Handling Qualities of a Blended Wing Body Aircraft

Handling Qualities of a Blended Wing Body Aircraft Timothy Shaw Peterson Master of Applied Science Institute for Aerospace Studies University of Toronto 2011 The blended wing body (BWB) is a tailless aircraft with the potential to use 27% less fuel than a conventional aircraft with the same passenger capacity and range. The primary purpose of the current study was to determine the handling qualities of the BWB, using piloted-handling trials in a moving-base simulator. The secondary purpose was to determine the effect of simulator motion on handling-quality ratings. De Castro conducted piloted-handling trials in a fixed-base simulator. De Castro’s tasks and flight model were modified in the current study. In the current study, three subjects rated the handling qualities as Level 1 or 2, depending on the task. Simulator motion did not have a significant effect on the results.


Contents
C l , C m , C n Coefficients of rolling, pitching, and yawing moments, body frame C T Coefficient of thrust C x , C y , C z Coefficients of forces along x, y, and z, body axes Coefficient of drag at zero lift C D δLG Coefficient of drag due to landing gear C i j k 2 nd partial derivative of coefficient of force or moment i, ∂ 2 C i ∂j∂k C i j Derivative of coefficient of force or moment i with respect to j, dC i dj

Literature Review
Handling qualities are defined by Cooper and Harper [5] as "those qualities or characteristics of an aircraft that govern the ease and precision with which a pilot is able to perform the tasks required in support of an aircraft role". Cooper and Harper developed a ten-point scale that is used to rate handling qualities. After performing some flight task, the pilot assigns a numerical Cooper Harper rating to the task based on the char-acteristics of the aircraft and the demands on the pilot. Handling qualities are affected by the cockpit interface, the aircraft environment, and pilot stress [5]. The military handbook MIL-HDBK-1797 [6] uses the term "flying qualities" somewhat interchangeably with "handling qualities". In the current study, only the term "handling qualities" is used.
In 2000, Cameron and Princen wrote about piloted-handling trials of a BWB commercial airplane that were conducted in the NASA Langley Visual Motion Simulator [7].
However, the results of these tests are not widely available. De Castro's handling trials [8] were conducted in a fixed-based simulator, without motion cues. Therefore, the subjects would have relied primarily on visual cues. Showal- Grantham [10] compared handling-quality ratings obtained in an in-flight simulator to those obtained in a ground-based simulator, the Langley Visual/Motion Simulator.
The aircraft that was modeled was a jet-transport airplane, with 31 configurations of longitudinal flight control systems. For those configurations that had a pitch-rate command response, the Cooper Harper ratings of the landing-flare task were generally lower in the in-flight simulator than they were in the ground-based simulator; handling-quality ratings obtained in a ground-based simulator tended to be conservative. The majority (81%) of the 31 configurations had ground-based ratings that were within ±1 Cooper Harper ratings of the corresponding in-flight ratings. Grantham also found that performance tended to be worse in the ground-based simulator than in the in-flight simulator.
Sink rates at touchdown tended to be approximately 3.5 ft/s greater in the ground-based simulator. While that study supports using ground-based simulation to obtain HQRs of an aircraft, it does not provide guidelines for the motion requirements of the ground-based simulator.
Soparkar and Reid [11] used a model of a generic jet trainer to determine the effect of simulator motion on HQRs. They found that pilots preferred simulator motion with low washout levels to the no-motion case. The task was lateral tracking under gust upsets applied to the roll axis. Schroeder [12]

Purpose
The primary purpose of the current research is to determine the handling qualities of a BWB aircraft, using piloted-handling trials in a moving-base simulator. The secondary purpose is to determine the effect that simulator motion has on the HQRs.

Equations of Motion
The rigid-body equations of motion that are implemented in the flight model are taken from Etkin [14]. Euler's equations of motion are rearranged to solve for the derivatives of translational and angular velocity. The derivatives are then integrated. The translational velocity V E B is the velocity relative to an earth-fixed frame expressed in a body axis. The earth-fixed frame, F E , is taken to be an inertial frame. Rotation and curvature of the earth are neglected. The translational equation of motion is taken as, where f B is the sum of the external forces, m is the aircraft mass, andω B is the skewsymmetric matrix of angular velocity expressed in body axes. The previous equation is The rotational equation of motion is also expressed in the body frame as, where G is the sum of all moments on the BWB. The angular momentum is given as and therefore the derivative of the angular momentum relative to the body frame, expressed in the body frame, isḣ where I B is the inertia matrix. The BWB is treated as a rigid body, sȯ The principal axes of the BWB are assumed to be coincident with the body axes, so

Aerodynamic Model
The aerodynamic forces and moments are calculated in stability axes. The stability axes are chosen so that the x -axis is aligned with V in a reference condition of steady, symmetric flight [14]. The forces and moments are then transformed to the body axes and summed with other forces and moments, such as those from the engine and ground contact.
The non-dimensional force and moment coefficients are first calculated and are then multiplied by the dynamic pressure and a reference area or volume to get the actual forces and moments. The following equations, from de Castro [8], give the force and moment coefficients, in the absence of ground effect:
The controls surfaces are the elevons and winglet rudders, which are grouped together to act like the elevator, ailerons, and rudder on a conventional aircraft. The aerodynamic control deflections are δ e , δ a , and δ r . into body axes as follows, where b is the wingspan, c is the mean aerodynamic chord, S is the trapezoidal wing area, q is the dynamic pressure, and L BS is the transformation matrix from the stability axes to body axes.

Ground Effect
The airflow around an aircraft is altered as the aircraft approaches the ground. Since the tasks of the piloted-handling trials in the current study include landings and takeoffs, ground effect is expected to occur. Etkin [15] states that, for conventional aircraft, significant changes in trim and stability take place due to ground effect. In ground effect the downwash at the tail of a conventional aircraft is reduced, the wing-body lift-curve slope is increased, and the tail lift-curve slope is increased. Clearly, tail downwash and tail lift are not applicable to the BWB, but the change in wing-body lift-curve slope may be significant.
During the piloted-handling trials, de Castro [8] found that the aircraft could not be rotated during takeoff at 180 kts (92.6 m/s). As a solution, the engine height was lowered to be coincident with the CG. Rotation at 180 kts might have been possible if ground effect increased the lift slope significantly or gave a positive pitching moment.
The ground effect model that was used by de Castro only changed the lift slope, C zα , and did not change the pitching moment. For small angles of attack, C zα ≈ −C Lα . The change in C zα with proximity to the ground from de Castro is given as, Staelens' results were used in the current research, with the belly flap retracted (δ bf = 0).
Ground effect was forced to be zero at a nondimensional heightĥ ≥ 1. In ground effect, the total drag and lift are found by adding to Equations 2.9 and 2.11 the following increments, Staelens [16] found the ground effect for nondimensionalized heights from 1 down to 0.057. However, Staelens only compared the polynomial functions for lift, drag, and pitching moment to the wind-tunnel measurements down to a nondimensional height of 0.115. This was the lower limit used in the current study. Ground effect is greater in Staelens' model than in de Castro's model [8]; at a nondimensional height of 0.115, the change in lift slope of the former is 0.96 rad −1 while that of the latter is 0.59 rad −1 .
The total pitching moment in ground effect is found by adding to Equation 2.13 the following increment, The ground effect on pitching moment includes the difference in static margins between the configurations of Staelens and de Castro. In the former, the static margin k ns = 0.105 while in the latter, k n d = −0.13.
gives the following equations: The coefficients a i are given in Table 2.1.  In the current study, the inclusion of ∆C m due to ground effect was sufficient to allow rotation at 180 kts, with the engines 2.8 m above the CG. While adding the pitching moment and lift increment due to ground effect helps achieve rotation, it changes the trimmed condition on the approach to landing. At an approach speed of 180 kts, ground effect lowers the trimmed angle of attack by 3.2 • and increases the elevon to trim by 4 • over the free-air case. Figure 2.2 shows the changes to trimmed angle of attack due to ground effect. The graphs are for the approach-to-landing configuration; the landing gear is extended and the aircraft is descending on the glideslope at 3 • . The impact of ground effect on longitudinal handling qualities is discussed in Chapter 3.

Landing Gear Struts
The as a starting point [17]. The front bogey was located 25.4 m in front of the centre of gravity (CG). The body bogeys were located 2.82 m behind the CG, and 5.5 m laterally from the CG. The distance from the CG to the ground was 4 m, when the landing gear struts are fully extended. The landing gear struts were modeled with non-linear springs and oleo-pneumatic dampers. The strut force for a single bogey is given by: where K 1 is the nonlinear spring force, and K 2 and K 3 are the non-linear and linear damping coefficients, respectively and is the strut compression.
The spring forces and non-linear damping coefficients are given in Table 2 determined assuming a strut stroke max = 0.673 m and a load factor n = 2.485 [18].
The shape of the spring force-deflection curve was kept the same as the Boeing 747.
However, the spring stiffness was increased to achieve the desired spring endpoint stiffness for the given load factor and strut stroke. The endpoint stiffness is given by the following equation, assuming equal loading of both main landing gear struts: (2.29) = 371280 kg · 9.81 m/s 2 · 2.485 2 · 0.673 m = 6.72e6 N/m The endpoint spring stiffness used in the flight model was slightly higher than this, 6.97e6 N/m. The front (or "nose") landing gear strut is designed to carry 40% of the force of one of the main landing gear struts. The required front landing gear strut endpoint stiffness is given as, The endpoint spring stiffness of the front strut used in the flight model was 4.04e6 N/m, considerably larger than what is recommended by Chester's [18] method.

Tires
The ground model calculates the forces acting on the landing gear tires. The tire model calculates the wheel's angular velocity, longitudinal and lateral slip, and normal deflection. The 2 front bogey tires and 10 main bogey tires are each treated as one "homogenous" or equivalent tire. The normal load on the equivalent tire is the strut normal load divided by the number of tires. The total force acting on that strut is found by multiplying the forces on the equivalent tire by the number of tires. The tire parameters are given in Table 2.3.
where λ is the longitudinal tire slip. Clover [19] describes a basic algorithm for implementing the wheel slip and wheel spin equations. The algorithm includes a step that multiplies ∂F x /∂λ and F x 0 by the sign of the wheel longitudinal velocity, u. When this step was added to the algorithm of Cheung [17], the instability whereby the aircraft accelerated backwards no longer occurred. One other change was made to the tire model.
When the tire is not in contact with the ground, the angular velocity of the wheel is set to 0.

Engine Model
The engines in the BWB flight model are the same as the engines in an existing Boeing calculates thrust as a function of air temperature, Mach number, altitude, and powerlever angle. Reid and Nahon model the lag of the engine response with a lag filter, which is given as, where EPR 1 is the commanded engine pressure ratio and EPR 2 is the actual EPR. The break frequency of the filter is a function of the compressor first-stage speed (N 1 ), and is given as, At N 1 = 50%, the filter time constant is 2.9 s. The Boeing 747 model has four JT9D-3 engines, while the BWB flight model has three. In the current study, the thrust of each JT9D-3 engine was multiplied by 2.2. The engines in the current study, being larger than the Boeing 747's JT9D-3 engines, would probably have a greater lag than the JT9D-3 engine. However, the break frequency of the low-pass filter was not changed in the current study.
De Castro [8] gives an engine static thrust at sea level of 418.4 kN. The static thrust at sea level of the engines in the current study is 460 kN. At sea level, the thrust-to-weight ratio is 0.38. Figure 2.3 shows the maximum thrust per engine as a function of airspeed.
The thrust decreases by less than 1% at an altitude of 5000 ft, the maximum altitude used in the current piloted-handling trials.  Table 2.4. The distance along the y body axis from the CG was given in Appendix D of de Castro's study, whereas the distance along the z body axis was measured from a diagram in that study.

Wind and Turbulence Model
The flight model included a wind and turbulence model. The turbulence model came from an existing Boeing 747 flight model, and was based on work by Reid and Nahon [20] and Robinson [21]. The wind and turbulence model in the current study consisted of a constant-direction mean wind that increased with height above ground, and patchy turbulent gusts. The turbulent gusts had 5 components, aligned with the body axes of the aircraft: u g , v g , w g , p g , and r g . Wind and turbulence were included in the pilotedhandling trials for two reasons: to increase the pilot's workload and to provide "ruse" motion for the reduced-simulator-motion condition. The mean wind at any height above ground level was a function of the mean wind at a height of 10 m above the ground U 10 and a ground roughness constant, ν, and was given as, A ground roughness constant of 0.19 was used, which is suitable for flat grasslands. The mean wind speed at 10 m above ground was set to 2 m/s. The turbulence intensity in the flight model would probably be classified as light turbulence or light chop. Selection of U 10 was constrained by the heave capability of the simulator. During the early stages of model development, a mean wind speed U 10 of 5 m/s was used. However, it was found that this speed was too high, such that the simulator actuator length was reaching the limits. For this reason the mean wind speed U 10 was reduced to 2 m/s. The wind direction varied between tasks, and is given in Section 5.1.
The five turbulent gusts (u g , v g , w g , p g , and r g ) are obtained by filtering white noise [20]. The spectral properties of the turbulence are described by the Dryden spectra, which are given as, The same equation is used for v g and w g . Non-Gaussian characteristics are obtained by modifying the kurtosis of the probability density function. The kurtosis of a Gaussian distribution is 3. Using a higher kurtosis increases the probability of low and high gusts while decreasing the probability of mid-range gusts. Reid and Nahon [20] experimented with several values of kurtosis and found that a value of 8.4 produced turbulence that was acceptable. This was the value of kurtosis used in the current study. The patchiness of the turbulence can be adjusted with a patchiness parameter R, which can range from 0 to 1. The length of a patch of turbulence increases as R decreases. Reid and Nahon used test flights of a 747 in the UTIAS flight-research simulator (FRS) and found that R = 0.1 gave realistic-feeling turbulence. The same value of R was used in the current study.
The lengthscales and intensities of u g , v g , and w g are functions of the height above while the lengthscales for the rotational, or asymmetric, gusts are given by, where z o is a surface roughness parameter, 0.1 for open grassland.
Above the gradient height (h ≥ 661 m), the intensities are given by, and the lengthscales above the gradient height are given by, The rolling and yawing gusts, p g and r g are found by an "equivalent one-dimensional power spectrum". The equivalent function integrates the asymmetric contribution from the variation of the u or w gust along the y axis. Gerlach and Baarspul [22] describe how to calculate the rolling and yawing gusts, assuming an elementary, two-dimensional flow field as depicted in Figure 2.4. The reference frame O e X e Y e Z e is earth-fixed.
Gerlach and Baarspul [22] give the equivalent one-dimensional spectral density function for the yawing gust r g due to variations in u g along the y e axis. An approximate  [22] expression is also given, and it is this approximation that was implemented in the flight model. The approximation to Iû g is given as, .53 uses graphs for Iû g (0, B), τ 1 , τ 2 , and τ 3 , which are functions of B [22]. The implementation of Equation 2.53 in the current research uses the asymmetric lengthscales L ug asym and L wg asym . The circular frequency of the turbulence ω t is related to the aircraft velocity V and turbulence spatial frequency Ω by: Similar to the yawing gust, the rolling gust p g due to variations in w g along the y e axis is described by the following spectral equation:

Predicted Handling Qualities
Bailey and Knotts [23] conducted in-flight experiments to determine the effect of the feel system on the lateral handling qualities of a fighter aircraft. They found that the feel system has an effect on handling qualities, distinct from the effect that the aircraft dynamics have on handling qualities. Part of the reason for the unique effect of the feel system is that the pilot senses both the force and displacement of an inceptor; Bailey and Knotts found that the feel system has an effect on handling qualities whether it is a force-sensing type or a displacement-sensing type. Mitchell et al. [24] recommended that changes be made to MIL-STD-1797A, to replace handling-qualities criteria based on flight phases with criteria based on mission-task-elements. They also recommended that the dynamics of the feel system be taken into account in handling-qualities analyses, because control force, not position, is the important reference for handling qualities.
However, their recommendations were not implemented in the handbook MIL-HDBK-1797A [6]. The handbook considers the feel-system dynamics for force-sensing inceptors, and excludes the feel-system dynamics for position-sensing inceptors. In the current study the inceptor dynamics are described and their effect on handling qualities is considered.

Inceptor Dynamics
The dynamic characteristics of the simulated BWB controls are described in this section.
The inceptor dynamics are then used to determine the expected handling qualities. Aside from control sensitivity (static gain), the dynamic characteristics of the controls used in de Castro's handling trials [8] were not given.
An Opinicus control loader provides the forces on the FRS controls. The control loader is a model-following system. Force transducers on the controls measure the net force acting on each control. The measured force is corrected for the gravitational force on the control, which depends on the control position. This corrected force is used as an approximation to the force exerted on the control by the pilot. By using a model A broad range of inceptor characteristics can be modeled. Though it was not done in the current study, the control forces can be made to vary with dynamic pressureq. In the current study, the controls were modeled as second-order, spring-mass-damper systems.
The differential equations governing the motion of the controls are given as, where The wheel radius is 0.165 m.

Longitudinal Flight Control System
The flight control system used in the Cranfield University piloted-handling trials [8] was a "Pitch-Rate-Command-Attitude-Hold" (PRCAH) controller. An input from the pilot, through the control column, produces a certain pitch rate. This is in contrast to a conventional response type, which typically produces a time-varying pitch rate for a certain column input [24]. A PRCAH FCS was used by de Castro to reduce the pilot's workload on approach to landing. During approach to landing, pitch angle is a good indicator of glide path angle. A PRCAH FCS would enable the pilot to select a pitch angle and have the FCS maintain it, without the pilot having to apply continuous control inputs. The PRCAH feedback structure is shown in  The BWB model used in the Cranfield piloted-handling trials had a negative static margin of -13% of the mean aerodynamic chord. To overcome this instability, angleof-attack feedback was used for stability augmentation. Pitch-rate feedback increased the short-period-mode damping. In order to maintain a constant pitch attitude with no control input (by eliminating the steady-state pitch-rate error), an integrator was used.
A disadvantage of the pitch-rate-command-attitude-hold response is that the pitch angle is held constant, even if airspeed decreases and angle of attack approaches the stall angle.
Therefore, some type of speed protection is desirable. While de Castro's FCS did not include a speed controller, Rahman [25] developed a speed controller for the BWB. A speed controller was not implemented in the current study.
De Castro developed the PRCAH FCS to meet four goals. The first goal was a shortperiod-mode damping ratio of 0.7. The second goal was a Generic Control Anticipation Parameter [26] of 0.6 rad·s 2 ·g. The third goal was to satisfy the Gibson dropback criterion [27], and the fourth goal was a phugoid damping ratio of greater than 0.1. De Castro found two sets of PRCAH FCS gains, one for a speed of 100 m/s and the other for a speed of 180 m/s. The gains that de Castro used in the piloted-handling trials were approximately equal to the gains obtained by interpolating the two sets of gains at 150 m/s. In the current study, however, the set of gains for a speed of 100 m/s were used, because the speeds specified for all the approach tasks (92.6 m/s) and most of the takeoff task is close to 100 m/s. The gains, pre-filter, and actuator dynamics that were used in the current study are given in Table 3.5.

Longitudinal Handling Qualities
De Castro [8] predicted the longitudinal handling qualities using the bandwidth/phasedelay criterion, dropback criterion, and the Generic Control Anticipation Parameter (GCAP) criterion. These criteria use the pitch-attitude-to-stick-displacement transfer function of the aircraft-plus-controller system. Mitchell et al. [24] recommend using the first two criteria, together, for predicting the handling qualities of an aircraft with a pitch-rate-command-attitude-hold flight control system. The third criterion, GCAP, is suitable for use as a design tool only.
The handling qualities will now be predicted using the first two criteria, including the dynamics of the simulated BWB column. The handling qualities are evaluated at a low speed and MTOW only. These are the conditions under which de Castro's pilotedhandling trials were conducted, and are also the conditions used in the current study. De Castro's trials used takeoff and landing tasks, with a maximum specified speed during climbout of 250 kts (129 m/s). The landing approach speed was 180 kts (92.6 m/s). This is considerably greater than the approach speed of 150 kts specified by Liebeck [1]; the difference in approach speed may be due to the absence of clamshell-type drag rudders in de Castro's model.

Bandwidth/Phase Delay
One measure of an aircraft's handling qualities is its stability margin when operated in a closed, pilot-in-the-loop compensatory pitch attitude tracking task [8]. The input is the commanded pitch rate, and the output is the pitch attitude, θ. At frequencies above the stability boundary it is assumed that the phase can be represented as a linear function, Then the phase delay τ p is approximately: where ω 180 is the frequency at the stability boundary, and Φ 2ω 180 is the phase at 2 · ω 180 .
The handling qualities can be related to the pitch-attitude bandwidth and the phase delay, as shown in The bandwidth in the first case was limited by the phase margin to 2.1 rad/s, as shown in Figure 3.3a. In the second case, the bandwidth is also limited by the phase margin,

Dropback Criterion
The dropback criterion for longitudinal handling qualities is based on the relationship between pitch-rate and pitch-attitude overshoot following a step control-column input.
The peak pitch rate q peak is plotted relative to the amount of pitch-attitude overshoot ∆θ peak . Both are normalized with respect to the steady state pitch rate q ss . The handling qualities were evaluated using the dropback criterion at an airspeed of maximum ground effect (ĥ = 0.115). Figure 3.5 shows a linearized simulation of the pitch-rate and attitude response to a step pitch-rate command. The input is set to zero halfway through the simulation. Although the pitch attitude is large in the simulation, in real-time handling trials the maximum pitch-rate input would be much less than 1 rad/s.   Table 3.6. Comparing these values to Figure 3.4 shows that Level 1 handling qualities are expected at 90 m/s, for both ground-effect cases. As discussed in the previous subsection, the column is not expected to affect longitudinal handling qualities because the column has a sufficiently-high natural frequency.

Lateral Flight Control System
The lateral flight control system developed by de Castro [8] is shown in Figure 3 it is statically unstable at angles of attack below 5 • . If disturbed from symmetric flight, the un-augmented BWB will diverge. The second feedback loop uses the yaw rate, r. The washout filter is a high-pass filter, and it acts as a yaw damper. It also allows the aircraft to perform a normal turn, in which the steady-state yaw rate is non-zero. There is an interconnection between the aileron and rudder channels, which reduces the amount of aerodynamic cross-coupling. The controller gains that were used are given in Table 3.7.
Different gains were used in the current study relative to de Castro's as the gains used in de Castro's handling trials yielded a Dutch roll damping of 0.045, for level flight at 90 m/s, which was considered too light. The modified gains shown in Table 3.7 yield a Dutch roll damping of 0.14.
De Castro includes a sideslip-reference input to the lateral FCS. In effect, the gain K multiplies the commanded rudder deflection d r (the input to the four-way summation block) by (1 + K β K). The gain K, which was used by de Castro to allow a larger sideslip in response to a given rudder-pedal deflection, was set to zero in the current study. The commanded rudder deflection d r (Figure 3.6) is a function of the gain on the simulator's rudder pedals (converting a linear displacement of the pedal to a commanded angular displacement of the rudder). Because the maximum-commandable rudder deflection in the current study was much larger than that in de Castro's study, using K = 0 was not expected to have much effect on directional control. The maximum-commandable rudder deflection in the current study was 68 • , while in de Castro's study it is 25 • . With K β = 3 and K = 0.5, the effective maximum-commandable rudder deflection in de Castro's study is 62.5 • . In both studies all control surfaces, including the rudder, were limited to ±30 • of deflection. In effect, in both de Castro's study and the current study, the rudder would reach saturation when a pilot displaced the rudder pedals less than half of their available travel.

Lateral Handling Qualities
De Castro [8] applied several criteria to determine the lateral handling qualities of the BWB. Most of the criteria were from the military specification MIL-F-8785C [28]. In the current study, reference is made to the military handbook MIL-HDBK-1797 [6], which is based on the material of MIL-F-8785C. De Castro assessed the roll-axis control power by finding the time taken to roll the aircraft 30 • . De Castro predicted Level 1 handling qualities for airspeeds of 120 m/s and higher, and worse than Level 1 handling qualities for speeds of 80 m/s. Mitchell et al. [24] state that time-to-roll is a suitable criterion for predicting handling qualities of large-angle maneuvers. Since the tasks used in de Castro's handling trials and the current study do not require aggressive, large-angle maneuvers, the time-to-roll analysis was considered irrelevant.
The roll-oscillations analysis of de Castro is also not relevant when the lateral FCS gains of the current study are used (Table 3.7). The handbook MIL-HDBK-1797 states that the roll oscillation criterion is applicable to aircraft with two properties: a medium to high ratio |φ/β| d , which is the ratio of the magnitudes of the roll-to-sideslip response of the Dutch roll mode, and marginally low Dutch roll damping. The ratio |φ/β| d is considered low if it is less than 1.5. The minimum Dutch roll damping for Level 1 handling qualities is 0.08. For trimmed, level flight at 90 m/s, the augmented BWB has a ratio |φ/β| d of 1.9, and Dutch roll damping of 0.14. The ratio |φ/β| d is not much larger than 1.5 and the Dutch roll damping is not especially low.
Mitchell et al. [24] make a distinction between criteria that are applicable to smallangle maneuvers and criteria that are applicable to large-angle maneuvers. Mitchell et al.
also define four categories of mission-task-elements, based on the level of precision (precise or non-precise) and aggressiveness (aggressive and non-aggressive). The handling trials of the current study have two types of tasks: takeoffs and approaches, including landing.
Mitchell et al. categorize these as non-aggressive mission-task-elements. As such, smallangle criteria are suitable to predict the handling qualities. The takeoff and landing are non-precise, while the approach is a precise mission-task-element. The wheel of the simulated BWB has a fairly low natural frequency, and this may cause considerable phase error. As a first-order approximation, the wheel of the simulated BWB can be modeled as a pure time delay of 0.14 s. This approximation is valid at low frequencies only, below 2 rad/s. Since the tasks of the piloted-handling trials generally require low-bandwidth control, it is possible to predict the lateral handling qualities using the roll time delay criterion.  handling qualities would be expected. Table 3.8 summarizes the differences in Dutch roll mode characteristics between sets of FCS gains.

Flight Research Simulator
The UTIAS Flight Research Simulator has a modified DC-8 simulator cockpit with outthe-window displays, speakers and a digital synthesizer to generate sounds, loaded controls, and an EFIS display for flight instruments. The cockpit is mounted on a CAE 300 series hydraulic motion system [29].   The out-the-window visual display is created by three high-end PCs with NVIDIA GTX 285 processors. The scenery and terrain models are created by X-Plane. H x 1024 V pixels, and the refresh rate is 60 Hz. The total transport delay is 58 ms (measured from the receipt of data to half the field of view of the visual system) [30].    Subjects were informed of this limitation before the handling trial started. The interior and exterior of the FRS are shown in Figure 4.3.  The aircraft's angular velocity components in the aircraft body frame are the second set of inputs to the MDA. Although angular accelerations are sensed by humans, the dynamics of the human vestibular system are such that it acts like a velocity sensor in the bandwidth of interest [31]. A fixed gain of 0.5 is applied to both the specific force and the aircraft angular velocity. The specific force is transformed from an aircraft-fixed frame to an inertial frame. The specific force is then converted to an acceleration and high-pass filtered, to remove steady-state accelerations that cannot be simulated with the motion base's limited travel. The aircraft angular velocity is converted to Euler rates and then is also high-pass filtered. Sustained specific forces in the x and y body axes are simulated by tilting the simulator cab relative to the gravity vector -this is called tilt-coordination. The amount of tilt required is found by low-pass filtering the x and y components of the specific force and dividing by the gravity vector. In order to prevent false rotational cues, the angular velocity of tilt coordination is limited. The baseline limits applied to tilt coordination are 3 deg/s. It may be acceptable to pilots to use a higher tilt coordination rate limit for pitch, as pilots are more sensitive to roll-tilt rate than pitch-tilt rate [20]. The Classical Washout MDA is shown in Figure 4.4.
As a starting point, Reid and Nahon's [20] Classical Washout Parameters set 2 or "CW2" parameters were used for the high-pass filters. The CW2 translation and rotation high-pass (HP) filters are defined by the following transfer functions: HP z = s 3 (s + 2 · 4s + 4 2 )(s + 0.01) (4.4) The starting point of the low-pass filters was changed from the CW2 parameters. Reid and Nahon [13] used second-order low-pass filters with a natural frequency of twice the natural frequency of the high-pass filter. Subsequent to the work of Reid and Nahon, Grant [32] developed a method to tune simulator motion. Grant found that experienced  [13] airline pilots objected to the jerky surge and sway motion that was generated using the CW2 parameters. Grant found that lowering the low-pass natural frequencies reduced jerkiness. The following low-pass filters, with a lower natural frequency to prevent jerk response, were used as a starting point in the current study: LP x = 2.5 2 s 2 + 2 · 1 · 2.5s + 2.5 2 · 100 s + 100 (4.6) LP y = 3 2 s 2 + 2 · 1 · 3s + 3 2 · 100 s + 100 (4.7) The low-pass filters have the form of a second-order filter in series with a first-order filter.
The first-order low-pass filter, with a break frequency of 100 rad/s, is used so that the tilt-coordination angular acceleration can be limited.

Frequency Response
The frequency response of the classical-linear-washout MDA depends on the nature of the aircraft motion input. If there is no crossfeed between the translational and rotational motion, such as for a heave maneuver, yaw maneuver, or coordinated roll maneuver, the response depends only on the high-pass filters. If there is crossfeed, so that tilt-coordination is used, then the frequency response is a combination of the low-pass and high-pass filters. If there is no tilt-rate limiting, then the transfer function for surge or sway is given by the following equation [33]: The subscripts ss and aa designate the simulator and aircraft-fixed frames, respectively. The superscript i is replaced by x to indicate surge, and y to indicate sway.
Similarly, the subscript i on the low-pass (LP) and high-pass (HP) filters designates which component of specific force is being considered. As described in Section 4.2.1, the gain k i is 0.5 for all specific force components (as well as for the angular velocity filters).
The fidelity of the simulator motion can be estimated by applying Schroeder's [12] criteria to the frequency response at 1 rad/s. Applying this criteria allows the filter gains and cutoff frequencies to be tailored to each task, to provide the pilot with appropriate motion cues.

Tuning the Parameters for Each Task
The motion drive algorithm must be tuned for the particular tasks to be flown in the BWB simulator. Tuning involves choosing the filter gains and break frequencies to minimize the motion "error" perceived by the pilot. Since the Classical Washout Parameter set 2 (CW2) were optimized for the Boeing 747, improvements to the motion fidelity can be made by customizing the filter parameters to the BWB and the specific tasks to be flown. The tasks are given in Table 4 Group at UTIAS, has extensive experience with simulator motion [33]. He was unable to frequency ω HPz of 4 rad/s was too high. The cutoff frequency was therefore reduced to 2 rad/s. He was also unable to feel any pitch rate cue, with ω HPq at 1 rad/s. The high-pass pitch-rate cutoff frequency was lowered to 0.5 rad/s to make the pitch-rate filter less restrictive. The second expert was a former air-force pilot with experience in turboprop transport aircraft. He commented that, while completing the level-off phase of Task 1, with ω HPz = 2 rad/s the motion felt jerky and sensitive. The input from these two experts led to the changes given in Table 4.4. To accommodate the conflicting opinions regarding heave motion, ω HPz was set to 4 rad/s for the duration of takeoff tasks (1 and 4) and lowered to 2 rad/s for the approach tasks until 200 ft AGL was reached.

Simulator Motion Fidelity
As discussed in Section 4.2.2, the fidelity of the simulator motion can be predicted based on the frequency response of the motion-drive algorithm. To assess the fidelity of the MDA parameters described in Section 4.2.3, the amplitude ratio and phase of the translational and rotational channels were found at 1 rad/s. Figure 4.5 shows the amplitude ratio and phase of the 3 specific force channels and the 3 angular rate channels, for each set of filter parameters. Low motion fidelity is expected for all motion except for pitch rate, which is expected to be medium fidelity. While it may seem unnecessary to use motion at all if it is generally expected to be low fidelity, handling-trial experiments have been successfully conducted in simulators with hexapod motion bases [10] [11]. Except for very specific low-motion tasks, Schroeder's criteria will generally predict low motion fidelity for simulators with hexapod motion bases.

Reduced Motion
The secondary purpose of the piloted-handling trials is to determine the effect of simulator motion on HQRs. There are six tasks. Each task is flown twice, once with full simulator motion and once with a reduced amount of simulator motion. Reduced simulator motion is used instead of no motion, as some pilots might have a bias for or against motion in simulators. A complete lack of motion would be easily detected by the pilot, even if the pilot was not informed of this condition. Reduced motion is used as part of a ruse, where the pilots are told that several flight control systems are being presented to them. In reality, it is the amount of simulator motion that is being varied within tasks.
The reduced motion is an approximation to the aircraft's response to a gust along the aircraft's vertical axis. Etkin [15] describes a method by which the change in normal load factor is found by aerodynamic transfer functions. The aircraft is treated as a point. The aerodynamic transfer functions are given as [15], where u g and w g are the Laplace transforms of the nondimensional turbulence gust components along the aircraft's x and z body axes, and the system matrix of longitudinal  force and moment coefficients is [15], where K α , K p , and K i are the proportional feedback gain on angle of attack, proportional gain on pitch rate, and integral gain on pitch rate, respectively. Note that actuator dynamics are not included in the equations, as they have a much higher natural frequency than the longitudinal modes of the aircraft (actuator ω ac = 30 rad/s). The state vector is given as [15], y = ∆V ∆α q ∆θ ∆δ e T (4.11) The system matrix on the right-hand side of Equation 4.9 is [15], The change in normal load factor from equilibrium due to a gust is caused by two longitudinal components, u g and w g . Since the response due to w g is much larger, u g is neglected. The transfer function for the change in load factor due to a gust w g along the z body-axis is given as [15], The exact transfer function has a frequency response that is similar to a high-pass filter. The simplest approximation is therefore a first-order high-pass filter. The cutoff frequency and gain of the high-pass filter are functions of the aircraft's speed, V . The equation for the filter is given in as, where the scaling factor is The reference speed, V e , is 90 m/s. This is chosen somewhat arbitrarily, for the purpose   Clearly there is a significant phase error, although the phase error is not expected to affect the HQRs. The purpose of the reduced simulator motion is simply to generate some reasonable heave motion to mask the fact that the simulator motion is not presenting most of the aircraft-motion cues to the pilot.

Experimental Setup
Three subjects took part in the handling trials. They were recruited by word-of-mouth and were not paid for participating. Their experience is given in Table 5.1. The subjects were all "high-time" pilots, using Showalter and Parris' [9] definition of "high-time", which is 500 hours of flying time. Only Subject B had flown heavy aircraft. None of the subjects had used the Cooper Harper scale before, although Subjects B and C were familiar with the scale. Each subject was sent an electronic copy of the pilot briefing at least a week before the handling trial. The pilot briefing stated that two flight control systems were being evaluated. Only one flight control system was implemented in the current flight model. The claim that two flight control systems were being evaluated was a ruse; its purpose was to hide the fact that two simulator motion conditions were used.
Since some pilots might have a bias for or against simulator motion, the ruse was intended to minimize this potential bias. The pilot briefing is included in the Appendices. Upon arriving at UTIAS, the subject was asked to read a simulator safety briefing and then sign a consent form. Before entering the FRS, the subject was told of any changes (which were minor, if any) to the pilot briefing. The subject was given a hard copy of the pilot briefing upon which to mark HQRs. With the motion turned off, the subject practiced a normal takeoff, a takeoff with an engine failure, and a normal approach before starting the evaluation phase.
Each task was performed twice, once with full motion and once with ruse motion.
The run order was randomized and consisted of 12 runs (six tasks, each with two motion conditions). Randomizing the order of tasks was done to minimize any training effect on the results. Subjects could repeat any run before evaluating the task. Desired and adequate levels of performance were specified in the pilot briefing. It was up to the subject to decide whether the task performance was desired, adequate, or inadequate.
Self-assessment of performance was used in Schroeder's experiments [12], but in the handling trials of Soparkar and Reid [11], pilots were told of their performance after each run. An audio recording device was used to record any comments made by the experimenter and the subject. The evaluation phase lasted between 1.3 and 2 hours.
The tasks chosen for the current research were taken from the handling trials of de Castro [8]. Runway 08L of Honolulu International Airport (PHNL) and the surrounding area was used for all tasks. The runway is 12300 ft long by 150 ft wide. The first task was a normal takeoff. The aircraft started 700 ft from the end of the runway, aligned with the runway. The wind was from the right, perpendicular to the runway, at a speed of 2 m/s (at a height 10 m above ground). De Castro [8] did not include any wind in any of the tasks that were evaluated for handling qualities. The first task specified a rotation speed, V R , of 180 kts. Subtask 1a consisted of a climb to 5000 ft with a pitch attitude of 10 • and an indicated airspeed of 200 kts. Subtask 1b required the aircraft to accelerate to 250 kts while maintaining 5000 ft. Subtask 1c called for the aircraft to maintain that airspeed and altitude.
The second task was a normal approach. The aircraft started 8nm from the runway, at an airspeed of 180 kts and an altitude of 2230 ft. At this altitude the aircraft was below the ILS glideslope. The landing gear was initially retracted. The mean wind was the same as in Task 1. The initial aircraft heading was set so that the aircraft's course over ground was aligned with the runway. Subtask 2a consisted of capturing the glideslope and then following the approach to 200 ft, maintaining 180 kts. Subtask 2b was the landing, including the flare and touchdown. The touchdown zone was centered at a point in the middle of the runway, 720 ft from the threshold. The same touchdown zone was specified for all landing subtasks.
The third task (called Task 4) was a takeoff with an engine failure. The initial position and heading were the same as in Task 1. To lower the difficulty of this task, the wind was made to be a headwind, aligned with the runway. The left or right outer engine was suddenly failed at an airspeed of 150 kts (77 m/s). Which engine failed was determined by a random number generator. Subtask 4a required the pilot to maintain the runway centreline after engine failure. The aircraft was to be rotated at 180 kts. Subtask 4b called for a straight, wings level climb. The initial climb attitude was to be 5 • , until 1000 ft, at which point the airspeed was to increase to 220 kts and the pitch attitude to 7 • .
The task ended at an altitude of 5000 ft.
The fourth task (called Task 5) was an approach with an engine failure. The initial condition was the same as Task 2, but with the wind direction aligned with the runway, head-on. The left or right outer engine was suddenly failed at an altitude of 1700 ft.
Subtask 5a required the pilot to maintain the runway heading and wings level until 200 ft. Subtask 5b consisted of the landing, including flare and touchdown.
The fifth task (called Task 6) was an approach with a longitudinal offset. The initial condition, including wind, was the same as for Task 2. The pilot was instructed to deliberately fly one dot above the glideslope, until 600 ft AGL. Subtask 6a consisted of regaining the glideslope while maintaining airspeed. Subtask 6b was the landing.
The sixth task (called Task 7) was an approach with a lateral offset. The initial condition was the same as Task 2. The pilot was told to fly one dot to the left of the localizer, until 600 ft AGL. Subtask 7a required the pilot to regain the runway centerline by 200 ft AGL. Subtask 7b was the landing.

Handling-Quality Ratings
The handling-quality ratings the subjects assigned using the Cooper Harper Scale are shown in Figure 5.1. Green markers are used for the full-motion condition, and red markers for the ruse-motion condition. Task 1, normal takeoff, was evaluated on three subtasks. All other tasks were evaluated on two subtasks. The ratings assigned to each subtask are given in Table 5.2. The mean Cooper Harper ratings with full motion and ruse motion are given in Table 5.3, as are the ratings from the equivalent task in de Castro's study. The mean ratings are across all subjects and all two (or three) subtasks.
The run order of the tasks is given in Table 5.4.   Task6  6a  3 3 3 3 3 3   6b  2 3 4 2 3 4   Task7  7a  3 2 3 3 2 3   7b  3 3 4 2 3 4 The pilot briefing defined the desired and adequate levels of performance for each subtask. The levels were the same as those used in de Castro's handling trials. The desired and adequate performance levels are taken into consideration by pilots when they assign a Cooper Harper rating to a task. As discussed in Section 5.1, subjects were not informed of the pilot-vehicle performance level following a task. The aircraft states were recorded during each task. The desired and adequate performance criteria were applied to each task, and a numerical rating was assigned to each subtask. Desired performance was rated a 1, adequate performance a 2, and inadequate performance a 3. It was found that subjects did not try to touch down in the prescribed zone, which was defined in the pilot briefing. The criterion of longitudinal touchdown position was therefore removed, and only the lateral touchdown position was evaluated. Table 5.5 shows the performance level for each task. A dash is shown for subtasks that were not completed.  in the descriptive phrase. Ratings of 1 to 4 specify desired performance, ratings of 5 and 6 specify adequate performance, and ratings of 7 and above specify inadequate performance. Figure 5.2 shows the correspondence between performance level and HQR.
The height of each bar shows the number of times the HQR (in terms of desired, adequate, and inadequate) matched each level of performance. By far the greatest number (45) of subtasks were rated at the desired handling quality level when the performance was also at the desired level. The second most frequent pair was a rating of desired performance when the actual performance was adequate. This occurred for 17 subtasks. The third most frequent pair was a rating of desired when the performance was inadequate, which occurred 6 times. That desired ratings were assigned when the performance was adequate or inadequate could be due to the fact that a subject felt that he or she could achieve desired performance with a reasonable increase in effort.

Analysis of Variance
There was considerable variability in the ratings assigned to tasks and their subtasks, with ratings between subjects differing by as much as 7. Table 5.6 shows the mean rating and variance of each subject. The greatest variance occurred for Task 4, the takeoff with engine failure. The first subtask was to maintain the runway centreline after engine failure. The second subtask was to establish a straight, wings level climb after takeoff.
With full motion, Subject C rated the second subtask a 7, while the other two subjects rated it a 3. With ruse motion, Subject C rated the first subtask a 7, compared to a 3 for the other two subjects. Subject C rated the second subtask a 10, compared to a 3 for the other subjects, with the ruse motion condition. When Subject C performed Task 4 with ruse motion, the aircraft diverged in roll and the task was stopped before completion.
The airspeed was 176 kts when control was lost, which is below the rotation speed of 180 kts.
The piloted-handling trials in the current study are a two-factor within-subjects experimental design. The factors were the task and the motion condition. The task factor had 13 levels (rateable subtasks) while the motion condition had two levels.
The mean ratings and standard deviations are shown in Figure 5.3a for the ruse motion condition and in Figure 5.  This makes it possible to fly the aircraft with some spare capacity to devote to other tasks.

Comparison with Previous Handling Trials
Because the ANOVA found no significant effect of motion on handling quality ratings, the ratings obtained in the current study with full motion and ruse motion are combined.
The mean ratings of each task (across motion conditions) are compared to de Castro's results. The effect of differences between the flight models and tasks of the current study and de Castro's study are discussed below.
There was generally good agreement between the current Cooper Harper ratings and de Castro's results [8]. Except for Task 7, the ratings were within ±1 of the previous results. The changes to the ground-force model and the inclusion of the engines' pitching moment in the current study did not seem to affect longitudinal handling qualities. In fact, for both motion conditions, Tasks 1, 2, and 6 were all rated lower in the current research than in previous trials. These three tasks require primarily longitudinal control.
Since the PRCAH controller holds the pitch attitude constant, so long as the control surfaces aren't saturated, the additional pitching moment from the engines has no effect on longitudinal handling qualities.
The lateral handling qualities were rated considerably different from de Castro's re- sults. In the current study Task 7 was rated 2.00 lower with full motion and 2.17 lower with ruse motion. Two factors may have caused the ratings for Task 7 in the current study to be lower than de Castro's results.
The first factor is that the gains for the lateral controller were different in the current study than in de Castro's study ( Apart from the reasons discussed above, there are three additional reasons why the ratings for some tasks differed from de Castro's results: different inceptor characteristics, subject differences, and random effects. Additional factors that could have led to higher ratings in the current study were the atmospheric turbulence and a constant-direction, variable-velocity wind. However, given that the turbulence intensity and wind speed were fairly low, these factors were not expected to significantly raise the HQRs.

Comparison with Predicted Handling Qualities
The longitudinal handling qualities were predicted to be Level 1. Since subtask 7a required primarily roll control, this is the subtask that is compared to the predicted handling qualities of Chapter 3. Subtask 7a was rated as Level 1. This agrees with the prediction of the roll-mode-time-constant criterion and the roll-time-delay criterion. Subtask 7a was rated better than the Level 2 handling qualities predicted by the Dutch-roll-mode criteria.

Effect of Simulator Motion
The mean Cooper Harper ratings with full motion differed by 0.5 points or less compared to the ratings with ruse motion, except for Task 4. This small difference and the ANOVA suggest that motion did not have a significant effect on HQRs. Subjects B and C commented on the simulator motion. Subject B noticed the lack of motion cue after touching down in Task 6 with ruse motion, and asked if turbulence had been introduced in Task 2 with full motion. Subject C commented after touching down in Task 2 with full motion that it "felt like a landing". Among these two subjects, for the tasks that they commented on the motion or lack thereof, there was only one difference in HQ rating.
Subject B rated the landing in Task 2 a 5 with full motion, compared to a 4 with ruse motion. No subjects asked about or commented on the fictitious second flight control system. It seems that the ruse of the fictitious second flight control system was effective; for while some subjects commented on the simulator motion, they did not state that the motion affected their HQRs.
With ruse motion, Tasks 4 and 5 were rated worse in the current study than in de Castro's trials [8]. Tasks 4 and 5 might have been more difficult in the current study than in the previous trials because of a greater yawing moment for the engine failure tasks in the current study. Appendix D in de Castro's study shows that, in the simulator, the engine moment was made to be 3.5 times less than what it would be in the aircraft.
Three of the four pilots commented in de Castro's trials that the engine failure was not immediately noticeable. In the current study, Task 4 (in particular subtask 4b) gave the greatest amount of spread in ratings. This is understandable given that a slight decrease in airspeed may have caused rudder saturation and loss of control. Subject C's ratings of Task 4 decreased by 3 points with full motion. Subject C completed Task 4 with ruse motion before full motion. Subject C's comments indicate that outside advice influenced his or her second performance of Task 4. So while Task 4 with ruse motion was rated worse than in de Castro's trials, it is probably safest to assume that any difference in ratings of Task 4 were due to some combination of training effect and the greater engine yawing moment.

Limitations
The current study has several major limitations. First, the handling-trial subjects were not test pilots, and as such they did not have experience using the Cooper Harper scale.
may have introduced some bias and scatter into the handling-quality ratings. A third limitation is the small number of subjects and the large variance of their ratings, which limits the statistical power of the results. It was not possible to show a significant effect of motion on the handling-quality ratings. Fourth, according to Schroeder's criteria [12], the predicted motion fidelity was low, except for the pitch rate. The pitch rate was predicted to be medium fidelity for all MDA parameters. The fact that no motion-effect on handling-quality ratings was found may have been due to the relatively-low motion fidelity. A fifth limitation is that the flight model of the BWB treated the aircraft as a rigid body; aeroelasticity was not modeled. Carlsson [35] found that the control effectiveness of a BWB decreased with an increase in airspeed due to static structural deformation.

Conclusion and Recommendations
The handling qualities of a hybrid BWB were found to be Level 1 or 2, depending on the  [8]. The HQRs of the current study were not significantly affected by ground effect. It was shown in Chapter 3 that ground effect had only a small, beneficial effect on longitudinal handling qualities. Ground effect was therefore not expected to be a significant factor in the HQRs. The significant differences in HQRs for the takeoff-with-engine-failure task (Task 4), compared to de Castro's results, were probably due to a much larger (and more realistic) yawing moment in the current study. In addition, the maximum static engine thrust was 10% greater in the current study. The larger yawing moment in the current study likely made the task more challenging, leading to higher HQ ratings. The lower HQ ratings in the current study for the approach with a lateral offset (Task 7) might have been due to higher Dutch roll damping in the current study.
As discussed in the Introduction, Grantham found that handling-quality ratings in a ground-based simulator were conservative. The handling-quality ratings that were obtained in the current study are therefore predictive of the handling qualities of an actual blended wing body. The handling-quality ratings of the current study apply only to the six low-speed tasks that were evaluated, and only to a BWB with the flight control system as described by de Castro and adjusted in the current study.
Simulator motion did not have a large effect on HQRs. The largest difference in mean HQ ratings with full motion compared to ruse motion was 1, and this may have been due to effects other than the motion condition. The reason for the small effect of motion on HQ ratings may be that the simulator motion was predicted to be low fidelity, except for the pitch-rate motion, which was predicted to be medium fidelity. The main benefit of motion in the piloted-handling trials seemed to be as a means for the pilot to know when the aircraft touched down during landing. Schroeder [12] suggests that if the simulator-motion fidelity is not adequate, then the task should be re-designed.
The small effect of motion on HQ ratings may also have been because the tasks were such that the pilot generally created his or her own motion. Schroeder reviewed two studies that found that simulator motion does not seem to be "beneficial for tasks in which the pilot creates his own motion." Except for the engine-failure-during takeoff task, the tasks in the current trials required tracking in fairly calm air. Showalter and Parris [9] found that motion cueing had a statistically-significant effect on pilot performance of a landing-with-wind-shear task. In their study the wind shear was strong enough that severe Dutch roll of the aircraft was induced. Simulator motion would be more important to evaluating handling qualities for tasks with more turbulence or wind shear.
Based on the results of the piloted-handling trials, there are four suggestions for future trials. First, different tasks could be evaluated. Tasks at high airspeeds such as cruise and initial descent could be performed in future trials. If a hexapod motion system is to be used, flight tasks should be designed such that the simulator motion can be tuned to achieve high (or at least medium) fidelity on the Schroeder criteria. Second, static aeroelasticity should be modeled if high-speed tasks are to be used. Third, incorporating drag rudders, as used on NASA and Boeing's X-48B, in the aerodynamic model might help with low-speed control. Data exists on the yawing moment of a flying wing with drag rudders, which might be applicable to a BWB [36]. Fourth, a different flight control system could be used. Goldthorpe et al. [37] describe the "de-augmentation/reaugmentation" controller used on the X-48B. The controller uses dynamic inversion to de-couple the forces and moments of the elevons, allowing the pilot to execute pure roll or yaw maneuvers. Using a different FCS than that used in the current study would obviously affect the handling-quality ratings.
Appendix A Stability Derivatives and Data

BRIEFING FOR PILOTED HANDLING TRIALS Objective
The objective of the handling trials is to obtain pilot evaluations of the handling qualities of a blended wing body aircraft. Two flight control systems will be assessed by the subjects.
Subjects will be asked to perform six low speed flight tasks, and rate the handling qualities using the Cooper Harper Scale after each task. The results will be compared to similar trials conducted at Cranfield University in 2003 in a fixed-base simulator.

Experimental Procedure
The subject will first read this briefing describing the objective and procedures of the experiment. The subject will then read the UTIAS simulator safety instructions. Next, the subject will read an informed consent form that they must sign in order to participate. The subject will then be led into the UTIAS Flight Research Simulator, seated and belted in. The motion base will then be energized. The subject will be given about 15 minutes to familiarize themselves with the aircraft and tasks before the evaluation phase begins.
The entire session will take approximately two and a half hours. The subject can take breaks as needed. During the experiment, the subject will be wearing a headset through which they can communicate with the experimenter. A cockpit voice recorder will be used and subjects are encouraged to make brief comments about the handling qualities during and after a task.
Upon completion of the session the subject will remain seated until the motion system is fully de-energized and the experimenter informs the subject it is safe to undo the seatbelts. The subject will then be led out of the simulator.

TASK DESCRIPTION AND EVALUATION
The flying instructions for each task are given on the following pages. Desired and adequate performance is specified for each task.   The inner two power levers directly control the outer engines, 1 and 3. The centre engine power setting is the average of the two inner levers (see Figure 2).

TASK 1 -Normal takeoff, including climb and acceleration segments
If the EFIS touchscreen landing gear control is difficult to use, an alternate switch is available. The thumb switch on the control wheel can be used instead of the touchscreen to raise or lower the landing gear. If the thumb switch is used, the landing gear lever diagram on the touchscreen will not agree with the actual gear state (up or down).