The NACA and NASA were not routinely called in to give advice on all accident investigations, but in some cases in which the cause of a crash could not be readily determined or when a new experimental design was involved, the appropriate government organizations requested assistance. Accident investigations are important not only to correct defects in existing designs that might cause further accidents, but also to provide information for safer designs in the future. I was directly involved in a number of these investigations, some of which are sufficiently interesting to include in this autobiography.

In this discussion, the term "dynamic pressure" is used. The dynamic pressure is approximately the increase of pressure above the surrounding atmospheric pressure caused by the flow impinging on a flat surface normal to the airstream. It is given by the formula one half times density times the square of true airspeed. Such a value of pressure occurs at the nose of an open-ended tube facing into the airstream, called a pitot tube, and is used to measure airspeed as well as to determine loads on the airplane caused by the airflow. The term "indicated airspeed" is the airspeed shown by a standard airspeed meter when it is supplied with this pressure and a reference pressure called the static pressure, which is the true ambient pressure in undisturbed air. The instrument is calibrated to read true airspeed for the density at standard sea level conditions. The true airspeed, required for navigation purposes, is greater than the indicated airspeed at altitudes above sea level because the density decreases with increasing altitude.

The abbreviation MAC stands for mean aerodynamic chord, a reference line based on the wing plan form that is used in deriving center of gravity location and in computing stability derivatives.

 

BAC-111 Crash Near Omaha, Nebraska, on August 6, 1966

 

From time to time, in investigations of crashes of civil aircraft, the NACA or NASA were contacted by the Civil Aeronautics Board (CAB), later called the National Transportation Safety Board (NTSB), to assist with accident investigations. One of these accidents that involved considerable study on my part was the crash of a BAC-111. This airplane was a small twin-engine jet transport, constructed by the British Aircraft Corporation (BAC), that was used rather extensively on short-haul routes in the United States (figure 15. 1). In the course of the study, I attended a meeting with the British officials in England and other meetings in various locations in the United States, including Omaha, Nebraska, and Cleveland, Ohio.

 


[
138]

FIGURE 15.1.

FIGURE 15.1. The BAC-111 Transport Airplane.

 

The crash occurred when the airplane, which was flying at 5000 feet altitude (4000 feet above the terrain) and carrying 38 passengers and 4 crew members on a scheduled Braniff Airways, Inc. flight, burst into flames in the air and crashed in a field in a flat, upright attitude with no forward motion. As is usual in such crashes, there was little information available to determine the cause of the crash.

One of my first duties was to attempt to determine, from the locations of various pieces of wreckage around the crash site, the approximate altitude and airspeed of the sections at the time of breakup in the air. This kind of analysis requires calculating the trajectories of the various pieces as affected by their weight, drag, and surface area. In this case, the horizontal and vertical tail unit (a T-tail arrangement) and an outer wing panel were separated from the main wreckage. At that time, digital computers were available, but I had little experience in using them. I used a program called the Mimic program, which allowed the analysis to be conducted with the aid of a block diagram similar to that which would be used in making an analog computer solution and which had an unusually clear manual to describe its use. The solutions came out in good agreement with the expected altitude and speed of the airplane, which indicated that the airplane broke up in normal flight without warning. Later, the CAB representative requested copies of the computer program to use in their analyses of other crashes.

All possible causes of the crash were investigated, both by BAC and by the CAB. These causes included such things as failure of the control system or feel devices, use of improper construction materials, use of incorrect design procedures or requirements, and fire due to fuel or hydraulic leaks. A contributing factor involved in many studies was that the airplane was flying through or above a roll cloud when it burst into flames and was a few miles away from an approaching line of heavy thunderstorms.

The cockpit voice recorder was recovered, but the only information on it was various noises, such as a rushing sound of air following the breakup and a sounding of various warning horns during the descent. There was no crash recorder such as is now required on commercial airplanes to record airplane motions or control positions.

In listening to the record on the voice recorder, it was noted that following the breakup, there was noticeable variation in pitch or "wow" as it is called by audio enthusiasts. I conceived the idea that the records obtained might be used to get some idea of....

 


[
139]

FIGURE 15.2.

FIGURE 15.2. Rolling acceleration and rolling velocity of BAC-111 airplane as determined from voice recorder analysis.

 

....the motions of the airplane. The recorder had a flywheel in the tape drive that had its axis aligned with the longitudinal axis of the airplane. As a result, the speed of the flywheel would be affected by rolling accelerations of the airplane. A similar voice recorder was obtained for calibration tests in the NASA Instrument Research Division. The record obtained from the accident had a faint 800 hertz tone, probably a harmonic of the aircraft power supply, that formed an excellent time base. I laboriously analyzed the data, one cycle at a time, to determine the movement of the tape and was able to plot a record of rolling acceleration as a function of time. This record was then integrated to determine rolling velocity. The data showed an abnormally rapid rolling velocity to the left building up to over 100 degrees per second in less than I second following the start of the "rushing air" noise (figure 15.2). Though this record was interesting, it could not be relied on entirely because the tape speed was also affected by large values of linear acceleration.

To simulate the maneuvers following the breakup, engineers at the Langley 20-Foot Vertical Spin Tunnel had a crude model of the airplane built with the tail assembly and part of the left wing missing, as found on the wreckage (figure 15.3). This model was launched by a catapult from the top of the Langley Lunar Landing Facility (now Langley Impact Dynamics Research Facility), a girder structure 250-feet high, which for the scale of the model used, corresponded closely to an altitude of 4000 feet. The model would go into violent tumbling motion, but before hitting the ground would settle into a flat descent, sometimes without rotating and sometimes in a slow flat spin. The model always landed right side up, probably because of the wing dihedral.

Another interesting phase of the investigation was a number of lectures by a well-known meteorologist, Dr. Tetsuya Fujita of the University of Chicago, who discussed mesoscale turbulence and who had coined the term "downburst" to describe the downward rush of air cooled by heavy rain in a...

 


[
140]

FIGURE 15.3.

FIGURE 15.3. Model of damaged BAC-111 used for catapult tests to determine motion following breakup.

 

....thunderstorm. He pointed out that this air would spread out on the ground and would advance at a high speed several miles ahead of the main storm, which would result in roll clouds or in heavy clear-air turbulence.

The conclusion of the accident board was that there was no fault in the design or construction of the airplane and that the breakup had been caused by an abrupt gust with velocity well beyond the design requirements. Members of the BAC calculated that a gust of 144 feet per second angled at 45 degrees upward and 45 degrees from the right would have been required to break off the T-tail in the manner that occurred in the accident, whereas the design gust velocity at that time was 88 feet per second. Following loss of the tail, the airplane would immediately pitch down to large negative values of acceleration, which would cause the outer wing panel to break off.

The official accident investigation concluded that the airplane met all design requirements and that no changes in design were required. Warnings were issued concerning the possibility of high turbulence at low altitudes several miles from heavy thunderstorms. Subsequent experience showed that the conclusions were probably correct because no similar crash of a BAC-111 ever occurred. A complete copy of the National Safety Board report was published in Aviation Week (ref. 15.1).

I proposed a wind-tunnel study of turbulence of the type encountered in the accident, by using a large downward jet of air directed at the floor of a wind tunnel operating at very low speed. This study was never made. I have always wondered whether the airplane would have broken up flying anywhere into the storm front at the altitude and vicinity of the crash, or whether the gust was a single, isolated phenomenon with very small probability of being encountered by an airplane.

 

Crash of Convair B-58 Bomber

 

The Air Force started a program about 1952 to develop a supersonic bomber. The result was the Convair B-58 Hustler, a delta-wing aircraft with four engines mounted in nacelles under the wing (figure 15.4). The airplane was first flown in 1956. A large pod, almost as large as the fuselage itself, could....

 


[
141]

FIGURE 15.4.

FIGURE 15.4. B-58 Hustler bomber airplane.

 

....be carried under the fuselage. This pod was intended to be developed into a large nuclear weapon or a carrier for smaller munitions. The B-58 was a remarkable airplane in many ways in that it had much larger range, installed power, and load-carrying ability than any supersonic airplane developed before that time. By 1960, about 13 airplanes had been built. It is not surprising that, with such an advanced design, problems would be encountered in its development. By 1960, however, five airplanes had been destroyed in accidents, and the Air Force was engaged in a frantic effort to increase the safety of the vehicle and its operation.

Of the accidents that occurred, the first four involved such unrelated causes as a ground fire during fuel transfer, pilot error, a crash due to tire failure, and structural defects. The fifth accident, however, occurred in flight while an experienced test pilot was testing the engine control system. The airplane went to a large angle of sideslip and disintegrated in the air. This event might indicate a serious defect in the design. The Air Force set up their own investigating board and also set up a joint Air Force-industry board, of which I was a member.

Several of the board members were men I had worked with previously on the Society of Automotive Engineers (SAE) A-18 committee, known as the SAE Aerospace Control and Guidance Systems Committee. This committee was a remarkably informative organization of the top control engineers of various companies, equipment manufacturers, and universities. Meetings were held four times a year, usually at very attractive locations. The members had very few inhibitions on discussing their companies' latest projects because it was realized that a free interchange of ideas would stimulate development. Another feature that encouraged attendance was that no formal proceedings were published. Instead, each speaker brought enough copies of his presentation to a meeting so that each member could have one. As a result, the attendees immediately had a stack of useful information on the latest developments.

Two of the men prominent in the SAE committee who assumed leadership roles in the accident investigation were Duane McRuer and Irving Ashkenas, officials of a small company called STI (Systems Technology Incorporated) whose business was studying stability problems of airplanes. I was assigned the job of investigating the longitudinal stability, and Irving Ashkenas headed the study of lateral and directional stability. The Convair Corporation at Fort Worth, where the board met, furnished computing facilities to aid in the investigation.

[142] I was very impressed by the effectiveness of this small group of experts, who knew exactly how to proceed with investigating the problems encountered in the accident. I thought how desirable it would be to have such a capable group assigned as my own engineers for conducting the projects in my division. Of course, such an arrangement would not be possible because the group consisted of the top men from each company or organization, each one of whom normally supervised a group of less experienced engineers.

The proceedings of this board were classified secret. As a result, I was not allowed to take out any notes or documents. My account of this investigation is therefore given entirely from memory and may be subject to errors. I will attempt, however, to give a correct general impression of the investigation.

The emphasis, of course, was on the control system of the airplane. This control system had been designed by the Eclipse-Pioneer Division of the Bendix Corporation and incorporated many novel features. An attempt was made to place the various subsystems, such as the air data system, the gyroscope and accelerometer unit, the amplifier and computer assembly, the engine controls, and the autopilot into an integrated system. The various systems all came together in the power control linkage assembly or PCLA. This complex mechanism, which fit into a space about 2-feet wide by 3-feet long, contained all the equipment for receiving the pilot's inputs and the autopilot signals and provided control feel, ratio changing, commands to the control surfaces, and various safety and monitoring provisions. Usually, such functions are distributed throughout the airplane, and the complexity of this tightly packed unit must have caused a nightmare for maintenance. In fact, as I recall, the unit was intended to be returned to the factory for maintenance. In accordance with the state of the art in that time period, the system used all hydraulic power controls and all analog computation.

As it turned out, the lateral-directional system soon came under suspicion as the cause of the accident. All the quantities sensed by the autopilot were fed to the control system by a control law that required changing gains as a function of flight condition. These gains were varied by servo-driven potentiometers, or pots, in a manner similar to that used in analog computers of the time. Investigation of the wreckage showed that one of the pots in the autopilot was against its stop, which indicated that the electric servo driving the pot had probably experienced a hard-over signal. Normally, it would be expected that such a failure would change the handling of the airplane somewhat, but would not cause a catastrophic failure. In this case, however, when the pot was driven against its stop, the wiper on the pot ran off the end of the resistance element and produced zero output. As a result, the gain of this particular feedback signal was zero. Stability calculations showed that with this feedback gain at zero and the others at their normal values, the airplane would have a highly unstable lateral, or Dutch roll oscillation. Furthermore, with this particular combination of gains, the normal control actions of the pilot in attempting to keep the wings level would further destabilize the oscillation. It was therefore concluded that when the failure occurred, the airplane went into a divergent lateral oscillation and reached sideslip angles sufficient to break off the tail.

The findings of the board were undoubtedly transmitted to the proper officials in the Air Force, and correction of the problem might involve simply limiting the travel of the servos driving the pots to avoid exceeding the ranges required in flight. This accident, however, illustrates the problems that caused great reluctance of airplane manufacturers to adopt electronic controls for airplanes and that required 20 years or more before commercial airplane manufacturers would use such devices in primary control systems, despite the improved capabilities and lower cost and weight of these systems.

[143] For many years, the design of each new airplane had been supervised by a chief engineer, who was sufficiently familiar with all elements of airplane design and construction that he could understand the operation of these systems and could bring his long experience to bear in avoiding dangerous features. As soon as electronic control systems were introduced, the operation of these systems was beyond the knowledge of the chief engineer. In the case of the B-58, a company skilled in autopilot design was given responsibility for the design of the control system. The electrical and electronic engineers, however, did not have long experience with the safety problems of airplanes or of the possible catastrophic effects of a very minor failure. In my experience in flight research, I worked with airborne radar equipment that employed several hundred vacuum tubes. Usually two or more tubes would have to be replaced after a few flights, yet the designers of this system considered it to be highly reliable. Later, more extensive studies of reliability requirements for control systems were made. The basis of these studies were that no failure should occur, not in the lifetime of an airplane, but in the lifetime of the entire fleet of airplanes of a given design. This requirement amounts to a time between failures of the order of 109 flight hours. The meeting of such requirements is still a major problem, but with redundant systems and digital control equipment, these reliability goals can be met. Just the realization on the part of the designers of the importance of this degree of reliability represents a great advance in attaining the desired degree of safety.

Following the accident investigation, B-58 bombers continued to have crashes. Accident number 6 was a tail failure. Number 7 was an interesting case of an unanticipated problem resulting from the novel design of the flight control system. A feature of the longitudinal control system was that the airplane was automatically in trim at each value of airspeed. This feature is desirable on supersonic airplanes because large trim changes often occur in going through the transonic speed range. Of course, this feature means that the airplane has no speed stability. Normally, a statically stable airplane without an autopilot can be trimmed to fly at a given airspeed and, because of the stable variation of pitching moment with angle of attack, it will tend to return to the angle of attack corresponding to the trim condition. The B-58 did not have this capability, but this problem was overcome by use of the autopilot system to automatically hold any desired value of airspeed.

One of the requirements in Gilruth's requirements for satisfactory handling qualities is that an airplane should require an increase in pull force on the control stick to reduce the speed to the stall. This feature is stated to provide a valuable form of stall warning. In the case of the B-58 in a flight in icing conditions, the airspeed head iced up so that the pilot had no correct instrument indication of airspeed. Also, a delta-wing airplane like the B-58 has a gradual stall with no buffeting until large values of angle of attack are reached. The result was that the B-58 gradually slowed down without the pilot's knowledge until it stalled and fell out of control. This experience shows that the provisions of Gilruth's flying qualities requirements should be kept in mind when devising new control systems. If the new systems fail to provide these features, some modifications should be made to provide an equal degree of safety.

Eventually, the problems of the B-58 were worked out and pilot error was decreased with the introduction of training versions of the airplane. In all, over 80 of these airplanes were manufactured. They never saw service in war because the expense of operating them was found to be very large when compared with that of subsonic bombers like the B-52. The development of the weapons pod was dropped, but the airplanes were used for high-altitude reconnaissance and provided valuable surveillance information in the Cuban missile crisis and in a volcanic eruption in Alaska.

 

[144] Accidents of the Republic F-105 Thunderchief Airplane

 

A number of cases had occurred in which the Republic F- 105 airplane, following a pull out from a dive-bombing run, experienced a large amplitude pitching oscillation. In one case at Nellis Air Force Base, three airplanes were performing rocket missions in a race track pattern. One of the airplanes made a pull up at 500 feet altitude, and a series of three longitudinal oscillations was observed with accelerations of 7g and -1g. The following airplane in the pattern, after pulling out at 300 feet altitude, experienced a similar oscillation. At the end of the third cycle, the airplane went to a large pitch attitude and crashed.

I made a trip with an instrumentation engineer, R. H. Sproull, to visit the Republic Aircraft Corporation at Farmingdale, Long Island, New York, and then to the Aeronautical Systems Division, Wright-Patterson Air Force Base, to discuss the accidents. Prior to the trip, the Republic Corporation had already made modifications to the control system and had run flight tests at Eglin Field to study the problem. The main problems facing the company were whether to unground the airplane and whether to place restrictions on maneuvers.

The stabilizer control system incorporated a pitch damper in series with the stabilizer actuator. To protect against pitch damper failure, the pitch damper was originally designed to cut off automatically if the acceleration exceeded 4g or -1G, followed by a pitch damper actuator deflection of greater than 90 percent full travel in the direction to extend the g value beyond these limits. The pitch damper could be caused to disengage, therefore, by a pull up exceeding 4g, followed by a rapid push down which caused the damper to oppose the negative pitching velocity.

The airplane also had very light stick forces and low longitudinal stability in the conditions existing in dive pull outs. Such conditions are known to be favorable to pilot-induced oscillations. The control feel was considered desirable by pilots, however, and it was generally considered that the problem of oscillations could be avoided by pilot training. The Republic Corporation had realized that a pitch damper disconnect in the middle of a maneuver might trigger a pilot-induced oscillation. The pitch damper system had therefore been modified to prevent pitch damper disconnect except in case of an actual failure of the pitch damper. Though details of this modification are not available, the change involved installing a duplicate pitch gyroscope and a monitor circuit.

In examining the data from tests of the modified system, the NASA representatives noted some inconsistencies between the measured stick forces and the movement of the control stick. These data, if correct, could indicate a malfunction of the control system. This discovery caused some consternation among the Republic engineers, but it was soon realized that the data might be incorrect if up and down forces on the control stick could affect the strain gauge instrument that measured stick force. Such forces might be applied by the pilot in pull-up and push-down maneuvers to prevent his head from bouncing against the canopy. Tests were made that confirmed this source of error.

The final recommendation by the NASA representatives included removing the grounding restrictions on the airplane, but adding restrictions on dive angle and g value level during pull outs. This accident investigation reemphasized the knowledge that has been gained from numerous cases, namely, that stabilizing devices should not be disconnected as a safety measure, particularly in situations where the airplane is performing violent maneuvers.

 

[145] Analytical Study of the F-84 Following Accident in High-Speed Flyby

 

Several airplanes have broken apart in the air during high-speed flybys. One of these was the Republic F-84 Thunderjet, an early unswept-wing jet airplane with tip tanks. Such an accident, which is tragic enough in any case, is doubly embarrassing to the company because it occurs during a demonstration of the airplane to high officials of the government or the Air Force. Movies of the crash seem to show that the airplane exploded in the air, but a frame-by-frame analysis of the movie shows that the airplane actually pulled up very rapidly to a high value of normal acceleration, about 8g, before the breakup occurred.

I was curious about the cause of this and other similar crashes. Though I was not associated with the accident investigation board, I made a brief analytical study of the stability characteristics that might contribute to a crash of this type. The study was never published, but it resulted in a memorandum for the Chief of Research dated October 11, 1948, with the title "Effect of Approach to the Wing Divergence Speed on the Longitudinal Stability of an Airplane."

Wing divergence is a phenomenon caused by the lift on a wing acting at a point ahead of the wing torsional axis. The torsional axis is defined as the axis location at which a lift force will produce no twist of the wing. Ordinarily, unswept wings have an aerodynamic center, or point of application of the lift force due to change in angle of attack, at about 25 percent of the chord, but this point would be still further forward with tip tanks. The torsional axis is usually at 35 to 45 percent of the chord. During a maneuver, the wing therefore twists in a direction to increase the lift and torsional moment above those of a rigid wing. If the structural restoring moment exceeds the destabilizing effect of the air forces, the wing behaves normally.

Beyond some value of dynamic pressure, however, the wing becomes unstable and diverges until the structural strength is exceeded.

The dynamic pressure for divergence is defined as the dynamic pressure at which the wing diverges when the wing root is rigidly fixed. Airplanes are flown so that this dynamic pressure is not exceeded in normal operations. The fuselage of an airplane in flight, however, does not provide a fixed restraint because of the change in angle of attack of the fuselage during a maneuver. Some effects on the motion of the airplane might therefore be expected at speeds below the divergence speed.

Equations for the longitudinal motion of the airplane were set up with the assumption that in addition to the usual degrees of freedom (vertical motion and angle of attack) a third degree of freedom representing wing torsion was included. The twist angle, of course, varies with the position along the wing span. To account for this effect, a technique known as the semirigid method, described in some British reports, was used (ref. 15.2). A variation of twist angle along the span was assumed, as well as a distribution of torsional moment of inertia. By use of a formulation of the equations of motion called Lagrange's equations, which integrates these effects, the torsional mode may be specified by a single variable representing, for example, the twist angle at the tip.

For a complete analysis of this problem, wing bending should also be included, but it was reasoned that wing bending would have less effect on the motion because wing bending does not directly affect the pitching moments acting on the airplane. Because of the lack of computational facilities at the time of this study, it was necessary to limit the number of variables. With the variables assumed, the resulting motion consisted of the short-period rigid-body motion and the torsional mode of the wing. With the assumption of constant airspeed, the analysis therefore required the solution of a fourth [146] degree equation, for which as mentioned previously, solution techniques were available.

The airplane characteristics assumed did not represent any given airplane, but were representative of a high-speed fighter airplane. The wing divergence speed was assumed to correspond to Mach number of 0.9 at sea level. The center of gravity was assumed to be at 31 percent of the mean aerodynamic chord (MAC), which gave a stability margin in maneuvers at low speed of 2.5 percent MAC. The point of application of lift due to wing torsion was at 25 percent MAC. The results showed that when the wing torsion was taken into account, the airplane became longitudinally unstable in maneuvers above a Mach number at sea level of 0.61, and at a Mach number of 0.8 the time for the longitudinal motion to double amplitude was less than I second.

A point of interest is that most flight tests of high-speed airplanes are made at altitudes greater than 15,000 feet. Often, a high-speed low-altitude pass to demonstrate the airplane is the first time that conditions of maximum dynamic pressure have been encountered. This circumstance may account for the number of accidents that have occurred in such demonstrations.

Problems of the type studied were encountered on the Republic F-84 and the Northrop F-89 Scorpion, both of which had large tip tanks. The cure in the case of the F-84 was to put large triangular fins on the outside rear of the tanks, thereby adding a stabilizing effect to the torsional moments acting on the wing, and as an added benefit, reducing the induced drag of the airplane. This modification was installed on the F-84E and later models. The success of such empirical fixes probably accounts for the lack of interest in more detailed analytical studies of this problem.

A nonlinear structural phenomenon also was evident on many airplanes of this period that resulted in abrupt, unexpected incidents of both wing flutter or airplane instability. The wings of airplanes of this period where relatively thick with thin-skinned monocoque construction. These wings showed adequate torsional stiffness when tested on the ground. When the wing was subject to high loads in flight, however, the skin wrinkled extensively, which resulted in a loss of torsional rigidity. In a high-g-value maneuver, the airplane could therefore be placed instantly in a very unstable condition that was not evident in flight preceding the maneuver. Fortunately, these problems have largely disappeared on modern airplanes that have thinner wings and thicker wing skin.

 

Accidents of the Beech Bonanza Involving In-Flight Structural Failures

 

Chapter 10 contained a discussion of the problem of airplane crashes resulting from spiral instability in flight under instrument conditions. This type of crash usually resulted from a pilot with insufficient experience in instrument flight becoming disoriented and entering a spiral dive. With early types of light airplanes, the airplane often was destroyed by crashing into the ground. If the airplane emerged from the clouds with sufficient altitude, the pilot could usually regain his orientation and recover. In the case of more modern personal airplanes with higher wing loading and lower drag than earlier types, excessive speed builds up rapidly in a spiral dive, which sometimes results in structural failure in the air regardless of the visibility conditions.

Accidents of this type involving in-flight structural failure have been encountered with all types of high-performance personal airplanes, but one type, the Beech Bonanza, had an accident rate much higher than other similar airplanes. This airplane, with a distinctive V-tail, was first introduced in 1947 (figure 15.5). The company stopped building the V-tail model in 1982 after 10,405 had been produced. By 1986 there had been.....

 


[
147]

FIGURE 15.5.

FIGURE 15.5. Beech B-35 Bonanza airplane as tested at Langley.

 

....232 reported in-flight structural failures of these airplanes, which resulted in deaths of over 500 people. Already in 1950 the high accident rate of this airplane had become apparent. As a result, the Civil Aeronautics Administration (CAA) requested that the NACA conduct flight tests on the Bonanza to try to determine any unusual characteristics that could contribute to the accidents. This investigation was conducted by personnel in my branch and a report was published (ref. 15.3).

The flight tests of the Bonanza were conducted by NACA test pilots and, in accordance with safety procedures followed in all NACA flight tests, were limited to the placard values of airspeed and acceleration, which were an indicated airspeed of 200 miles per hour and a normal acceleration of 4g. Within these limits, no unusual or dangerous characteristics were observed. Special tests were made in which the pilot in a hooded cockpit was given control of the airplane in inverted and other unusual attitudes and was required to recover with the most elementary flight instruments, the so-called needle-ball-airspeed group. A skilled pilot encountered no difficulty in making these recoveries. The pilots did note that because of the clean design of the airplane, speed could build up beyond the placard limit rather rapidly if the airplane were nosed down.

Accident data showed that many of the inflight breakups were encountered at values of airspeed well beyond the placard limit. Any such accidents are not considered the fault of the designer, because the pilot is never supposed to exceed the placard limit. Furthermore, the CAA design rules did not specify any design requirements for safety or handling qualities beyond the placard limits.

In studying the flight data, I observed that the curve of control force against dynamic pressure in straight flight had a downward curvature in the direction of increasing push forces with increasing speed. This plot, for a rigid airplane, should be a straight line. Various types of aeroelastic distortion can result in the plot deviating from a straight line. The Bonanza, in common with some other light planes, had a large negative setting of the stabilizer to allow the nose to be raised on takeoff at a forward center-of-gravity location. This stabilizer setting resulted in a down-elevator deflection for trim in cruising or high-speed flight. Such an elevator deflection would cause the stabilizer to twist by an amount that increased directly with the dynamic pressure. In a V-tail design, the twist would be increased when compared with a conventional stabilizer because of the....

 


[
148]

FIGURE 15.6.

FIGURE 15.6. Measured and extrapolated stick force in straight flight as a function of dynamic pressure for Beech B-35 Bonanza airplane.

 

....increased length of the tail panels. With a rigid stabilizer, the curve of control force against indicated airspeed would be a parabolic curve, which varied as the square of the speed, because the dynamic pressure varies as the square of the speed. The effect of the stabilizer twist would be to add to this curve an amount varying as the fourth power of the speed. The result of this addition would be to make the push force required for trim in 1g. flight increase rather rapidly at high values of airspeed.

The plot showing experimental values of control force versus dynamic pressure for two values of trim speed is shown on figure 15.6. Because of the limitation of the flight data to 200 miles per hour, extrapolation of the data to higher values of airspeed is difficult. Nevertheless, two extrapolated curves are shown. From the upper extrapolated curve, the one less affected by stabilizer twist, two plots of control force versus indicated airspeed for various values of normal acceleration are shown in figures 15.7 and 15.8. The values of control force are little affected by bank angle or pitch angle. These curves therefore apply during a spiral dive as well as in straight flight.

The conclusion from this analysis is that the airplane has a great tendency to pull up to higher values of normal acceleration if the airspeed increases beyond the placard speed in a spiral dive. For example, if the pilot holds zero control force, the acceleration at 280 miles per hour would be 6g with a trim speed of 145 miles per hour or 8.5g with a trim speed of 120 miles per hour. Normally, if the pilot becomes disoriented in a spiral dive, he will exert a pull force on the control wheel, further increasing the loads on the airplane. It is considered very unlikely that he would exert a push force to reduce the loads on the airplane. The results therefore indicate that structural failure might result because of the normal acceleration exceeding the design strength of the wing or of the tail. These results are based entirely on analytical extrapolation of the data obtained within the placard limits. No actual tests have been.....

 


[
149]

FIGURE 15.7.

FIGURE 15.7. Estimated stick force as a function of indicated airspeed and normal acceleration for Beech B-35 Bonanza airplane. Trim speed 120 miles per hour. Crosshatched area indicates limits of test data.

 


FIGURE 15.8.

FIGURE 15.8. Estimated stick force as a function of indicated airspeed and normal acceleration for Beech B-35 Bonanza airplane. Trim speed 145 miles per hour. Crosshatched area indicates limits of test data.

 

[150] ...made to study these problems in the speed range beyond the placard limits.

One cause of the increased accident rate of the Beech Bonanza may therefore be the unusual tendency of the airplane to pull up in flight at high speed beyond the placard limit. Though no design requirements are specified for this regime of flight, a large improvement in safety might be attained with very little penalty in weight or performance by considering the handling qualities in this regime of flight. Most modern airplanes would not suffer catastrophic failure in a high-speed dive unless the acceleration increased beyond the structural limits.

At the time these studies were made, they were discussed with members of the Beech Corporation and of the CAA. No immediate action was taken, because of the philosophy that consideration of handling qualities beyond the placard limit was not required. In later years, models of the Bonanza with increased power and gross weight were produced, and some increases in the structural strength of the stabilizer were made by the Beech Corporation. Accidents continued, however, until in 1984, an extensive study was made by a task force at the Transportation Systems Center of the U.S. Department of Transportation. The purpose of this study was to assess the facts associated with the controversy that had arisen regarding the design and certification process for the Bonanza. At this time, I transmitted the studies made by the NACA in 1950 to the director of the study. The results of the study by the Transportation Systems Center are contained in a comprehensive report (ref. 15.4).

Despite these extensive efforts, no change in design philosophy or certification procedure has been made. Though the final report of the Bonanza task force contains considerable study to find any case of failure within the placard limits, no mention is made of the possible failures beyond the placard limits, and no consideration is given to the idea of improving the characteristics in this range. Production of the V-tail Bonanza was discontinued in 1982 and later models of this airplane with conventional tail surfaces have shown much reduced accident rates. With the present lack of any effort to give consideration to characteristics beyond the placard limits, I believe that problems of this type may occur with future designs and that the lessons learned from the Bonanza should not be forgotten.


previousindexnext