The handling qualities simulation program for the … handling qualities simulation program for the augmentor wing jet STOL research aircraft by WILLIAM B. CLEVELAND N ABA-Ames Research

The handling qualities simulation program for the augmentor wing jet STOL research aircraft

by WILLIAM B. CLEVELAND

N ABA-Ames Research Center Moffett Field, California

INTRODUCTION

Aircraft have been simulated on computers for a variety of reasons. The training of pilots and crews on operational flight trainers, for example, is a common use of simulation. Subsystems of aircraft are often simulated to firm up the design of the hardware and frequently the whole aircraft must be simulated to help in the design of the subsystems. This is the case in the simulation of the Augmentor Wing Jet STOL Research Aircraft, a modified de Havilland C8-A Buffalo, in which the total aircraft was simulated to determine final design values for control systems and devices which augment control of the aircraft. For research and development simulations such as this one, simulation software and hardware must have general application while in piloted "man-in-the-loop" simulations speed of computation is the overriding concern. Thus the aircraft model and computer software and hardware must be merged to provide an accurate simulation which meets the needs of the research objectives.

The STOL problem

Short Take-Off and Landing (STOL) aircraft are designed to use 1500 ft. runways as opposed to 10,000 ft. runways commonly used by commercial jet transport aircraft. To meet this requirement it is necessary to fly at a slower speed with a resulting steeper flight path angle. Due to the slow speed requirement all aerodynamic control of the aircraft is reduced as aerodynamic control power is proportional to the square of velocity. For example:

Lift cc V 2CL

The coefficient of lift C L is a function of the shape of fixed parts such as the fuselage and wings but it is varied by movable surfaces such as flaps and spoilers.

213

As the aircraft velocity decreases or increases C L

must be increased or decreased to maintain the required value of aerodynamic lift. When conventional means cannot produce sufficient lift special devices are a necessity. For example, if insufficient lift is obtained from aerodynamic properties, direct thrust lift from the jet engines may be employed. Similarly, lateraldirectional (roll-yaw) control is reduced in the same manner as lift at the lower speeds making the STOL class of aircraft in general dependent on special aids in roll-yaw control as well as lift. For the C8-A a special aerodynamically high lift flap is used in conjunction with vector able jet engine thrust to provide the necessary lift at low landing approach speeds.

Handling qualities

Along with loss in control effectiveness the stability of flight maneuvers is also reduced and improvements in stability as well as control must be introduced to bring the aircraft "handling qualities" up to an acceptable level. "Handling qualities" is a general term in which the aircraft characteristics are rated by pilots on a scale ranging anywhere from' 'uncontrollable" to "optimum." The pilot must be asked to perform a specific task. For instance, he may rate the handling quality of an aircraft in a normal landing approach as "optimum." However, the same approach may be difficult or nearly impossible with an engine failed and his rating would be consequently lower. The ratings are nearly purely subjective, as the rating is the pilot's opinion. Normally several pilots are used to avoid personal biases and obtain a crude statistical sampling for a handling quality rating. Since one of the primary outputs of a simulation of this type is the subjective pilot evaluation, systematic investigations require exacting simulation models and the ability to introduce or repeat any combination of initial conditions, failure

From the collection of the Computer History Museum (www.computerhistory.org)

214 Fall Joint Computer Conference, 1971

Figure 1-The augmentor wing jet STOL research aircraft

modes, control effectiveness, air roughness, etc., desired to obtain reliable pilot ratings.

STATEMENT OF PROBLEM

The augmentor wing aircraft

The Augmentor Wing Jet STOL Research Aircraft is sponsored jointly by the National Aeronautics and Space Administration and the Department of Industry, Trade and Commerce of Canada. Major modifications of a de Havilland C8-A Buffalo are presently under way to provide the research vehicle, Figure 1.

To meet the· short field requirements set by the Federal Aviation Agency for STOL aircraft the airplane contains several novel pieces of hardware. The augmented jet flap is a high lift device which gets its name from the blowing of a flat jet of air down the slotted flap as seen in the wing section diagram, Figure 2.

c AIR DUCT ~

FLAP SYSTEM

Figure 2-Cross section of the augmented jet flap wing

Cold air from the fan-jet engines is ducted to ejector nozzles to provide the air jet. The ailerons also contribute lift by drooping in conjunction with the flaps but only up to one half the total flap deflection. The aileron is a boundary layer control device in which air is blown over the surface to improve the aerodynamic force characteristics. Normal roll control by the ailerons is provided by differential aileron deflections about the common flap-aileron angle operating point.

The effect of air blowing on the aileron may be seen in Figure 3, the function of coefficient of roll C laa versus aileron deflection oa and blowing coefficient, CJa• The coefficient CJa is a measure of the cold air thrust, Te,

non-dimensionalized through division by the product of dynamic pressure and wing area, (C J a = Tel ij S) .

As CJa increases for any down going aileron angle (oa>O) so does the rolling moment on the aircraft. It is apparent that boundary layer control adds great effectiveness over the non-blown aileron. Since the roll

. coefficient curves represent an individual aileron the total rolling coefficient for both ailerons is C laa (port) -C laa (starboard) .

The engines themselves are remarkable in that they provide the air for the flap blowing and aileron boundary layer control and more so since the engine thrust is deflectable. The thrust angle, under pilot control, is normally directed aft for cruising but it may be directed straight down to provide direct engine lift at slow speeds.

APPROACH OPERATING POINT

.2 ~

CJa=·OI6

CJa=·008

Figure 3-Rolling coefficient CIBa, a function of aileron deflection, 8a and blowing coefficient CJa for one aileron, downgoing is positive


Objectives of the program

An early piloted simulation of the modified Buffalo was conducted to determine just how the pilot might best control the aircraft in takeoff, transition to cruise and back to landing, and landing itself. The results showed that the aircraft overall had acceptable handling qualities, but the pilot's workload was higher than desirable for a commercial STOL aircraft. However, it was felt that the use of Stability Augmentation Systems (SAS) in lateral (roll) and directional (yaw) would reduce the work load to a satisfactory level. Systems that were designed to help the lateral characteristics provided turn coordination and dutch roll damping. In order to evaluate these control aids the aircraft was simulated on a moving base simulator in as much detail as possible. The test pilots had not only the normal flight instrumentation but good visual and motion cues to make their evaluation of the proposed stability systems. Various effects such as control system hydraulic failures, SAS servo failures, and engine failures were needed in addition to the usual disturbances such as air turbulence to evaluate the handling qualities of the aircraft in both normal operating and failure modes.

Computer requirements of the handling qualities simulation

The problems discussed in these previous sections place certain requirements on the simulation and at the same time allow some few concessions. To be specific the simulation of this vehicle must provide the following:

1. Six-degree-of-freedom equations of motion completely rigorous and no approximations; a flat earth is acceptable in a landing study.

2. Accurate and detailed aerodynamic derivatives including all coupling effects. This is required to access handling qualities.

3. Engine performance model. 4. Stability augmentation system to help make the

aircraft easier to fly in the lateral-directional modes.

5. Air turbulence model, wind shear, gust upset, and/or steady state winds.

6. System failures, engine failures, SAS failures of several types.

7; Non-steady aerodynamic effects; for example, the effects· of a time delay from when the air flows over the wing until it reaches the horizontal stabilizer.

8. Landing gear model for landing and roll out.

Handling Qualities Simulation Program 215

Other simulations, especially of high speed and very large aircraft, would require a representation of body bending modes and aero-elastic effects.

The amount of computation required to accomplish these items listed above made it impractical, if not near impossible, to do with analog computers (at least with the desired accuracy). Thus this simulation was done using a digital computer.

SIMULATION HARDWARE AND SOFTWARE

Simulation computing system

The Ames Research Center simulation computing system is made up of both analog and digital computers. The principal component of the system is the EAI 8400 Digital Simulation Computer. This computer is a 32 bit, 32,000 word machine with a real-time interval timer and floating point hardware. Its input/output to the analog domain consists of 32 bits "in" and 32 bits "out" of discrete on-off signals as well as 64 channels of multiplexed analog-to-digital converters (ADC) and 64 DACs. Peripheral equipment includes four magnetic tapes, line printer, card reader, disk file, and typewriter. Of this equipment 29,500 cells, 64 DACs, 16 ADCs, 17 discrete bits "out" and 12 bits "in" were used in addition to the computer peripherals. The memory was allocated as follows: 6,500 for the simulation monitor system, 21,000 for the simulation program, and 2,000 for the simulation software (user provided). Because most of the cockpit instrumentation was developed in the past for all-analog simulations, data communication with the simulator cab is entirely through the ADC-DAC linkage and the discretes. No digital instruments were used.

In addition to the digital computer an analog computer (EAI 231-R) was used. Whereas the digital computer computes the aircraft model as its primary work, the analog computer is used as a buffering device between the digital computer and the analog recorders, motion and visual simulators, and the various cockpit instruments and controls.

Theoretically the analog computer was not needed since no part of the aircraft was simulated on it but with the myriad of devices requiring data transfer to and from the digital computer it is impractical not to have some sort of analog computer for a data trunking center. No claim is made to call this a hybrid computing arrangement due to the nature of the workloads on the two systems, however, it is interesting that the requirement exists for analog components to be available for the "digital" simulation.



Figure 4-Flight simulator for advanced aircraft located at NASA-Ames Research Center, Moffett Field, California

Simulator system

To obtain the most valid handling quality evaluations the pilot must be subjected to as many motion, visual and aural cues as he would experience in flight. While this is impossible to achieve on a ground based simulator the most important cues of STOL aircraft can be faithfully duplicated on the large motion generator at Ames called the Flight Simulator for Advanced Aircraft .(FSAA), Figure 4. This motion simulator has a fully Instrumented cockpit and six-degrees-of-freedom travel capabilities of ±50 feet in lateral, ±4 feet in vertical a~d longitudinal, and at least ±22.5 degrees in roll, pItch, and yaw. The FSAA was chosen for its large lateral travel which proved most useful in the roll-yaw control handling qualitiBs study and also in the engineout maneuvers. In the simulator the pilot has the capability of flying on instruments or by visual contact. The visual scene is an out-the-window pictorial repres~ntation of a landing field and surrounding countrySIde. The scene is a tBlevision representation in which the model of the landing field is scanned by a color television camera mounted in gymbals so that in addition to three translations the angular motions of roll pitch, and yaw of the airplane are displayed to the pilot:

A mes simulation software

The simulation software system at Ames has evolved from manned simulation requirements. With a "manin-the-Ioop" all work must be performed in real time. Historically, manned simulations havB been done on analog computers with their fast parallel computing

capability and only recently have digital computers been used widely for the real-time problem. Execution time is of prime importance, thus the software used in simulations must meet rigorous execution time requirements. At present the Ames simulation software is of two types: hardware support and program support. The program support software is a group of programs under the label FAMILY 1.1 The main features of FAMILY I include the basic Ames simulation monitor, integration packages, real-time magnetic tape data dump and two special systems-MOTHER (Monitor Time Handling Executive Routine) and CASPRE (Comprehensive Aid to Simulation Programmers and Engineers). The hardware support software serves to make analog type operations tractable from the digital computer. This software is critically important to the operational efficiency of Ames' simulations.

Program support software

A real time scheduler called MOTHER was developed in response to the problem arising from the speed limitations of the EAI 8400 digital computer. It was recognized that programs that contained high frequency systems or that sampled high frequency analog inputs must sample and solve the system equations at high rates to produce sufficient accuracy. However, the size of our simulations indicated that one big program loop solving all the systems would normally produce integration and sampling step size too large to accuratBly reproduce the high frequency portion of the problem. The easiest solution to this problem would be to get a computBr fast enough, but when the problem is not too large and the frequencies are not too high, there are less drastic solutions. The approach used in the Ames' simulations is to do the high frequency operations more often than the low frequency ones using a special piece of software called MOTHER.

MOTHER is a scheduling and executive routine which schedules subroutines to operate at synchronized time intervals or to operate within given time constraints. Subroutines are defined to MOTHER to operate within specific time constraints~ By organizing all the high frequency computations into one list and the remainder into a second list one may define to MOTHER what the computation rate is to be on each list. Both lists are called constrained processes because they must he completed within time constraints as opposed to syn~hronized processes such as input! output of analog data which must be performed at specific time intervals.


In Figure 5 a two loop program is shown as it might be run under a MOTHER schedule. Assume that the I/O for loops A and B must be executed every 10 and 20 milliseconds respectively and that the computation blocks, A and B, must execute within the same 10 and 20 milliseconds.

Since the I/O processes are synchronized, they are executed first. After the I/O the constrained processes begin; process A has been scheduled to execute first since it must execute within the shorter time constraint. At the completion of A, process B begins; at the time of 10 milliseconds the I/O of A interrupts the constrained process and proceeds to execute. Note that A executes again before process B resumes its execution and finally is completed. At this point all the synchronized and constrained processes are completed for this period. The period as used iIi this context is the shortest time into which all the various process times must divide integrally. The ability of MOTHER to do this scheduling is a great aid to the simulation programmer since all he must do is to provide the simulation subroutines and specify the calculation rates. The scheduling burden has been removed.

Other items that have proved useful are MOTHER's executive features such as the servicing of simulation mode control and discrete signal inputs. In the typical operating environment a request for a mode change may come from any of three areas: The analog domain, the program itself, or from the digital control console. M OTHER receives the mode requests and services them by first setting a user-provided mode word and then executing the process list defined for this mode. Discrete bit signals from the analog domain are serviced

TIME, m/sec a 10 20

I/O FOR A

1/0 FOR B

LIST A

LIST B

EXECUTIVE

MODE

WAIT

Figure 5-MOTHER schedule for two computation loops


by either setting a corresponding Fortran word high or low according to the condition of the input bit or by executing special subroutines not in the lists of defined constrained processes. Figure 5 shows the mode and discrete bit servicing periods which follow execution of the defined processes.

The digital simulation computers are completely dedicated to simulations as Ames' batch work is performed at a separate central facility. Hence the emphasis in the simulation laboratory is not on computer throughput but on computer/simulator uptime. In this context computer uptime means not only that the hardware be operational but also that the program be operationally useful. Experienced simulation engineers will certainly agree that changes to programs occur at a high rate and in a seemingly never ending stream. In this state of flux a means of quickly updating and changing data and equations is a necessity. Since the simulation programs are written in Fortran, recompilation is the only practical means of making lengthy changes. However, small changes may not warrant delaying an operational simulation to make a time consuming compilation. The software used to make minor changes is called CASPRE.

Changes of data constants~ for example, are very simple, e.g., if it was desired to set the weight of the aircraft to 100,000 pounds the computer operator need only type + WEIGHT = 100000$. The elements of this statement fall into several categories. The +, =, and $ are commands or operatives defined to perform specific tasks on the character strings WEIGHT and 100000. The plus sign indicated to CASPRE that a Fortran name was to follow, the equal sign is a command to set the "weight" cell to the following data concluded by the dollar sign. This flexibility and ease of change is essential to the operation in research and development simulations.

Some of the many directive codes in the CASPRE system include typewriter display and modification of a data cell's contents in floating point, octal, or integer formats and even in Binary Coded Input (BCI) if the cell happens to contain such data. The display is also available on the line printer making it possible to provide short data print-outs. In practice this print capability is rarely used for more than program checkout.

Small program changes are made with a combination of machine language instructions and symbolic addressing. Making these changes requires some knowledge of the machine language codes but the nature of program patches normally requires only arithmetic plus a few conditional branch instructions. For example, a very common request is for a sign change in an equation. This is easily done by a simple operation



SCALE

SWITCH I OISCRETE DIGITAL COMPUTE R, O· AC o--B=IA.:..:;.S_-t-.;.;,.IN.;.;,.PU-:TS--\ r----.,-----,,-i f7\

CRE E a, • 100 (J!. + a ) V IN AS INSCAl SUBROUTINE IR. a mox BIAS

SWITCH 2 DECREASE

Figure 6-Calibration of cockpit instruments using INSCAL

such as changing an add to a subtract instruction. Large changes such as the insertion of an equation into the program are made by a jump from the compiled program to an area of memory set aside as a patching area, where the equation is patched in machine language and then a jump back to the equation stream.

Hardware support software

One of the operational problems associated with flight simulators is the calibration of the cockpit instruments. The instrument readings are, for the most part, linear with voltage input but the zeroes and scaling vary with each instrument. It is a daily process to calibrate and check each instrument used. Calibration consists of both a bias and a gain on the problem variable. It was the practice at Ames to provide a DAC channel scaled in some convenient manner and with an analog computer do the necessary voltage scaling and biasing before sending the signal to the instrument. This straightforward approach proved to have a major fault albeit a human one. Due to the general purpose nature of the computers and the simulators there is a large turnover of programs on the equipment. The analog equipment requires a high amount of human attention to keep up with changes which frequently resulted in improperly scaled instruments and control inputs thereby plaguing the uptime record of the facilities. The solution to this problem was to scale and bias the instrument drive signals in the digital program before sending them out via the DACs. INS CAL was devised to ascertain just what values of gain and bias were required. Figure 6 shows two switches and an instrument located in the simulator cabin. A computer operator selects the instrument to be scaled by typing the DAC channel number into the INSCAL routine. From this point one of the simulator personnel in the cockpit performs bias and scaling. When the operator sets switch 1 to Bias, the variable to that DAC is set to zero and the DAC output is only a bias voltage. If the instrument is not in its null position, switch 2 is used to increase or decrease the value of the

bias until it does null. Having determined the voltage bias of this instrument, switch 1 is set to Scale position. Internal to the INS CAL program a static test value has been previously assigned to the variable to aid in the determination of its scale value for the instrument. With switch 1 in the Scale position the variable is set to its static value. The operator again uses switch 2 to increase or decrease the scale factor until the meter reads the prescribed test value.

For most simulations this process is repeated for 10 to 30 instruments. However, once the initial cockpit calibration has been done the gains and biases, having been saved in the program, are available for subsequent setup of the program. Normally, subsequent calibration checks require no changes to the stored gain and bias values, thus dramatically reducing the setup and turnaround time for simulations.

A requirement for many channels of recorded data on analog strip chart recorders prompted the multiplexing of·2 variables onto one recorder channel. Multiplexing can be done by mechanically switching between contacts carrying the appropriate signals. This approach will require many DACs in the digital simulation. By multiplexing the variables in the digital computer the same effect is produced, but only one DAC is used to display two variables. Figure' 7 shows an example of what can be done with multiplexing. The input and output signals of a control system have been multiplexed onto one data channel for comparison. Normally high frequency and/or discontinuous signals are not multiplexed since in that form they would be very hard to read. Most variables can be multiplexed in aircraft simulations and the use of this routine has proved to be quite successful.

x-----S2+2twS+w2 y

Figure 7-Strip chart recording of X and Y multiplexed for comparison


AWFTV PROGRAM MECHANIZATION

The organization of the digital program is illustrated in Figure 8. The middle block contains MOTHER the , executive program for the aircraft and support subroutines. The upper block in the figure represents the main program of the simulation which defines to MOTHER the timing requirements for the execution of programs and input/output data transfer schedules.

The three lower blocks constitute the simulation which consists of two real-time calculation loops and a set of supporting subroutines. Because MOTHER can schedule the computations of fast variables more frequently and slow variables less frequently, MOTHER makes it possible to run complex programs with adequate dynamic fidelity than otherwise is possible using a straightforward seriai calculation of all variables. Two calculation loops proved satisfactory in this simulation. The high frequency loop contained the rotational kinematics, part of the rotational aerodynamics, the control system and the landing gear model. The low frequency loop contained the translational kinematics, the remainder of the aerodynamics, the engine model and the simulator drive calculations. The terms "high" and "low" frequencies are relative, of course, and the split of the workload is somewhat subjective. Usually the rotational behavior of an aircraft in flight contains higher frequencies than the translational. As a natural grouping of rotationally oriented work the control system and rotational aerodynamics were put into the fast loop with the rotational kinematics. The landing gear equations must be solved relatively fast due to the high transient frequencies present upon touchdown. The lower frequency work is essentially the remainder of the workload. One improvement that most probably will be made in the future is to solve for altitude in the fast

~EDULES AND DEFINITIONS I FOR MOTH E R

I I

MOTHER, MONITOR TIME HANDLING AND EXECUTIVE ROUTINE 1 I I I

HIGH FREQUENCY LIST LOW FREQUENCY LIST SUPPORT CAPABILITIES

ROTATIONAL DYNAMICS TRANSLATIO NAL DYNAMICS TRIMMING ROUTINE

LANDING GEARS AERODYNAMICS PRINT ROUTINES

AERODYNAMICS ENGINES DYNAMIC CHECKS

CONTROL SYSTEM SIMULATOR VARIABLES LIBRARY FUNCTIONS

Figure 8-0rganization of the digital simulation program


AERODYNAMICS

ENGINES

LANDING GEARS

BODY VELOCITY

Figure 9-Translational dynamics block diagram

loop for improvement in the simulated landing gear response.

The digital simulation depends upon several subroutines some of which _ do not run in real time. Data printout and aircraft trim calculations cannot run in real time. However, the three dynamic check routines which provide modal response checks must run in truetime. These routines are simply called by pressing a computer console push button. Other support software such as wind turbulence models, random number generators, and arbitrary functions of one, two, or three variables require a Fortran call in either the high or low frequency simulation loops.

Equations oj motion

The equations of motion chosen for this simulation are a six-degree-of -freedom rigid body set. The set assumes a flat non-rotating earth in which the linear accelerations are integrated in a local horizontal Euler axis system and the angular accelerations are integrated in the vehicle's conventional body axis. This set of equations is sufficient for landing studies in which the range traversed is only four or five miles or less and the maximum velocities are very low, 60 to 150 knots.

The translational equations are interesting in that all forces, not including gravity, are summed in the body axis frame then transformed to the local horizontal axis where gravity is easily added in before integrating to obtain ground velocities. Wind velocities may be easily inserted in this axis system in the north, east, and down directions. No resolution of winds or gravity from Euler axis to body axis is necessary in this formulation. This is illustrated in Figure 9. One advantage of



DYNAMIC PRESSURE

8SAS

1-------1 IAERODYNAMICS: I AND I

IEQUATIONS OFt : MOTION : L ____ .J

COMPENSAT ION FEEDBACK SIGNALS

Figure Io-Typical control system and stability augmentation system

integrating the forces in the inertial frame is that the w Xx terms present in body axis acceleration equations are omitted along with the corresponding higher frequency content in the angular velocity terms. In digital simulations this is to be desired as reduction of frequency content usually improves solution accuracy.

Aerodynamics

The aerodynamic equation set formulates all the effects of velocity, control deflections, etc. and produces three forces and three moments. The stability derivatives were formulated from wind tunnel data taken with respect to the stability axis; from which a rotation through the angle of attack, a, about the y axis produces derivatives in the body axis frame. The equations and data are representative of many simulations with just a few interesting exceptions. The effects due to angle of attack of the horizontal tail takes into account the downwash of air flow over the wings onto the tail and the variable time delay involved between the distance from wing to tail as a function of speed. The most interesting facet of the aerodynamics simulation is the separation of the effects of right and left ailerons, due to simulated engine failures which stop the air blowing on an aileron with the resultant loss of aerodynamic control force.

Two types of random disturbances were included in the simulation. The first type was a "wing drop" in which a roll acceleration was inserted for a specific period of time to produce a resultant roll angle. The second was a wind gust model which produced noise in the three rotational and three translational velocities using a Dryden model.

Control systems and stability augmentation systems

Figure 10 is a block diagram of a longitudinal control system. This figure shows "force feel" modification

under computer control as a function of dynamic pressure and control column movement (as shown in the figure) or by surface deflection. It also illustrates how the control actuator, in the forward control path, and the SAS, in the feedback control loop, serve to affect the amount of control surface deflection obtained by column movements. In each control mode lateral , , longitudinal, and directional, there is a similar control system to provide correct. pilot work loads.

In the actual aircraft the longitudinal control system is a purely mechanical system while both the lateral and directional systems have hydraulic power assist actuators. Provisions are made in the simulation to fail the hydraulic systems with resulting losses in aileron, spoiler and rudder effectiveness. In this vehicle stability augmentation is necessary in lateral and directional modes only. The lateral SAS system uses sideslip angle, roll rate and yaw rate feedbacks for roll stabilization while the directional SAS uses roll rate and sideslip rate feedbacks for stabilization in yaw. Wherever possible the linear filters are mechanized as simple difference equations using state space transform methods.2

Engines

The engine simulation is primarily a thrust computation in which the port and starboard engine thrusts are calculated separately from their individual throttle and diverter control levers. The engine thrusts contribute to body axis forces and moments for the aerodynamic computations. The jet engines produce hot thrust for propulsion and cold thrust for the flaps and ailerons.

The engine diagram, Figure 11, shows the basic ideas of the thrust simulation. The operation of this circuit

THRUST + DEMAND

Figure ll-Fan jet engine thrust block diagram


is based on non-linear rate limiting of a first order system. When a demand for increased thrust is made the thrust rate, T, is either T'l or T'2. Initially T'l is greater than T'2 so no limiting occurs, and the limited thrust, PL <65. Thrust, T, is a positive exponential function of time but as thrust builds up and P L becomes limited, the thrust will become linear with time until T'l5:.T'2. At this time the limiter acts to make T=T'l providing a first order exponential tail-off of the process. The rate limiting used whe,n thrust is decreasing is much simpler; the maximum thrust rate bares simply a square relationship to thrust. This scheme provides high thrust rates' at high thrust levels and quite low thrust rates at low thrust levels.

The thrust diverter dynamics are modelled with rate limiting and hysteresis between the control and diverter angle. For determination of pilot handling qualities, system failures were implemented providing thrust loss in either of the two engines and a "hard over" diverter lock to a specific angle.

Landing gear

The landing gear model is composed of equations for tire friction forces and oleo reaction forces and the proper resolution of the forces into body axis forces and moments. The friction forces are resolved into two components, one' in line with the tire and the other perpendicular to it. Coefficients of friction for both components are functions of gear velocities. The equations on gear compression and compression rate are rigorous but assume no tire deflection, so all reaction forces are due to the oleo. In the case of the C8-A three individual gears were simulated with the forces and torques on each summed to provide the total landing gear force and moment components.

Subroutine ICTRIM

ICTRIM is a subroutine which performs two separate functions. "10" refers to the calculation of Initial Conditions (ICs) for the velocity terms in the local horizontal or Euler Frame based on inputs of airspeed, Va, sideslip angle, /3, and angle of attack, lX.

"TRIM" refers to the capability of this subroutine to trim the aircraft longitudinally.

In the operation of simulations at Ames it has been found that while use of the Euler frame to integrate accelerations improved calculation accuracies it did pose operational problems. In the use of the simulations it was apparent that research people using the program were more used to-thinking in terms of total airspeed, for instance, than its components in body axes much


less those in Euler axes. Consequently a simple routine was written to accept Va, lX, /3 to calculate initial conditions for north, U E, east, V E, and down, WE velocities. Later it was modified to allow the flight path angle, 1', to be input along with lX to calculate the initial condition of pitch angle, fJ.

Starting the aircraft simulation run with a trimmed aircraft avoids requiring the pilot to waste time trimming the aircraft before starting his task. For a large class of simulation problems the trimming need only consist of nulling pitch acceleration and the aircraft longitudinal and vertical accelerations. When aileron, rudder and sideslip angles and the roll and yaw rotational velocities are zero, the remaining three accelerations are zero.

The trimming algorithm is an iterative scheme. In the routine's most conventional form elevator control, 8e, is used to null pitch acceleration, q; thrust, T, is used to null longitudinal acceleration, Az; and angle of attack, lX, is used to null vertical acceleration, A z•

As an example of the iterative process the current value of q is used to modify the current value of 8e according to the equation 8ei+l = 8ei+kqi where k is an appropriate gain predetermined from the aerodynamics of the aircraft. After modifying 8e the routine commands a cycle of calculation through all the aircraft equations and data using the new value of 8e to obtain a new value of q. Of course lX and thrust are being changed concurrently in the same manner. After sufficient iterations and for reasonable initial condition the routine determines the control inputs for which the accelerations are sufficiently close to zero for the airplane to be considered trimmed.

This basic scheme has proved to be sufficient for transport aircraft. However the C8-A has the capability of directing its thrust from 180 to 1160 from the aft horizontal and when using this thrust vectoring to obtain trim the throttle may be held constant so that the diverter angle then becomes the trim parameter affecting both Az and Az strongly. A more general scheme was devised for the iteration algorithm since thrust diverter angle influenced both vertical and longitudinal accelerations. Basically the chain rule of differential calculus was employed to introduce the effects of lX and thrust diverter angle, 'II, on Az and A z.

We say that

but since the desired Az is zero then

By the chain rule



In like manner

This is a set of two equations in two unknowns if we assume the partial derivatives can be found from the aero and engine data. Now we say that

and

where ~a and All are the solution of the two chain equations. The partials of Ax and A z with respect to 11

are merely thrust modified by the sines and cosines of the diverter angle, but the derivatives with respect to a are unknown analytically and difficult to determine exactly.

The approximation aAz/aa= AAx/ Aa was implemented by having the trim routine determine Ax and Az for some ai and again for ai plus one degree then using those results to calculate the approximations. Since the approximation to the partial derivative was made rather arbitrarily the calculated Aa and ~1I were multiplied by a constant less than 1 to assure convergence. Most generally .8 was used, for instance, ai+l = ai+ .8A,a.

This scheme in all other respects acts as the original more crude iterative scheme for conventional aircraft trim, iterating until the three accelerations are nulled.

Dynamic check routines

The subroutine DYNCHK is used to perform dynamic checks of the aircraft. The routine provides doublets and pulses in roll control and rudder control. In elevator control the optional disturbances are steps and pulses. The disturbances are input as pilot control variations providing a check on the control system as well as the aerodynamics. By recording the response data on eight channel strip chart recorders the user may view the results and determine the parameters in which he is interested.

Print routines

Two separate print routines are included in the simulation. One is a general purpose print routine for determining the status of a list of variables and is useful for printing initial conditions and trim conditions for documentation as well as a trouble shooting aid. The second routine is an attempt to determine the pilot's and aircraft's performance. It collects data in two ways as functions of altitude. Variables such as airspeed and

climb rate are saved at predetermined altitudes while the maximums of such variables as pilot control forces and guidance errors are determined within certain altitude ranges. If this routine is desired it automatically prints after completion of the run. This type of data is useful to correlate with the pilot's subjective ratings of the vehicle's handling qualities.

Support subroutines

Three subroutines from the computer library should be mentioned in the context of supporting simulations. They are WIND, MLTPLX and INSCAL. The latter two were described as hardware support routines. Probably the most important is WIND since it provides the atmospheric turbulence needed in simulations for aircraft handling qualities work. Turbulence is used primarily to assess the effects that real-life turbulence has on controllability, flying qualities, and ride qualities of an aircraft. The disturbance effects on the design of controls and stability augmentation systems are very important to insure that the airplane has sufficient control effectiveness to be manageable during flight in turbulence.

The turbulence model is the Dryden mode1.3 •4 In essence white noise is passed through filters to provide noise in the three translational and three rotational velocities which have good representations of the power spectral densities present in actual air turbulence.

EPILOGUE

The simulation program and simulator hardware provided test pilots a realistic representation of the modified C8-A Buffalo with the result that various control and SAS representations were evaluated using pilot handling qualities ratings.

Final design parameters were found for the aircraft which is scheduled for flight test in early 1972. The digital program has subsequently been used as a base for additional simulations of navigation and guidance of STOL craft in the air traffic control situation near airports and for studies of control and SAS in longitudinal motion for this class of airplane. Of current popular interest are the noise reductions made possible by the high angle landing approaches to the runway· which in turn are made possible by the slow approach speeds. At any ground point the STOL aircraft will be at a higher altitude then conventional aircraft thereby reducing the noise.

Due to the diverse uses of the program it is not con-


sidered to be fixed as present effort is aimed at determining better simulation techniques and models and at making the system software execute faster and be more responsive to the needs of the user.

REFERENCES

1 E A JACOBY J S RABY D E ROBINSON FAMILY I: Software for NASA-Ames simulation systems AFIPS Conference Proceedings Vol 33 Part 11968


2 J V WAIT State-space methods for designing digital simulations of continuous fixed linear systems Transactions of IEEE/PGEC Vol EC-16 No 3 1967

3 F NEWMAN J D FOSTER Investigation of a digital automatic aircraft landing system in turbulence NASA TND-6066 1970

4 C R CHALK Background information an user guide for MIL-F-8785B (ASG), "Military specification-flying qualities of piloted airplanes" AFFDL-TR-69-72 1970



Documents

The handling qualities simulation program for the … handling qualities simulation program for the augmentor wing jet STOL research aircraft by WILLIAM B. CLEVELAND N ABA-Ames Research