Development of Advanced Methods of Structural and Trajectory Analysis for

TransportAircraft

Annual Report

Dr. Mark D. Ardema

Principal Investigator

October 1, 1995 - September 30, 1996

Santa Clara University

Santa Clara, CA 95053

NASA Ames Research Center Grant NCC2-5167

 This report summarizes work accomplished under NASA Grant NCC2-5167,
"Development of Advanced Methods of Structural and Trajectory Analysis for Transport
Aircraft," during the first year of the grant October 1, 1995 - September 30, 1996. The
effort was in two areas: (1) development of advanced methods of structural weight
estimation, and (2) development ofadvanced methods of flight path optimization.

During the Spring of 1996 both graduate student research assistants working on the
project, HC. Chou and Mark Chambers, resigned to take positions in industry. This
required assigning three new Santa Clara people to the project: Dr. Lee Hornberger,
Associate Professor of Mechanical Engineering; Robert Windhorst, graduate student
research assistant; and Frank Dickerson, undergraduate student. These new people
inevitably required time to learn the ACSYNT code and the nature of the ongoing
research. The result is that some of the tasks in the work statement have not been
completed at this time, but will be at the completion of the Grant.

Dr. H. Miura and M. Moore were the NASA collaborators on the Grant.

Review of Results in Structural Weight Estimation

A report that was prepared under a previous grant was published in May 1996
("Analytical Fuselage and Wing Weight Estimation of Transport Aircraft," by
M. Ardema, M. Chambers, A. Patron, A. Hahn, H. Miura, and M. Moore, NASA TM
110392). A paper that summarizes this report has been accepted for presentation at the
World Aviation Congress to be held in October 1996; a copy of this paper appears in
Appendix A.

Throughout the year, integration of the structural weight computer code, PDCYL, into
ACSYNT has continued. Input variables used by PDCYL but already in ASCYNT have
been removed from PDCYL. Infrequently used input variables have been defaulted. Data
transfer has been modified so that optimization runs with ACSYNT can be done with
PDCYL as an integral part of the code.

The major effort to the first year of the grant was to develop an improved method of
estimating the weight of wing and fuselage structures made from composite materials.
This involved an extensive literature search, the coding of a composite materials
subroutine, and demonstrating the code. This work is discussed in detail in Appendix B.

Previously in ACSYNT, the weight of composite material structures was estimated

assuming quasi-isotropic materials, maximum stress failure theory, and smeared structural
elements. The capability being developed accounts for realistic lay-ups of unidirectional
fiber/matrix composites and uses a bi-axial strain failure theory. The new composite
routine has been implemented for the fuselage weight calculation and will be implemented
for the wing weight calculation in the second year of the grant.

A user's manual for the new composite subroutine may be found in Appendix B. As a
check case for the new subroutine, the weight of a composite fuselage ofthe ASA 2150
has been estimated.

The final effort in the structures area has been support of the project to design and
analyze a 150 passenger advanced transport airplane, the ASA 2150. PDCYL has been
used as an integral pan of ACSYNT to estimate the fuselage and wing weights of this
aircraft. Appendix C gives the details of the weight calculations for both Aluminum and
Graphite/Epoxy fuselage versions of the ASA 2150.

Appendix C shows that at a gross take-off weight of 152,181 pounds, the ASA 2150 is
estimated to have a wing weight of 10,315 pounds and a fuselage weight of 15,652
pounds when made of Aluminum. Figure 1 shows the ASA 2150 fuselage bending
moment distribution. The critical loading condition for most of the fuselage is either the
landing condition (L) or the runway bump condition (B), with a small portion governed by
the maneuver condition (M). The shell unit weight distribution is shown on Figure 2.
Approximately the first half of the fuselage is sized by minimum gage, with most of the
rest yield strength critical.

When the fuselage is made of composite material, the weight is estimated to be 15,375
pounds, a weight savings of about 2% relative to aluminum. The composite material is a
uni-directional tape made from Hercules AS4 carbon fiber in Fiberite 12K/938 resin. The
reason for this relatively low weight savings is that for relatively small and lightly loaded
aircraft such as the ASA 2150, the fact that the composite material thickness must be in
integer thicknesses of the basic stack thickness means that the structure is in many places
considerably overdesigned. The basic stack used was a quasi-isotropic lay-up of eight
unidirectional plys. Also, the nonoptimun factor used for the composite was 17% higher
than that for the Aluminum design. As for the Aluminum design, the composite fuselage
was sized by minimum gage and yield strength.

Review of Results in Trajectory Optimization

Because of the unexpected loss of a senior, experienced research assistant, the analysis
of the altitude jumps in energy climb paths could not be completed. Rather, a new, basic
look was taken at energy-state and related approximations. This analysis is related in
Appendix D and will result in an important new addition to the trajectory optimization in
ACSYNT.

Previously, the approach was to minimize a weighted sum of time and fuel
consumption:
 a, K ,+K:r)_,
J=fo(

Under energy-state approximation this may be written

where

p =v(r- D)
W

Thus the optimal climb path is given by

h
, II
K I + K2CT _*,_._

This analysis, however, is for range not specified. When range is specified, the optimal
path is determined by

max
h
( K_ + K:CT " - _x V
11 _=_,,

Where '_x is the adjoint variable associated with the range equation. It is determined
from

2x = KI + K:CcTc
Vc

Where C c, T c, and Vc are associated with the optimal cruise point, obtained by
maximizing the Brequet factor within the flight envelope

A search for the optimal cruise point, and the evaluation of gx, is being added to the
ACSYNT code.
A
0

C
O
em

.Q

on

C
0
E
0

e-
D_
"0
e-

IN qf" _ 0
 _v

m
i
(D

qb

G_

C_

o_

, T
C_ o
_m c_
 Appendix A

"Fuselage and Wing Weight of Transport Aircraft"

by

M. Ardema, M. Chambers, A. Patron, A. Hahn, H. Miura, and M. Moore

to be presented at

World Aviation Congress

October 22 - 24, 1996
Los Angeles, California

October 1996
 965583

Fuselage and Wing Weight of

Transport Aircraft

Mark D. Ardema, Mark C. Chambers, and Anthony P. Patron

Santa Clara Univ.

Andrew S. Hahn, Hirokazu Miura, and Mark D. Moore

NASA Ames Research Center

1996 World Aviation Congress

October 21-24, 1996
Los Angeles, CA

_
_---.,_
_
_ j_ The Engineering Society
For Advancing Mobility
Land Sea Air and Space
.!.z
1.4
INTERNATIONAL
American Institute of Aeronautics
SAE International and Astronautics
400 Commonwealth Drive 370 L'Enfant Promenade, S.W.
Warrendale, PA 15096-0001 U.S.A. Washington, D.C. 20024

For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics or SAE International.
Published by the American Institute of Aeronautics and Astronautics (AIAA) at 1801 Alexander Bell Drive,
Suite 500, Reston, VA 22091 U.S.A., and the Society of Automotive Engineers (SAE) at 400 Commonwealth
Drive, Warrendale, PA 15096 U.S.A.

Produced in the U.S.A. Non-U.S. purchasers are responsible for payment of any taxes required by their
governments.

Reproduction of copies beyond that permitted by Sections 107 and 108 of the U.S. Copyright Law without the
permission of the copyright owner is unlawful. The appearance of the ISSN code at the bottom of this page
indicates SAE's and AIAA's consent that copies of the paper may be made for personal or internal use of
specific clients, on condition that the copier pay the per-copy fee through the Copyright Clearance Center, Inc.,
222 Rosewood Drive, Danvers, MA 01923. This consent does not extend to other kinds of copying such as
copying for general distribution, advertising or promotional purposes, creating new collective works, or for
resale. Permission requests for these kinds of copying should be addressed to AIAA Aeroplus Access, 4th
Floor, 85 John Street, New York, NY 10038 or to the SAE Publications Group, 400 Commonwealth Drive,
Warrendale, PA 15096. Users should reference the title of this conference when reporting copying to the
Copyright Clearance Center.

ISSN#0148-7191
Copyright 1996 by the American Institute of Aeronautics and Astronautics, Inc. and SAE International. All
rights reserved.

All AIAA papers are abstracted and indexed in International Aerospace Abstracts and Aerospace Database.

All SAE papers, standards and selected books are abstracted and indexed in the Global Mobility Database.

Copies of this paper may be purchased from:

AIAA's document delivery service

Aeroplus Dispatch
1722 Gilbreth Road
Burlingame, Calffocnia 94010-1305
Phone: (800) 662-2376 or (415) 259-6011
Fax: (415) 259--6047

or from:

SAExpress Global Document Service

c/o SAE Customer Sales and Satisfaction
400 Commonwealth Drive
Warrendale, PA 15096
Phone: (412) 776--4970
Fax: (412) 776-0790

SAE routinely stocks printed papers for a period of three years following date of publication. Quantity reprint
rates are available.

No pa,'x of this publication may be reproduced in any form, in an electronic retrieval system or otherwise,
without the prior written permission of the publishers.

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE or
AIAA. The author is solely responsible for the content of the paper. A process is available by which discus-
sions will be printed with the paper if it is published in SAlE Transactions.
 965583

Fuselage and Wing Weight of Transport Aircraft

Mark D. Ardema, Mark C. Chambers, and Anthony P. Patron
Santa Clara Univ.

Andrew S. Hahn, Hlrokazu Miura, and Mark D. Moore

NASA Ames Research Center

PDCYL is able to model both skin-stringer-frame and

ABSTRACT
composite _,_ndwich shell fuselage and wing box construc-
tions. Numerous modilications werd' made to PDCYL and its
A method of estimating the load-bearing fu.,_lage weight
associated collection of subroutines. These modifications
and wing weight of transport aircraft based on fundamental
include the addition of detailed fu_lage shell geometry
structural principles has been developed. This method of
calculations; optional integration of a cylindrical fuselage
weight estimation represents a compromise between the rapid
mid_ction between the nose and tail sections; addition of
assessment of component weight using empirical methods
landing and bump maneuvers to the load cases sizing the
based on actual weights of existing aircrafL and detailed, but
fuselage; ability to introduce an elliptical spanwise lift load
time-consuming, analysis using the finite element method. The
distribution on the wing; variation of wing thickness ratio from
method was applied to eight existing sub_nic transports for
tip to root; ability to place landing gear on the wing to relieve
validation and correlation. Integration of the resulting com-
spanwise beading loads; distribution of propulsion system
puter program, PDCYL, has been made into the weights-
components between wing and fuselage; and the determination
calculating module of the AirCraft SYNThesis (ACSYNT)
of maximum wingtip deflection.
computer program. ACSYNT has traditionally used only
empirical weight estimation methods; PDCYL adds to
BRIEF DESCRIPTION OF ACSYNT
ACSYNT a rapid, accurate means of assessing the fuselage
and wing weights of unconventional aircraft. PDCYL also
The Aircraft Synthesis Computer program, ACSYNT, is
allows flexibility in the choice of structural concept, a.s well
an integrated design tool used in the modeling of advanced
as a direct means of determining the impact of advanced
aircraft for conceptual design studies.5 ACSYNT development
materials on structural weighL
began at NASA Ames Research Center in the 1970s and
continues to this day. Tim ACSYNT program is quite flexible
INTRODUCTION
aad can model a wide range of aircraft configurations and
sizes, from remotely pilo4cd high altitude (:raft to the largest
A methodology based on fundamental structural
transport
principles has been developed to estimate the load-carrying
The ACSYNT program uses the following modules, not
weight of the fuselage and basic box weight of the wing
necessarilyinthisorder:. Geometry, Trajecury, Aerody-
foraircraft,and has been incoq_oratedintotheAirCraft
namics, Propulsion, Stability, Weights, Cost, Advanced
SYNThesis program (ACSYNT). This weightroutineisalso
Aerodynamic Methods, and Takeoff.An ACSYNT mn would
available tonm independendy of ACSYNT, and isa modifi-
normally progre_ as follows: the Geometry module is called
cationof a collection of previouslydevelopedstructural
todefinetheaircraftshape aad configuration;theTrajectory
programs.I-4The main subroutinecalledby ACSYNT is
module thenrunsthe vehicle through a specifiedmission;
PDCYL. ThLs study has concentrated on modem Iransport
aircraft because of the detailed weight information available, finally the Weight and Cost modules are executed. To deter-
mine theperformance of the vehicleat each mission point, the
allowing the weights output from PDCYL to be compared to
Trajectory module will call the Ae_xlynamics and Propulsion
actual structural weights. The detailed weight staemenLs also
modules.
allow nonoptimum factors to be computed which, when
multiplied by the load-bearing structural weighL_ calculated by
PDCYL. will give good representative total structure weight
estimates. These nonoptimum factors will be compuled
* Work of the l-ast two authors was supported by NASA Ames
through a regregsion analysis of a group of eight transport
Research Center Grant NCC'2-._S8.
aircraft.
 Afterthe mission is comple_l, the calculated weight of the FINI'IE ELEMENT
air.-aft may be compar_ with the initial e_Umate and an
iteration scheme run to converge upon the required aircraft Finite element analysis is'the matrix method of solution of
weight. This process is nece._arily iterative as the aircraft a discretized model of a structure. This stracture, such as an
weight ACSYNT calculates is dependent upon the initial aircraft fuselage or wing, is modeled as a system of elements
weight estimate. connected to adjacent elements at nodal points. An element is
ACSYNT is able to perform a sensitivity analysis on any a discrete (or finite) structure that has a certain geometric
design variable, such as aspect ratio, thickness-to-chord ratio, makeup and .set of physical characteristics. A mxial force acts
fuselage length or maximum fuselage diameter. Sensitivity is at each nodal point, which is capable of displacement. A set of
defined as (change in objective function/value of objective mathematical equations may be written for each element
function) divided by (change in design variable/design relating its nodal displacements to the corresponding nodal
variable). As an example, if gross weight is the objective forces. For skeletal structures, such as those composed of rods
function and decreases when the wing thickness-to-chord ratio or beams, the determination of element sizing and correspond-
increases, then the sensitivity of thickness-to-chord ratio is ing nodal positioning is relatively straightforward. Placement
negative. It is important to note that while this increase in of nodal points on these simple structures would naturally fall
thickness-to-chord ratio lowers the gross weight of the aircraft, on positions of concentrated external force application or
it may also have a detrimental effect on aircraft performance. joints, where discontinuities in Itx:al displacement cw,:cur.
ACSYNT is also able to stse multiple design variables by Continuum structures, such as an aircraft fumlage or
optimizing the objective functtcm The objective function wing, which would use some combination of solid, flat plate.
represents the interactions Ix'tween design disciplines such as or shell elements, are not as easily discretizable. An approxi-
structures, aerodynamics _md propulsion. The automated mate mesh of elements must be made to model these struc-
sizing of design variables dunng the optimization pnx:ess is tures. In effect, an idealized model _f the structure is made,
accomplished using the gradient method. Two types of where the element selection and sizing is tailored to icx:al
constraints may be impos4M dunng the optimization process. loading and stress conditions.
These are pefformance-lxt,,ed omstraints such as runway The a-ssembly of elements representing the entire structure
length or maximum roll angle, and side constraints on design is a large set of simultaneous equations that, when combined
variables such as limiuttxm,, tm wing span or fuselage length. with the loading condition and physical constraints, can Ix
ACSYNT never violates ccm,,tramts during the optimization solved to find the unknown nodal forces and displacements.
process so that each itt.a'alntm I'wtttlut.'csa valid au'craft. The nodal forces and displacements are then substituted back
into the each element to produce stress and strain distributions
METHODS OF WEl(;lff I':-_;TIMATION for the entire structural model.

Two methods are ctmatmmi_, available to estimate the CLASSICAL PLATE THEORY
load-bearing fuselage weqeht and wing box structure weight of
airca'aft. These methods, m mcrea,,mg order of complexity and CPT has been applied to wing structure design and weight
accuracy, are empirical regre,,_m and detailed finite element estimation for the past 20 years. Using CPT a mathematical
structural analysis. Each metl_ ha_ particular advantages and model of the wing based on an equivalent #ate representation
limitations which will be lwtefly discussed in the fi)llowing is combined with global Ritz analysis techniques to study the
sections. There is an addntt(mal method based on classical structural response of the wing. An equivalent plate model
plate theory (CPT) which may be used to estimate the weight does not require detailed structural design data as required for
of the wing box structure. finite element analysis model generation and has been shown
to he a reliable model for low aspect ratio fighter wings.
EMPIRICAL Generally, CPT will overestimate the stiffness of more
flexible, higher aspect ratio wings, such as those employed
The empirical approach is the simplest weight estimation on modem transport aircraft. Recently, transverse shear
tool. It requires knowledge of fuselage and wing weights from deformation ha.,; been included in equivalent plate models to
a number of similar existing aircraft in addition to various key account for this added flexibility. This new technique has been
configuration parameters of these aircraft in order to produce a shown to give closer representations of tip deflection and
linear regression. This regression is a function of the configu- natural frequencies of higher aspect ratio wings, although it
ration parameters of the existing aircraft and is then scaled to still overestimates the wing stiffness. No fuselage weight
give an estimate of fuselage and wing weights for an aircraft estimation technique which corresponds to the equivalent plate
under investigation. Obviously, the accuracy of this method is model for wing structures is available.
dependent upon the quality and quantity of &tta available for
existing aircraft. AL,;o, the accuracy of the estimation will NEED FOR BETTER, INTERMEDIATE METHOD
depend on how closely the existing aircraft match the configu-
ration and weight of the aircraft under investigation. All of the Preliminary weight estimates of aircraft are traditionally
empirical regression functions currently in the ACSYNT made using empirical methods based on the weights of
program give total fuselage weight and total wing weight. existing aircraft, as has been demribed. These methods,
however, are undesirable for studies of unconventional aircraft
concepts for two reasons. First, since the weight estimating
formulas are based on existing aircraft, their application to Fuselage
unconventional configurations (i.e., canard aircraft or area
ruled bodies) is suspect. Second, they provide no straight- The detailed fu_lage analysis starts with a calculation of
forward method to assess the impact of advanced technologies vehicle loads on a station-by-station basis. Three types of
and materials (i.e., bonded construction and advanced loads are consideredmlongitudinal acceleration (applicable to

composite laminates). high-thrust propulsion systems), tank or internal cabin pres-

On the other hand, finite-element based methods of sure, and longitudinal bending moment. All of these loads
structural analysis, commonly used in aircraft detailed design, occur simultaneously, representing a critical loading condition.

are not appropriate for conceptual design, as the idealized For longitudinal acceleration, longitudinal stress resultants
structmal model must be built off-line. The solution of even a caused by acceleration are computed a,s a function of iongi-
moderately complex model is also computationally intensive tudinai fuselage station; these stress resultants are compressive
and will become a bottleneck in the vehicle synthesis. Two ahead of the propulsion system and tensile behind the propul-
sion system. For intemai pressure loads, the longitudinal
approaches which may simplify finite-element structural
analysis also have drawbacks. The first approach is to create distribution of longitudinal and circumferential (hoop) stress
detailed analyses at a few critical locations on the fuselage and resultants is computed for a given shell gage pressure
wing, then extrapolate the results to the entire aircraft, but this (generally 12 psig). There is an option to either use the
can be misleading because of the great variety of structural, pressure loads to reduce the compressive loads from other
sources or not to do this; in either case, the pressure loads are
load, and geometric characteristics in a typical design. The
second method is to create an extremely coar_ model of the added to the other tensile loads.
aircraft, but this scheme may miss key loading and stress The following is a summary of the methc_.ls used; the
concentrations in addition to suffering from the problems details may be found in Ref. 6.
associated with a number of detailed analyses. Longitudinal bending moment tlistributions from three
load cases are examined for the fuselage. Loads on the
The fuselage and wing structural weight estimation
method employed in PDCYL is based on another approach, fuselage are computed for a quasi-static pull-up maneuver, a
beam theory structural analysis. This results in a weight landing maneuver, and travel over runway bumps. These three
estimate that is directly driven by material properties, load load cases occur at user-specified tractions of gross takeoff
conditions, and vehicle size and shape, and is not confined to weight. Aerodynamic loads are computed a.,; a constant
an existing data base. Since the analysis is done station-by- fraction of fuselage planform area and are considered negli-
station along the vehicle longitudinal axis, and along the wing gible for subsonic transports. For pitch control there is an
structural chord, the distribution of loads and vehicle geometry option to use either elevators mounted on the horizontal tail
is accounted for, giving an integrated weight that accounts for (the conventional configuration) or elevons mounted on the
local conditions. An analysis basod solely on fundamental trailing edges of the wing. The envelope of maximum bending
principles will give an accurate estimate of structural weight moments is computed for all three load cases and is then used
to determine the net stress resultants at each fuselage station.
only. Weights for fuselage and wing secondary structure,
After the net stress resultants are determined at each
including control surfaces and leading and trailing edges, and
some items from the primary structure, such as doublers, fuselage station, a search is conducted at each station to
cutouts, and fasteners, must be estimated from correlation to determine the amount of structural material required to

existing aircraft. preclude failure in the most critical condition at the most
The equivalent plate representation, which is unable to critical point on the shell circumference. This critical point is
assumed to be the outermost fiber at each station. Failure
model the fuselage structure, is not used in PDCYL.
modes considered are tensile yield, compressive yield, local
METHODS buckling, and gross buckling of the entire structure. A
minimum gage restriction is aL_q imposed as a final criterion.
OVERVIEW It is assumed that the material near the neutral fiber of tim
fuselage (with respect to longitudinal bending loads) is
sufficient to resist the shear and torsion loads transmitted
Since it is necessary in systems analysis studies to be able
to rapidly evaluate a large number of specific designs, the through the fuselage. For the shear loads this is a good
metlxxls employed in PDCYL are based on idealized vehicle approximation ms the fibers farth_l from the neutral axis will
models and simplified structural analysis. The analyses of the carry no shear. Also, for beams with large fineness ratios

fuselage and wing structures are performed in different (fuselage length/maximum diameter) bending becomes the
routines within PDCYL, and, as such, will be di_ussed predominant failure mode.
The maximum stre_ failure theory is used for predicting
separately. The PDCYL weight analysis program is initiated at
the point where ACSYNT performs its fuselage weight yield failures. Buckling calculations assume stiffened sbelL,_
behave as wide columns and sandwich _ells behave as
calculation. PDCYL first performs a basic geometrical sizing
of the aircraft in which the overall dimensions of the a2craft cylinders. The frames required for the stiffened shells are sized

are determined and the propulsion system, landing gear, wing, by the Shanley criterion. This criterion is based on the premi_
and lifting surfaces are placed. that. to a first-order approximation, the frames act as elastic
supports for the wide column. 7
There are a variety of structural geometries available for
the fuselage. There is a simply stiffened _ell concept using
 longitudinal frames. There are three concepts with Z-stiffened GEOMETRY
shells and longitudinal frames; one with structural material
pro'portioned to give minimum weight in buckling, one with Fu_lage
buckling efficiency compromised to give lighter weight in
minimum gage, and one a buckling-pressure compromise. The fuselage is assumed to be composed of a nose
Similarly, there axe three truss-core sandwich designs, two for section, an optional cylindrical midsection, and a tail section.
minimal weight in buckling with and without frames, and one The gross density and finene,,_s ratio are defined as
a buckling-minimum gage compromise.
It is assumed that the structural materials exhibit elasto- '
PB= wB (I)
plastic behavior. Further, to account for the effects of creep, va
fatigue, stress-corrosion, thermal cycling and thermal stresses,
options are available to scale the material properties of
strength and Young's modulus of elasticity. In the numerical Rf._ = IB
D (2)
results of this study, all materials were considered elastic and
the full room-temperature material properties were used.
where WB is the fu_lage weight (W8 = gross takeoff weight
Composite matenals can be modeled with PDCYL by
excluding the summed weight of the wing, tails, wing-
assuming them to consist of orthotropic lamina formed into
mounted landing gear, wing-mounted propulsion, and fuel if
quasi-isotropic (two-dimensionally, or planar, isotropic)
stored in the wing), V B is the total fuselage volume, IB is the
laminates. Each of the lamina is assumed to be compo.sed of
fu_lage length, and D is the maximum fu_lage diameter.
filaments placed unidirectionally in a matrix material. Such a
The fuselage outline is defined by two power-law I'_xlies of
laminate has been found to give very nearly minimum weight
for typical aircraft structures. revolution placed back-to-back, with an optional cylindnc',d
midsection between them (Fig. 1). (For the present study, all
eight transports used for validation of the analysis used the
Wing
optional cylindrical midsection)
The horizontal tail is placed according to its quarter chord
The wing structure is a multi-web box beam designed by
location a.s a fraction of the lu_lage length.
spanwise bending and shear. The wing-fuselage carrythrough
Propulsion may be either mounted on the fuselage or
structure, def'med by the wing-fuselage intersection, carries the
placed on the wing. In the case of fu_lage mounted propul-
spanwise bending, shear, and torsion loads introduced by the
sion, the starting and ending positions of the propulsion unit
outboard portion of the wing.
are again calculated from their respective fractions of fu_iage
The load case used for the wing weight analysis is the
length
quasi-static pull-up maneuver. Theapplied loads to the wing
Similarly, the nov landing gear is placed on the fumlage
include the distributed lift and inertia forces, and the point
as a fraction of vehicle length; the main gear, on the other
loads of landing gear and propulsion, if placed on the wing.
hand, may be placed either on the fuselage as a single unit,
Fuel may also be stored in the wing, which will relieve
also as a fraction of fuselage length, or on the wing in multiple
bending loads during the pull-up maneuver.
units.
"Fae wing weight analysis proceeds in a similar fashion to
that of the fuselage. The weight of the structural box is
Wing
determined by calculating the minimum amount of material
requited to satisfy static buckling and strength requirements at
a series of spanwise stations. The covers of the multi-web box The lifting planforms are assumed to be tapered, swept
wings with straight leading and irailing edges. The planform
are sized by buckling due to local instability and the webs by
shape is trapezoidal as the root chord and tip chord are
flexure-induced crushing. Required shear material is computed
parallel. The wing is placed on the fuselage according to the
independently of buckling material. Aeroelastic effects are not
location of the leading edge of its root chord, determined as a
accounted for directly, although an approximation of the
fraction of the fuselage length (Fig. 2). It is assumed that
magnitude of the tip deflection during the puU-up maneuver is
specified portions of the streamwise (aerodynamic) chord
made. For the carrythrongh structure, buckling, shear, and
are required for controls and high lift devices, leaving the
torsion material axe computed independently and summed.
remainder for the structural wing box. The intersection of this
As for the fuselage, there are a variety of structural
structural box with the fuselage contours determines the
geometries available. There are a total of six structural
location of the rectangular carrythrough structure. The width
concepts, three with unstiffened covers and three with truss-
of the carrythrough structure is defined by the correslxmding
stiffened covers. Both cover configurations use webs that are
fuselage diameter.
either Z-stiffened, unflanged, or trusses.

4
 Y

11 "i

IB

Fig. 1 The body configuration.

For the transports in the pre_nt study, all the fuel is ate computed for each of the three loading cases and the
carried within the wing structure. An option is also available envelope of the maximum values taken as the design loading
to carry the fuel entirely within the fu_lage, negating any condition. The bending moment computation is given in detail
bending relief in the wing. in Ref. 4 and will only be summarized here.
Considering first the pull-up maneuver loading, the
LOADS motion is &sxumed to be a quasi-static pitch-plane pull-up of
given normal load factor n (nominally 2.5 for transport
Fuselage aircraft). The vehicle is trimmed with the appropriate contrt)!
surface (a horizontal tail for all eight transport u,_,d for vali-
Fuselage loading is determined on a station-by-station dation in the present study), after which the angle of attack is
basis _long the length of the vehicle. Three types of fuselage calculated.
loads are considered_iongitudinal acceleration, tank pressure, Landing loads are developed as the aircraft descends at a
and bending moment. In the present study, all three load types given vertical speed after which it impacts the ground; there-
ate assumed to occur simultaneously to determine maximum after the main and nose landing gears are assumed to exert a
compressive and tensile loads at the outer shell fibers at each constant, or optionally a ( i - cos(cot)), fi)rce during iLs stroke
station. until the aircraft comes to rest. The vehicle weight is _t equal
Bending loads applied to the vehicle fuselage are obtained to the nominal landing weight. Wing lift a.s a fraction of
by simulating vehicle pitch-plane motion during a quasi-static landing weight is specified, which reduces the effective load
pull-up maneuvex; a landing; and movement over a runway the landing gear carries. Likewi_, the portion of total vehicle
bump. Simpfified vehicle loading models are u_d where it is load the main gear carries ks specified. No pitch-plane motion
assumed that: (1) fuselage lift forces (nominally zero for is considered during the landing.
subsonic ttanslm_) are distributed uniformly over the Runway bump ioadx are handled by inputting the bump
fuselage plan area; (2) wing loading, det,'rmined load factor into the landing gear. Bump load factor is applied
independently, is transferred by a couple of vertical force and according to Ref. 8. This simulates the vehicle running over a
torque through the wing carrythrough structure; (3) fuselage bump during taxi. In a similar fashion to the landing, the wing
weight is distributed uniformly over fuselage volume; lift as a fraction of gross takeoff weight is specified, as is the
(4) control surface forces and landing gear reactions are point portmn of effective load input through the main gear. No
loads; and (5) the propulsion system weight, if mounted on pitch-plane motion is considered during the bump.
the fuselage, is uniformly distributed. A factor of safety
(nominally 1.5) is applied to each load case. The aircraft
weight for each case is selected as a fraction of gross takeoff
weight. All fuselage lift forces are assumed to be linear
functions of angle of attack. Longitudinal bending moments
 if x is behind it. Similarly, the total compressive stress
Wing resultant is

For the wing, only a quasi-static pull-up maneuver

if not pressure stabilized'
condition at load factor n is considered for determining loads.
At each spanwise station along the quarter chord, from the N_ = NxB + Nx A- 0,
Nxp, if stabilized "
wingtip to the wing-fuselage intersection, the lift load, center
of pressure, inertia load, center of gravity, shear force, and
(8)
bending moment are computed. For the inertia load, it is
assumed that the fuel weight is distributed uniformly with
if x is ahead of the inlet entrance, and
respect to the wing volume.
There is an option for either a trapezoidal or a Schrenk 9
if not pressure stabilized
lift load distribution along the wingspan: the trapezoidal
(9)
distribution represents a uniform lift over the wing area (which N'_ = NxB - 0, if stabilized
Nxp,
has a trapezoidal planform) while the Schrenk distribution is
an average of the trapezoidal distribution with an elliptical
distribution, where the lift is zero at the wingtip and maximum if x is behind it. Them relations are based on the premise
at the wing-fuselage intersection. Prandfl has shown that a true that acceleration loads never decrease stress resultants, but
elliptical lift load distribution will have a minimum induced pressure loads may relieve stress, if pressure stabilization is
drag, but a combination of the elliptical and trapezoidal chosen as an option. The stress resultant in the hoop direction
distributions will give a better repre_ntation of actual aircraft is
loading. 9 ".t

Ny = rPgKp (10)

STRUCTURAL ANALYSIS where Kp accounts for the tact that not all of the shell material
(for example, the core material in sandwich designs) is
Fuselage available for resisting hoop stress.
The equivalent isotropic thicknesses of the shell are given
Weight estimating relationships are now developed for the by
load-carrying fuselage structure. In addition, the volume taken
_ g_ (11)
up by the fuselage structure is also determined.
tSc =
Considering first the circular shell, the stre_s resultants in
the axial direction caused by longitudinal bending, axial
acceleration, and pressure at a fuselage station x are
Mr (3)
-tSr _u t
(12)

NxB = 1--7

_SG = Kmgtmg (l 3)
NxWS (4)
Nx A =""7-
for designs limited by compressive yield strength (Fcy),
ultimate tensile strength (Ftu), and minimum gage, respec-
= APs (5) tively. In EQ (13), tmg is a specified minimum material
Nxp p thickness and King i.s a parameter relating [SG to trng which
depends on the shell geometry.
respectively, where r = D/2 is the fuselage radius, A = nr 2 is A fourth thickness that must be considered is that for

the fuselage cross-sectional area, and P = 2xr is the fuselage buckling critical designs, i'SB, which will now be developed.
perimeter. In EQ (3), l_ = ltr 3 is the moment of inertia of the The nominal vehicles of this study have integrally stiffened
shell divided by the shell thickness. In EQ (4), for the case of shells stabilized by ring frames. In the buckling analysis of
fuselage-mounted propulsion, Ws is the portion of vehicle these structures, the _ell is analyzed ms a wide column and the
weight ahead of station x if x is ahead of the inlet entrance, or frames are sized by the Shanley criteria. "/Expressions are
the portion of vehicle weight behind x ifx is behind the nozzle derived for the equivalent isotropic thickness of the shell
exit. In EQ (5), Pg is the limit gage pressure differential for the required to preclude buckling, i'SB, and for the smeared
passenger compartment during cruise. The total tension stress equivalent imtropic thickness of the ring frames required to
resultant is then preclude general instability, iF . The analysis will be restricted
to the case of cylindrical shells. The major assumptions are
that the structural shell behaves as an Euler beam and that all
N+x = Nx B + Nzl , (6)
structural materials behave elastically.
For the stiffened shell with frames concept, the common
if x is ahead of the nozzle exit, and
procedure of assuming the shell to be a wide column is
adopted. If the frame spacing is defined as d and Yonng's
(7)
N +=NxB +Nxp +Nx A
modulus of the shell material is def'med as E, the buckling resultant N x on the cylinder. The buckling equation for these
equation is then frameless sandwich shell concepts is

(14) (22)

where m is the buckling equation exponent. Or, solving for

or, solving for t'SB
IsB

(15)
tsB = _ Ee
(23)

Fuselage structural geometry concepts are presented in

Table 1; valuesof the shell efficiency e for the various
This equation is based on small deflection theory, which
structural concepts are given in Table 2. The structural shell
seems reasonable for .,_andwich cylindrical shells, although it is
geometries available are simply stiffened, Z-stiffened, and
known to be inaccurate for monocoque cylinders. Values of m
truss-core sandwich. We next size the frames to prevent
and e may be found, for example in Refs. 10 and 11 for many
general instability failure. The Shanley criterion is based on
shell geometries. Table"2 gives values for sandwich structural
the premise that the frames act as elastic suplx3rts for the wide
concepLs available in PDCYL, numbers 8 and 9, both of which
column; this criterion gives the smeared equivalent thickness
are truss-core .sandwich. The quantities N x , r, and conse-
of the frames as
quently ?SB, will vary with fuselage'*station dimension x.
At each fuselage station x, the shell must _tisfy all failure
/ rtCFNx criteria and meet all geometric constraints. Thus, the shell
t'FB = 2r2al" (16)
thickness is .,;elected according to compression, tension,
Y KFld3EF
minimum gage, and buckling criteria, or

where CF is Shanley's constant, KFI is a frame geometry

?S = max(?sC,?ST,TSG 'tSB ) (24)
parameter, and E F is Young's modulus for the frame material.
(See Ref. 3 for a discussion of the applicability of this criterion
and for a detailed derivation of the equations pre_nted here.)
If ?S = tSB' the structure is buckling critical and the
If the structure is buckling critical, the total thickness is equivalent isotropic thickness of the frames, IF, is computed
from EQ (20). If t's > ?SB, tile sU'ucture is not buckling critical
[ = t-SB+ [FB (17) at the optimum frame sizing and the frames are resized to
make _S = tsB" Specifically, a new frame spacing is computed
Minimizing i" with respect to d results in from EQ (15) as

I
(25)

(18)
r =Fi7atK ,7%e3
and this value is u_d in EQ (16) to determine ?F.
The total thickness of the fuselage structure is then given
. by the summation of the smeared weights of the shell and the
- = --t (19)
tSB 4 frames

. tB =ts+tF (26)
t'_. = t (20)
"D

The shell gage thickness may be computed from

I ?g = ?S / King. TI_ ideal fu_lage structural weight ks obtained
by summation over the vehicle length
d=( 6r2 T__JPF
_I _"
(21)
(27)

where PF is the density of the frame material and p is the

density of the shell material, so that the shell ix three times as
where the quantities sub_ripteal i depend on x.
heavy as tim frames.
Since the preceding analysis gives only the ideal weight,
Frameless sandwich _ell concepts may aim be u.,_.xl. For
W I, the nonoptimum weight, WNO (including fasteners,
these concepts, it is assumed that the elliptical shell buckles
cutout.s, surface attachments, uniform gage penalties,
at the load determined by the maximum compressive stre_s

8
 Table 1 Fuselage structural geometry concepts

KCONsets
concept
number
Simply stiffened shell, frames, sized for minimum weight in buckling
Z-stiffeaed shell, frames, best buckling
Z-stiffened nell, frames, buckling-minimum gage compromise
Z-stiffened shell, frames, buckling-pressure compromise
Truss-core sandwich, frames, best buckling
Truss-core sandwich, no frames, best buckling

Truss-core sandwich, no frames, buckling-minimum gage-pressure compromise

Table 2 Fuselage structural geometry parameters

Structural concept m _. King Kp Kth

(KCON)

2 2 0.656 2.463 2.463 0.0

3 2 0.911 2.475 2.475 0.0
4 2 0.760 2.039 1.835 0.0
5 2 0.760 2.628 1.576 0.0
6 2 0.605 4.310 3.965 0.459
8 1.667 0.4423 4.820 3.132 0.405

9 1.667 0.3615 3.413 3.413 0.320

manufacturing constraints, etc.) h_ yet to be determined. The

Z = WBEND(Y) (29)
method used will be explained in a later section.
PZst

Wing
where WbEND is the weight of bending material per unit span
and p is the material density. WbEND is computed from
Using the geometry and loads applied to the wing
EQS (28) and (29). The weight per unit span of the shear
developed above, the structural dimensions and weight of the
material is
structural box may now be calculated. The wing structure is
assumed to be a rectangular multi-web box beam with the
pFs
webs running in the direction of the structural semispan. w;.n ty)-- = (30)
Reference 10 indicates that the critical instability mode for OS
multi-web box beams is simultaneous buckling of the covers
due to local instability and of the webs due to flexure induced where FS is the applied shear load and oS is the allowable
crushing. This reference gives the _lidity (ratio of volume of shear stress. The optimum web spacing is computed from 2
structural material to total wing box volume) of the least
weight multi-web box beams a.s

(28)

where and e depend on the cover and web geometries (31)

(Table 3), M is the applied moment, t is the thickness, E is the
elastic modulus, and ZS is obtained from Ref. I0. The solidity where subscripts W and C refer to webs and covers,
is therefore respectively. The equivalent i.,_tropic thicknesses of the
covers and webs are

9
 Table3 Wing structural coefficients and exponents

ec Ew Kgc Kgw
Covers Webs
0.556 3.62 3 0.605 1.000 0.407
Unstiffened Truss 2.25
0.556 3.62 3 0.656 1.000 0.505
Unstiffened Unflanged 2.21
0.556 3.62 3 0.911 1.000 0.405
Unstiffened Z-stiffened 2.05
0.600 1.108 2 0.605 0.546 0.407
Truss Truss 2.44
0.600 1.108 2 0.656 0.546 0.505
Truss Unflanged 2.40
0.600 1.108 2 0.911 0.5,46 0.405
Z-stiffened 2.25

used for validation in the present study, wc = D.) The quanti-

1
fies dw, tw, and tC are computed in the same manner a.,_for the
: M (32) box. The weight of the _ear material is
re- Zste cdw
)
(39)
WSHEARc = p FS w C
OS

(2-1-- (33)
where FSo = Fs(O). ,,
The torqueon thecarrythroughstructure
is

respectively, and the gage thicknesses are T = MO sin(As) (40)

(34)
tg C = KgctC and the weight of the torsion material is then

tg w = Kgw? w (35) pT(to+ CsR)Wc (41)

WTORSIONc = toCsRO S
of E, e, Ec, EC, EW, Kf_, and Kg c are found in
Values _ --- 0
Table 3 for various structural concepts, t If the wmg structural Finally, the ideal weight of the carrythrough structure is
semispan is divided into N equalTength segments, the total computed from a summation of the bending shear and torsion
ideal weight of the wing box structure is material, or
N
WC = WBENDc + WSHEAR c + WTORSION c (42)
(36)
2bs "_ [W' + W' )
WBox : Z.A B oi slieR,
i=1 As in the case of the fuselage structural weight, non-
optimum weight must be added to the ideal weight to ob_n
The wing carrythrough structure consists of torsion the true wing structural weight. The method used will be
material in addition to bending and shear material. The torsion discussed below.
material is required to resist the twist induced due to the sweep
of the wing. "I'he bending material is computed in a similar REGRESSION ANALYSIS
manner as that of the box except that only the longitudinal
component of the bending moment contributes. Letting Overview
to = tO' = 0) and M0 = m(y = 0),
Using fu_lage and wing weight statements of eight
subsonic transports, a reladon between the calculated load-
(37)
bearing structure weights obtained through PDCYL and the
actual load-bearing structure weights, primary structure
weights, and total weighu_ is determined using statistical
The weight of the bending material is then analysis techniques. A basic application which is first
describe_ is linear regression, wherein the estimated weights
(38)
WBEND c = p_,CCSRtOW of the aircraft are related to the weights calculated by PDCYL
with a straight line, y =mx + b, where y is the value of the
estimated weight, m is the slope of the line, x is the value
obtained through PDCYL, and b is the y-intercept. This line is
where wc is the width of the carrythrough structure. (When
termed a regression line, and is found by using the method of
the wing-fu_lage intersection occurs entirely within the
least squares, in which the sum of the _uares of the residual
cylindrical midsection, a.,_is the case with all eight transport

10
errorsbetween
actualdatapointsandthe corresponding points PDCYL analysis only predicts the load-carrying structure of
on fl_e regression line is minimized. Effectively, a straight line the aircraft components, a correlation between the predicted
is drawn through a set of ordered pairs of data (in this case weight and the actual load-carrying structural weight and
eight weights obtained through PDCYL and the corresponding primary weight, as well as the total weight of the fuselage, was
actual weights) so that the aggregate deviation of the actual made.
weights above or below this line is minimized. The estimated Structural weight consists of all load-carrying members
weight is therefore dependent upon the independent PDCYL including bulkheads and frames, minor frames, covering,
weighL covering stiffeners, and longerons. For the linear curve-fit, the
Of key importance is the degree of accuracy to which the resulting regression equation is
prediction techniques are able to estimate actual aircraft
weight. A measure of this accuracy, the correlation coefficient, Wactual = 1.3503Wcalc R = 0.9946 (46)
denoted R, represents the reduction in residual error due to the
regression technique. R is defined as This shows that the nonoptimum factor for fuselage
structure is 1.3503; in other words, the calculated weight must
be increased by about 35 percent to get the actual structural
(43)
weight. For the alternative power-intercept curve fitting
analysis, the resulting load-carrying regression equation is

where Et and Er refer to the residual errors associated with the l 0179
regression before and after analysis is performed, respectively. Wactual = 1.1304W_ c R = 0.9946 (47)
A value of R = 1 denotes a perfect fit of the data with the
regression line. Conversely, a value of R = 0 denotes no To use either of these equations.to estimate total fuselage
improvement in the data fit due to regression analysis. weight, nonstructural weight items must be estimated inde-
There are two basic forms of equations which are pendently and added to the structural weight.
implemented in this study. The first is of the form Primary weight consists of all load-carrying members as
well as any secondary structural items such as joint.s fasteners,
Yest = mXcalc (44) keel beam, fail-safe straps, flooring, flooring structural
supplies, and pressure web. It also includes the lavatory
The second general form is structure, galley support, partitions, shear ties, tie rods,
structural firewall, torque boxes, and attachment fittings. The
a linear curve fit for this weight yields the following primary
Yest =raXcal c (45)
regression equation

Fuselage Wactual = 1.8872Wcalc R = 0.9917 (48)

The analysis above is used to develop a relationship

The primary power-intercept regression equation is
between weight calculated by PDCYL and actual wing and
fuselage weights. The data were obtained from detailed weight
breakdowns of eight transport aircraft 12-16 and are shown in Wactual = 1.6399WelL141 R = 0.9917 (49)

Table 4 for the fuselage. Because the theory used m the

Table 4 Fuselage weight breakdowns for eight transport aircraft

Weight, lb

Aircraft PDCYL Load-carrying structure Primary smacture Total structure

B-720 6545 9013 13336 19383

B-727 5888 8790 12424 17586
B-737 3428 5089 7435 11831
B-747 28039 39936 55207 72659

I)(2-8 9527 13312 18584 24886

MD-II 20015 25970 34999 54936

MD-83 7443 9410 11880 16432

L-1011 21608 28352 41804 52329

11
 Wing
The total fuselage weight accotmts for all members of the
-'body, including the smacun'al weight and primary weight. It The same analysis was performed on the wing weight for
does not include passenger aoeomm(xlatious, such as seats,
the sample aircraft and is shown in Table 5. The wing box,
lavatories, ki_, stowage, and fighting; the electrical
of Ioad-casrying strucun'e, consists of spar caps, interspar
system, flight and navigation systems; alighting gear, fuel and coverings, spanwise stiffeners, spar webs, spar stiffeners, and
propulsion systecas; hydraulic and pneummic sysu=ms; the inte_parribs.The wing box linearregression equationis
communication system; cargo accommodations; flight deck
- acccmmodmions; airconditioningequipment.;theauxiliary Wactual= 0.9843Wcatc R = 0.9898 (52)
pow_ system;and e_ncrgcncysystems.Linear regression
results
in thefollowingtotalfuselageweight equation
so that the nonoptimum factor is 0.9843. Power-intercept
regression results in
Wacma/= 2.5686Wcatc R = 0.9944 (50)

Wactuat = 1 3 34 2W cal,c
09701 R =0.9902 (53)
This shows that the nonoptimum factor for the total
fuselage weight is 2.5686; in other words, the fuselage
stmetnre weight estimated by PDCYL must be increased by Wing primary structural weight includes all wing box
about 157 percent to get the actual total fuselage weight. This items in addition to auxiliary spar caps and spar webs, joints
nonoptimum factor is used to compare fuselage structure and fasteners, landing gear support beam, leading and wailing
weight estimates from PDCYL with total fuselage weight edges, tips, structural firewalL bulkheads, jacket fittings,
estimates from the Sanders and the Air Force equations used terminal fittings, and attachmenL_. Linear regression results in
_t
by ACSYNT.
The total fuselage weight power-intercept regression Wactual = l'3442Wcatc R = 0.9958 (54)

equationis
Power-intercept regression yields
WacmaJ = 3.9089W9:78 R = 0.9949 (51)
Wacma/= 2-1926Wc0_9:34 R = 0.9969 (55)
Plots of actual fuselage component weight versus
PDCYL.ealeulated weight, as well as the corresponding linear
regressions, are shown in Figs. 3-5.

40000 , , , ]

35000

30000
Wpc
%
//! MD-11

25000

"0

--J 20000
m
B

15000

10000
n71

50OO

5000 10000 15000 20000 25000 30000

0

Fig. 3 Fuselage load-carrying strucuae and linear regression.

12
 60000

50000

L-11

J 40000
)11_

Waclu_ 1.8872 WpDcfYL

MD-11

30000

20000

10000

0 5000 10000 15000 20000 25000 30000

Fig. 4 Fuselageprimary structureand linear mgre_ion.

80000

747

70000

60000
MD-11

50000

40000

30000
DC-8
720
20000
727 D
MD-S3
737 5
10000

0 5000 10000 15000 20000 25000 30000

Fig. 5 Fuselagetotal ra-octm_a_! .near rc_n.

13
 Table 5. Wing weight breakdowns for eight transport aircraft

Weight, lb

Aircraft PDCVL Load-carrying structure Primarystructure Total structure

B-720 13962 11747 18914 23528

B-727 8688 8791 12388 17860
B-737 5717 5414 7671 10687
B-747 52950 50395 68761 88202

DC-8 22080 19130 27924 35330

M'D-11 33617 35157 47614 62985

MD-83 6953 8720 11553 15839

28355 36101 46233

Dimussion
The total wing weight includes wing box and primary
weight items in addition to high-lift devices, control surtaces, Both fuselage and wing weight linear and power regres-
and access items. It does not include the propulsion system, sions give excellent correlation with the respective weights
fuel system, and thrust reversers; the electrical system; of existing aircraft, a.,_evidenced by the high values of the
alighting gear; hydraulic and pneumatic systems; anti-icing correlation coefficient, R. It should be noted that even though
devices; and emergency systems. The resulting total weight the power-ba.,_-d regressions give correlations equal to or
linear regression equation is better than the linear regressions their factors may vary
distinctly from the linear ca._s. This is due to their powers
Wactu d = 1.7372Wcah- R = 0.9925 (56) not equaling unity.
Because estimates of non-load-bearing primary structure
This shows that the nonoptimum factor for the total wing are generally not available at the conceptual design stage, and
weight is 1.7372; in other words, the wing box weight esti- because nonprimary structure is probably not well estimated
mated by PDCYL must be incre&_ed by about 74 percent to by a nonoptimum factor, EQS (48) and (54) are recommended
get the actual total wing weight. This nonoptimum factor is for estimating the primary structural weights of the respective
used to compare wing box weight estimates from PDCYL with transport fuselage and wing structures (Figs. 4 and 7).
total wing weight estimates from the Sanders and the Air A comparison may be made between weight estimates
Force equations used by ACSYNT. from weight estimating relationships currently used by
The power-intercePt equation for total wing weight is ACSYNT, PDCYL output, and actual aircraft component
weights. Figure 9(a) shows a comparison between fuselage
._ "_AaAUP0"9268
Wactual = _._'"caZc R = 0.9946 (57) weight estimated from the Sanders equation, the Air Force
equation, and PDCYL output with the actual fuselage weight
of the 747-21P. Figure 9(b) shows a similar compari,mn for the
Plots of actual wing component weight versus PDCYL-
wing weight.
calculated weight, as well as the corresponding linear
regressions, are shown in Figs. 6--8.

14
 60000
' ' ' ' 1 ' ' ' ' I

50000
F
I I , I /
t
!/=__
/'ql

i ! _ /'

! : i /'
40000 waca= d = uJu_ WpDcY L i
" i ._ .O-ll/i "
! i ;
U1011

30000 i _ e/

] i _

20000 i _ " _ i i
/ .,,,,, ,
.o-_ iJ" 720 i
10000
737

. ._l.J-

0 10000 20000 30000 40000 50000 60000

Fig. 6 W'mgload-carrying structure and linear re_ion.

70000
.... ' .... ! .... I .... ' .... J 747

60000

Wa,=_ ,, 1.3442 Wp _L i
" rX . un.1, / '

!
50000

:
J\/ i
40000

: !i/ ....
30000 I

]
20000 I
I
I
I
t
10000 i

I
.... l .... I , , , , I , , , , I , , t .... ,
0

0 10000 20000 30000 40000 50000 60000

Pig. 7 W'mg Frimary suucua_ and liuear regrcssioe.

15
 90000 .... , " e.747
, + , ,
,
! " .
/]
80000
it//

70000
MD-11 /

0/ :

+oooo W.=..h=_
- 1.7372Wp_ _t
50000 ;1011 /

40000 ' i

I
30000

20000 - I
M t
i I i '_'+ i
i i t I I I

10000

0 10000 20000 30000 40000 50000 60000

Fig. 8 Wing total strucmn'eand linear regression.

0.12

0.10

0.08

0.06
C] Body
W/WGTO [] Shell
0.04
Frames

[] Non<_+n_
0.02

0.00
PDC_ Actual Sandem Air Force

Fig. 9(a) Fuselage weight estimation comparison for 747-21P.

16
 0.20-

0.18"

0.16"

0.14-

0.12- - f_.

W/WGTO
0.10-

0.08-
(

0.06"

[] Body
0'04 t

Et
[] Non-Optimum

""'1= =',==

PDCYL Actual Sanders Air Force

Fig. 9('o) Wing weight estimation comparison for 747-21P.

REFERENCES 9McCormick, B. W.: Aerodynamics, Aeronautics, and

Right Mechanics. John Wiley & Sons, 1979.
1Ardema, M. D.: Body Weight of Hypersonic Aircrgt:
Part 1. NASA TM-101028, Oct. 1988. lCxawford, R. F.; and Burns, A. B.: Strength, Efficiency,
and Design Data for Beryllium Structures. ASD-TR-61-692,
2Ardema, M. D.; and Williams, L. J.: Transonic Trans_rt Feb. 1962.
Study - Slrucun'es and Aewodynamics, NASA TM X-62,157,
May 1972. 11Crawford ' IL F.; and Burns, A. B.: Minimum Weight
Potentials for Stiffened Plates and Shells. AIAA ]., vol. 1,
3Ardema, M. D.: Structural Weight Analysis of Hyper- no. 4, Apr. 1963, pp. 879-886.
sonic Aircraft. NASA "IN D-6692, Mar. 1972.
12Niu, M. C.-Y.: Airframe Structural Design: Practical
4Ardema, M. D.: Analysis of Bending Loads of Hyper- Design Information and Data on Aircraft Su'uctures. Conmflit
sonic Aircraft. NASA TM X-2092, 1970. Press, 1991.

5Moore, M.; and Samuels, J.: ACSYNT Aircraft 13york, P.; and Labell, R. W.: Aircraft Wing Weight
Synthesis Program - User's Manual. Systems Analysis Build-Up Methodology with Modification for Materials and
Branch, NASA Ames Research Center, Sept. 1990. Construction Techniques. NASA CR-166173, Sept. 1980.
6Ardema, Mark D.; Chambers, Mark C.; Patron, 14Thomas, R. B.; and Parsons, S. P." Weight Data Base.
Anthony P.;Hahn, Andrew S.;Miura, Hirokazu;and Moore, The Boeing Company, Commercial Airplane Division, Weight
Mark D.: Analytical Fuselage and Wing Weight Estimation of Research Group, Doc. No. D6-23204 TN, 1968.
Transport Aircraft. NASA TM-110392, May 1996.
15McDonnell Douglas Aircraft Company, Detailed
7Shanley, F. R.: Weight-Strength Analysis of Aircraft Weight Statement for MD-11 Transport Aircraft, June 1987.
Structures. Second Edition, Dover, N.Y., 1960.
t6McDonnell Douglas Aircraft Company, Detailed
8Megson, T. H. G.: Aircraft Structures for Engineering Weight Statement for MD-80 Transport Aircraft, July 1991.
Students.Second Edition, Halsted Press, 1990.

17
 Appendix B

Designing Composite Transport Aircraft

by

L. Homberger, M. Ardema, and F. Dickerson

 DESIGNING COMPOSITE TRANSPORT AIRCRAFT
by
Mark Ardema, Frank Dickerson and Lee Hornberger
Mechanical Engineering Department
Santa Clara University

ADDING COMPOSITES TO PDCYL

Light weight materials such as fiber reinforced plastics (composites) and

bonded honeycomb sandwiches have become more and more common in
airplanes in the last two decades (1). Designers value the unique
properties of these materials, particularly their high stiffness to weight
ratios. They must, however, balance these assets against the additional
cost of these materials and their manufacture. To aid designers with this
analysis, a composites subroutine has been added to the PDCYL structures
weight calculation code. This subroutine sizes the thickness of a
particular composite necessary to withstand the required aircraft loads,
and provides this information to PDCYL which calculates the resultant
weight of the aircraft.

TYPICAL AIRCRAFT COMPOSITES

The selection and use of composites on transport aircraft is an evolving

process. A variety of composites have been tested in both military and
commercial aircraft in the last 25 years (1). These composites typically
consist of a strong, stiff fiber such as glass, graphite or kevlar, and a
protective, adhering, inexpensive plastic matrix such as polyester or
epoxy.

Glass fibers embedded in a polyester matrix have been the dominate

composite for military and civil aircraft in the past. Currently, the
aircraft industry prefers the stiffer and higher temperature composites
made from carbon fiber in an epoxy matrix. However, the grade of carbon
fiber and epoxy seems to change from year to year and from airplane
manufacturer to airplane manufacturer. The current favored carbon fibers
are AS4 (Hercules/Hexcel), IM6 and IM7 (Hercules/Hexcel) . The AS4 is an
economical, high-strength carbon fiber and the IM6 &7 are high-modulus
expensive fibers. These three carbon fibers have been used on military
aircraft and in research, but are not on commercial vehicles. The T-800
fiber (a Toray equivalent to IM7) has recently been used in some
commercial applications (1-6).

Epoxies, particularly the 350F curing systems, are the least expensive
high temperature options for matrix materials. Several epoxy systems
have been developed and tested for use with specific fibers. There is a
current trend to use rubber modified epoxies such as 8552 and 3900 to
increase the toughness of the composite system and its resistance to
impact. Fiber-resin combinations currently in use by airplane designers
and researchers are:
AS4/938 (ICI Fiberite) -Boeing Advanced Composites Program Door
Panel(2)
AS4/8552 (Hexcel/Hercules), -Boeing Adv. Comp Fuselage (6-7)
AS4/8551 (Hexcel/Hercules) (6)
AS4/3501-6 (Hercules) -McDonnell Douglas Adv. Technology
Composite Wing program (8)
AS4/3502 (Hercules) Military Aircraft (6)

COMPOSITE STRUCTURAL ANALYSIS

Composite materials were originally added to the options in the
PDCYL program in 1995. This was done by simulating these materials by
homogeneous structures with uniform mechanical properties (strength
and modulus of elasticity) in every direction. This approach limits the
code to evaluation of only the simplest and weakest type of composites
called random mat1. Random mat composites are made by stacking the
reinforcing fiber in all direction throughout the thickness of the
material. In this type of composite the elastic properties and strength of
the layup are roughly the same in every direction but the fiber density and
reinforcement is low in any specific direction.

Random mat composites are not favored by aircraft designers because of

their low strength to weight ratios. The preferred type of composite for
these applications are ones in which the properties of the material are
customized to meet the specific directions and magnitudes of the
structural loads. This yields the minimum weight composite for the job.
To accomplish this, composite designers specify a layup pattern for a
composite laminate relative to a major axis of loading.

1 See Appendix A for definition of composite terms

A typical composite laminate is made of a stack of 4-16 plys. A ply is a
single layer of parallel reinforcing fibers embedded in a partially cured
matrix of plastic. The location of each ply in the stack is defined
relative to the angle its fibers makes with a major axis, such as the x-
axis. For instance, a 0/90/90/45/0 layup is one in which the fibers of the
outer and inner layers are parallel to the x-axis, the next two plies have
fibers perpendicular to this axis and the fibers of the third layer are at an
angle 45 clockwise to the x axis. This type of composite would have
reinforcing fibers to sustain tensile and compressive loads in the x and y
directions but would be weakest in the 45 direction. Composites walls
for structural parts such as aircraft are often made from stacks of these
laminates.

Analysis of a multilayer stack is more complex than that of homogeneous

materials such as aluminum or random mat and requires the use of a
macromechanics approach to determine elastic properties and strength.
The macromechanics approach used in the COMPOS part of the PDCYL code
is that presented in most textbooks on composite design (9-11) . In this
approach the stiffness of a particular laminate is calculated by summing
the contributions of each layer (ply) in the stack to the stiffness of the
laminate in a particular direction. The composite stiffness in each major
direction is then used to calculate the net strain of the composite in that
direction due to the applied loads. From the net strain, the strain on each
layer (ply) parallel and transverse to its fiber is derived. The resulting
strains are then compared to the failure strains of the ply material and
from this the potential for the failure of the stack is determined. The
details of implementing this approach in PDCYL are described in the
following section describing the COMPOS (composites) code addition.

COMPOS CODE
COMPOS is a section of code which has been added to PDCYL program to
calculate the minimum laminate thickness required to withstand the
forces imposed at each section of the airplane.
Assumotions within COMPOS
The laminate is symmetric and orthotropic. (This type of layup is
commonly used in aircraft design to minimize warpage of the layup).
Every ply in the stack is composed of the same resin- fiber material.
The stack is a minimum of 3 plys. (A ply is usually .003-.007 inches
thick depending on the material.)
 The modulus of the material is the same in compression and tension. (if
the compression modulus is different than its tensile modulus, the
smaller of the two values is selected for all calculations.)
Failure of the composite laminate occurs when any single ply fails.
Failure of a ply occurs when it reaches the maximum strain transverse
or parallel to the fiber direction in tension, compression or shear (1 1)
Maximum strain theory is invoked in this analysis because it is currently
believed to be the most predictive failure theory for composites (3,4,8)
The minimum gage thickness for the composite material is assumed to
be the thickness of the initial laminate (a stack of plies).
All loads are applied in the plane of the ply. This means that there are
no z direction loads in tension, compression or shear.
The buckling equations used in PDCYL to analyze the frames and
stringers made from homogeneous materials apply to these heterogeneous
materials. For buckling analysis the modulus of the laminate in the
direction of load is used. This is a very course assumption and maybe
somewhat optimistic for quasi isotropic composites manufactured with
adhesive joints but seems highly unlikely for symmetric orthotropic
laminates with heterogeneous properties. However, buckling analysis of
complex composites structures is still in the developmental stage.(12)

Calculation Procedure
Calculations for Compressive and Tensile Loads
Once the maximum tensile and/or compressive loads per unit width (Nx
and Ny) at any given aircraft section are determined in the PDCYL code,
they are transferred to the COMPOS subroutine. The effect of these
normalized forces on the composite laminate strain is calculated using
the following relationship for an orthotropic symmetric laminate (9) :

[N] = [A] x [Eo] (1)

Where:
[N] = Matrix of forces on the composite section (Nx, Ny and Nxy)
[A] = Stiffness matrix of the composite
[o] = strain matrix of the composite (ex, ey, exy)

The components of the stiffness matrix [A] are determined in the code
through the following relationship (9):

Aij=T_.n K=] (QBij)k(hk) (2)

Where:
QBij = component of each ply's stiffness in the i and j's directions
h = thickness of k ply
k= ply number in the laminate

The stiffness contributions, QB values, of each ply are determined from

the initial ply properties, El, E2, v12 and the ply angles, 0, specified by
the user in the input file for a particular laminate construction. (Here, the
"1" direction is taken parallel to the fiber and the "2" direction
transverse to the fiber).

Once the average laminate strain is determined from equation (1), this
strain is then transferred to each ply and transformed into strain parallel
and transverse to each fiber as well as shear strain. These strains are
then divided by the mating failure strains for the material (supplied by
the user in the input file) to determine the R value of the layup.

Rij= alleij/eij

Where:
alleij=allowable components of strain in principle ply direction
eij = components of strain in principle ply directions

If the R value for all plys in all the principle directions is more than 1,
the laminate thickness is adequate to support the load and is left
unaltered. If R is less than one on any ply in any of the principle
directions, the thickness of the laminate is increased by giving it the
value of it initial thickness divided by R.

Calculations for Buckling

PDCYL currently determines critical buckling loads from the modulus of
elasticity of the material. COMPOS calculates the modulus of the
laminate in the direction parallel to the buckling force and passes this
value back to PDCYL. As mentioned in the assumptions portion of this
report, the buckling calculation of PDCYL may not be valid for composites
as they were developed for isotropic materials. Little research has been
done on composites in buckling so the authors advise caution in
interpreting this result particularly with non-isotropic layups.

NON-OPTIMUM FACTORS
 Unfortunately, few all composite planes have been built so it is
difficult to find planes to use as checks for the composite section of the
code (8). The all composite planes listed in the literature (8) are:

Windecker Eagle in 1967 which was glass fiber reinforced

Learfan in 1981 which used glass, carbon and kevlar fibers
Piaggio Avanti in 1986 with carbon fiber parts
Beech Starship in 1986 with carbon fiber
Grob GF-200- all composite
Slingsby T-3A Firefly -all composite

A literature search and personal interviews failed to turn up much

information directly useful in determining non-optimum factors. (These
factors are used to multiply the results of theoretical calculations to get
weights of practical structures.)

One reference was found which had this type of data (12). In this
reference, a theoretical analysis gave 8640 pounds as the weight of a
composite wing box whereas the actual wing was estimated to weigh
11,284 pounds giving a non-optimum factor of 1.306. Using the non-
optimum factors for aluminum structures (13) this number can be used to
estimate non optimum factors for carbon fiber-epoxy structures. If it
is assumed that the non-optimum factors for the fuselage primary
structure increase in the same proportion as wing structure relative to
aluminum, and that the increments for secondary structure and non-
structural are the same for graphite-epoxy composites and aluminum,
then the following non-optimum factors for the composite result:

Primary i Primary &Secondary Total

Structure Structure IAssembly

Fuselage 1.792 2.329 3.010

Wing 1.306 1.666 2.059

There are many composite components in commercial and military

structure as well as some from research on advanced composites. It may
be possible to compare these components to predictions of the code.
REFERENCES
1. Vosten, L. F., "Composite Chronicles: Past Performance and Future
Prospects," Fourth NASA/DoD Advanced Composites Technology
Conference, 1993, Salt Lake City, Utah, vol 1. p 1
2. Russell, S., Vastava, R., Ley, R., Polland, D and Mabson, G., "Design Cost
Modeling of Fuselage Door Cutout Structure," Fifth NASA /DoD Advanced
Composites Technology Conference August 22-25 1994. NASA Conference
Publication 3294 Volume I, Part I. p 127.
3. Personal communication from Professor. Steve Swanson, Univ of Utah,
Mechanical Engineering Department
4. Personal communication from Professor Mark Tuttle, Univ of
Washington, Mechanical Engineering Department
5. Personal communication from George Lallas of Hexcel/Hercules
6. Personal communication from Jim Stearns of NASA Langley
7. Scholz, D., et al, "Material and Processing Development for Composite
Fuselage Sandwich Structure," Fifth NASA /DoD Advanced Composites
Technology Conference August 22-25 1994. NASA Conference Publication
3294 Volume I, Part I. p 257.
8. Hawley, A.V., "Preliminary Design of a Transport Aircraft Composite
Wing, " Fifth NASA /DoD Advanced Composites Technology Conference
August 22-25 1994, proceedings of which are contained in NASA
Conference Publication 3294 Volume I, Part I,p. 736
9. Jones, R., Mechanics of Composite Materials, McGraw-Hill, New York,
1975.
10. Tsai, S., Composite Design, 4th edition, THINK COMPOSITES, Dayton,
1988.
11. Agarwal, B., and Broutman, L., Analysis and Performance of Fiber
Composites, 2nd ed., John Wiley and Sons, New York. 1990.
12. Swanson, G., Wishart, R and Eastland, C, "Compression Test Results
for Stiffened Composite Fuselage Structure," Fourth NASA /DoD Advanced
Composites Technology Conference June 7-11 1993, Salt Lake City, Utah
NASA Conference Publication 3229 Volume I, Part I, p125.
13. Ardema, M.D. et al., "Analytical Fuselage and Wing Weight Estimation
of Transport aircraft, NASA TM 110392, May 1996.
 _.EEI [12J,Y _
COMPOSITES TERMINOLOGY

Random Mat- equal fibers in every direction

]_L_,,._._- equal fibers in orthotropic directions yield a composite with
identical properties in 2 principal directions.
J,_.y..ELm_.t.r._-A symmetric laminate is one in which for each ply above the
center of the stack there is and identical one at an equal distance below
the center. For instance, a 0/-45/90/90/-45/0 is a symmetric layup
but a 0/-45/90/ 0/-45/90 is not.
Quasi-lsotrooic- Layups which are designed to have only two independent
elastic constants, the modulus of elasticity and Poisson's ratio. These
materials have the same values in every inplane direction. To meet this
criteria fiber (ply) layups must have the following conditions:
Total number of plies must be 3 or more
Individual plies must have identical stiffness [Q] matrices and
thickness
Layers must be oriented at "equal" angles (if total number of
layers is n, than each layer is pi/n relative to the next). If the
laminate is constructed from several groups of laminates,, the
condition must be satisfied for each laminate group
Typical laminates which satisfy these rules : [0/60/-60], [0/45/-
45/90]
 Appendix C

Weights of ASA 2150 With Aluminum and Graphite/Epoxy Fuselage

 _Oo_ Lq
O O
4" 4"

GO
OD 0_, 0
,'_ 00 _o
Or, .,_ 0 o t_

,"1CN ,,-'t

000 o o

II II II II II

o o
4- 4-

co _0

(NCNO
_0

ddJ o o

II II II II II

0 C_ 0

E_

2 + + + + + 4

0_0_
0_00 GO

_D ,.D

0000 oo
II II II III 11 II

_ _ _ 0

_ "__ ____. _

o_ 0 0 O0 00000 _ 0

gJJJ oo II II II II II II II II II II II II II II II II II II II II II
II II II II II II

II II II II II II
n_ ,q
t,_ 0 0
0 "_
O_ ._ _ _ 0 _ _ "_ _ _

oo
00

r..
_ _0 _, _ ._
0000 oo o _ _ r_
_ 0 I_
II II II II II II
_ "_
OO_ 0 ,-1
_J
0 _
 oo _ oo

O0 O0 O0 O0

000000000000

_00_00000000
_00_00_00_00

_ O0
O0 O0
O0 O0
O0 O0
O0

00000000000000_000000000000
00000000000000_ 000000000000

OOOO _ 0000 0
0000 _ _0_ 0
0000 _ 00_ 0
0000 _ _0_

II II II II II II II II II II II II II II II II II II II II II II II II II II II II II II II II II
++ ++ ++ ++

O0 O0 O0 O0
O0 O0 O0 O0

oo_oo_oo_o
O0 O0

_oo_oo_
U
o _
+ +

E
0 0
_o o o_oo

oooo go oo
_oo _o
00 00 0 o oo _o _o _o _o _ _ _ _
_oo o_o o_
_o oo oo oo oo _ _ _o _o

oooooooooooo_oo_oo_oo_oo ___
_00_00_00_00000000000000 _00_00

_ _ _ _ _g _g _ _ oo oo oo oo
_ _ O0 O0
O0 O0 O0 O0 O0 O0 O0 O0 _ _ O0 00_

_ _ _g _ oo _ o
o o
o g_ g_
O0 O0 O0 O0 O0 O0 O0 O0
O0 O0 O0 O0 O0 O0 O0 O0

_ _ _o _o
 IN

o o o oo oo o o o o _ _ _ _ m o _ _ _ _ _

ooooooooooooo___

Dq

ooooooooooooooooooooooo

_66666666666686666666666
O_

C3

m o

ooooooooooooooooooooooo

66666666666666666666666

oooooo_____
m

E"

0 00000000000
0 000000_0000
0 0_0000_00_0

0 O_ _ O0

H r..

II II II II II II II II II II II II II

m_

o _

OH
 0 '_
_'r--

_m

000000000000000000 _
0
c;
_o
m

_o
H m,-_ C; o _g
_o o

ii1

o_o_ooo_ooo_ o_ o_ o

000000000000000000
.................. _ _ o

0 '_
m ',_

f,l

oooooooooooooooooo _
:o

g o
0
m r_ E_m
_0_r _.

_ . .

cN _m
_ . 0 _n
u_

"_ _ r_
B.,
0

_o

C_m
l.n
o

0_. _ o
_
0 I-_ I_

0 H .
,-1 o_
o

0 _ _ _ I_ _

_o__o__o 0_ _
0
o_6_ooo_ooooo_ooooooooooooo_oooooooooooooo

IIIIIIlllllllll

000000000000000000000000___0_0__

lllllllllll1111111111111111111111111111111111111
 ++++++++++++++++++++++++

OOQQOOOQO0
66666666o66o666666666666666666

0_0_0_0_

_0_0_

,,?7,,7,,

66666666_W 666dd6JJ66ddJJJ666666666666666

__ _o 0 ooooooooooooooooooooooooo_g?0

I I I

5_

H u'l

__ 0_" TM I I I I I I I I I I I I I I I I I

, _

_o.__

,___ I I I I I I I I I I I I
o _

p..

1_
_;o_o_ggNo_o_o_NN_No_NNo_ggo_ggg_ggNg_g
-t- + + + -I- + + + + + + + 4- + + + + + + + + + + + + + + + + +

___=0_o_.__o_

666_6666_
IIIIIIIIII
o t_l I_11 ' I I I I I I I I I I I I I I I I I I I i i i i i i

_000_

N ,._

,...1 m
 e%
m:
[.9 q
L) u_

0000000000000000000000000000,-4 0 0 0 0 0 0 0 0 0 0 0

o
_n

_g
0 0 0 0 0 0 0 0 0 0 0

IIIIIIIIIIIIIIIIIIIIIIIIII1_

0 0 0 0 0 0 0 0 0 0 0

IIIIIIIIIIIIIIlllllllll _ _

'''''77_

I _._

+******,

00000000000000000000000000000 _ _
IIIIII

m_
 oooooooooooo_o_o_o_o6oo_o6_o6oo6o_6o

000oo00000oo000000o0o0oooo0000o0o0oo0ooo0oo0o060
C_
('4

_0____0__0__0__0_ 0

Illlllllllllllllll}llllllllllllllllllltl

_______0___0_0_0_ _

IIIIIIIII111111111111_111111111111

_0_0_0____0_0__0__0__ _

ooooooooooooooooooooo_oo_oo_od_oo_o_o_o_oooo_
 Z

0000000000000000000000000000000000000000

b-l 0

_0 ........................................

_. ___0____0000___

H 0
:Z 0
.

_0

0000000000000000000000000000000000000000

DH Z

0 0000000000000000000000000000000000000000
0%

C_

0
0
0

Mm

0000000000000000000000000000000000000000

o _-i
 oooo

oooo
oooo

r_i-4-_oooo

o o_

m_ _ mq on

_ _oo_
o_o_

oooo

_H_oooo

oooo

_o_o_o_

D_oooo

0_0_00_0_0_000_0
000_0000000_00000_0 _
0_0_00_0_0_000_0 i

___ _

00000000000000000000

00000000000000000000
00000000000000000000_ _

_o_o_

_o_o_mo_ooooo_o

_)-,_ _o_o
I_ I-I
000000000000000000000000000000000000000000000000

000000000000000000000000000000000000000000000000

000000000000000000000000000000000000000000000000

000000000000000000000000000000000000000000000000

__0___0_0_000000_000_0_0000_0000000_
______00000_000_0_0000_0000000_

000000000000000000000000000000000000000000000000
000000000000000000000000000000000000000000000000

000000000000000000000000000000000000000000000000
 oo_oo
_oooooo

_d66dd_

H t9 [_ _O_O_O_
_O_O_

_m _O_O_O_

,_ ,_ o o o o o o

oooodooo

o_
_o _o_
_0_0000
.o_ _t _o_oooo
O_OO_O_
t- o (D
0000000 _O._
_O_ O_OOOOO
_o_
I.--t (D %O
U'%
,-4 ,-_

o o o o o o o o

o o o o o o o o r_l "_ _ U
_O_O_
_ _ rn

00 o o 0 o I_ o _
000 o o o _'_ o
LOOO0OO 1'_ o O O
_n

_0 0
_-_

%1_ _'-_ P- I'N oo "3

n_ H

0 0
OOO__
 0

_ 0

.k
.k

4, 0

.k

_ 0ooo

.k

.k

.k
'k

"k
z _

"k

t-4 "1_

0
*--4 "k .,.-t

0 '_
.,-I

0 "_

"k
t_

e-t

0
CD _

_._ _
 0
ooooooooooooooooo ooooooooooo ooo

ooooooooooooo ooooooooooo ooo

Xooooo

ooooo

o_o____

O_O_O_O_O_O_
U
ooooo

OO_O__OOO
_-I :Z .....
O r. .....
[_ O .....
_I_ O_ O t'- t_10 ['- r- ,-4 t,- O u_ r- o o u-_ [.7-I .....
cn O H .....
,-4

_ ca ....

r,1 ................ _g _ _::::

_ O ....
_I_ ............... tm _-_ O -,-I ....
r, ......... r.0 ..... IZ
:: :_ : : : : : :> _ _ ....
:_:::: :o _ _ o _._ _ .... E_
_I
_
M: _J
O
....
._
_-I ,_ ._

O
_ -,.,t

a)

r_
0
1_

,<

Q;

0 0
 0 000000 0 00 000000000000000

0)
.,.-I .,-I

_o_ ....... _5_ o .................

_oo_oo_o_o_
03

,,_ ,

O
n_ n_
ul 0 ul .,-t
C
.,-4

-,'4
. O 'O
I _ E
0 0 0 c, 0
a.J al= a,I ._ ,-_
_ 0 _ ._ ._ ,-1 .,_ _ O_ _ 0
0A 0 E-_ -,._ t.4 1.4
0 _' E
l_ 0 0
"_ 0
m _ t..) _
0 _' 0 _J

O O;
,_ I_
..g_ 0
O 0 _

1 m *.)
0
L;
0000000 0000 0
ZZZZZZZ ZZZZ Z

_0000_0

_0_0_00 _

C
0

t_
o ._

_ _ o

_0 _
 _J

o o

_o o ,-I o
rq 0 0
o u_

tN
L._

v-t r'_l ,_-t Ln

000 oo
II II II 11 11

_ C_ 0

0 0

O0 CO

t_,lxl 0
t_1 XI 0

oo
II II II II II

0_0_
0_00
0_00
0_00
0_0_

_JdJ de;
I
II II II II II It

O_U_

0000
0000 _

IN e_ __
_ _ -__4_._ ____ -_oooo ._ "_ ._

tl II II II II II
JJJ_ II II II II II II tl II II II II II II II II II II II II II II

II II It II II II

a_

_ U

c_ m

_ U

_o
X
_0
o _ _
-_ 0 i_
._ ,_ II II II II II II II ,-t I_ -,-t
m
00_ 0 'q C; m
U U

_ 0
 o
o o
o o
o

_0 _0 O0 O0

_6 _6 _6 _6

000000_00_00

_,_,__,
....
_00_00_00_00
_00_00_00_00

_66_66_66_66
%0
_ o
o co

_o_
0000 0 000_ 0
0 Ul 0000 0 0000 0
0000 0 0000 0

**_ _ oooo
_0_ o
0

0 _'_ 0 _. . _oo_oo_oo_oo
6666 6 oo_ 66 666 _ 6666 66 66 66 66

II II II II II II II II I_ II II II II II II II II tl II II II II II II II II It II II II II It II
_, _,,OoO, oo
O0 O0 O0 O0

OoO
oo OoO
oo
_oo_o_o
!!

o
o o
o _ o
o

C
0

o o o o

m r_ _0 _0 O0
O0 O0 _0 _0
x
0 0
e
_u
 _g _g
14 I+

o _o go go o o oo oo oo

gg _g g_ g_ _g _ gg g_
_ _ _o
, _o
0 ,
0 0 0 0 _0 _0 _0 _0

gg

og_go_
_00_0

o ,,o
o,, Ooo ,,oo
o,,ooo,,oo o,,oooo,, ,,,, ,, ,Jo ,,o
ooo II II II II II II II II II

gg o ooo
o oOoO OoO oOO ooo o oooo ooo OoO oo oOO
gg g o ooo
o ooo
o oOoO oooo oOoO oOoO oo oo oo oo
oo o o oo oo oo oo oo oo ._ _. oo
o

+ + + + I+ t+ I+ I+ I+ I+

oo oo _ _N
dg do do do oo oo oo oo
O0 O0 O0 O0 O0 O0 O0 O0

gg gg _g _ _ gg g_ gN
 _JJJJJJJJJJJJJJJ_jjjjj
_0

c_ 00000000000000000000000

_ .......................
o
p-
o_

_ 00000000000000000000000

00000000000000000000000

t'-
:n

o_

o _n

_o

,-4

o
o m:z_
u'1
o ooooooooooo t_
o ooooooooooo
o

-_,_o_.o._o__ P_

II II II II II II II II II II II II II

r_

_n _, o
o _

59

f]l
 +

0 ',_

II1
m:

000000000000000000 _ _

dd66d6_6dd_66_gg_ _N6 _

_D

_666_6_6o_66_666 _ o

_o _;

o
C_
0,_
m

o I

000000000000000000 _

0
m _-
m[ar-

_ _ .

ao

('4

_ _o

m _

o, _o _ . ._ _

_ _ o

t.n U) tn tn

0 _ _ r_ I_ _

_,

,@
 000000000000000000000000000000000000000000000000

000o000000000ooo000oo000000000oo006ooooooo000000

IIIIIllllllllllllll

__00_0_0_0_0___0__0__00_

IIIIIItllllllllllllllllllllltll

IIIIIIIIIIIIIIIIIIIIII1111111111111111t111111111
 +++++++

0000000000 o_o_ooooo_oo_o_do_ooooooooo

IIIIIIIIII

. , . , ,
000000000 t"- oooooooooooooooooooooooooo_66o
_ _ en

mg

Ill

.......... _ _

_ _0

.......... _

o'3 IIIIIIIIIIIIlllll

, ,g

Illlllll]lll
IIIIIIIII
0 o

I::l ._

0000000000
++++++++++++++++++++++++++++

_00_0_

I'N

?????????? _ lllllllllllllllllllllllllll
+++++++++++_+++

00000000000000000000000000000

000000__

_0_
f,%
m +_MM
C9_

0000000000000000000000000000,-4 0 0 0 0 0 0 0 0 0 0 0

_n

_o

0 0 0 0 0 0 0 0 0 0 0
__ZZ_Z__+_Z_Z__ _ _ _o

0 0 0 0 0 0 0 0 0 0 0

rllllllllllrllllllll Ill _ III

_m
I

_$_NNN$__NNNNNNNNNNN_
llllllll_lll_llllll__ _
llllllllll

_ o
_ _
_

00000000000000000000000000000 ++++++++++

.............................
00000000000000000000000000000
_IIIIII
 '____g__N_g_;_N_
IIIIIIIIIIIIIIIIIIIIIIIII

++++++++++++++++++++++++++++++++++++++++++++++

666_6666666666666666666_6666_6666666666666J
 Z

oooooooooooooooooooooooooooooooooooooooo

_-_ o
o

_o _ ...............................
m

_ o _

'_ o

_ o
_ un

o 0000000000000000000000000000000000000000

DH _

0000000000000000000000000000000000000000

f-I

o
o

_0 _ ........................................

0000000000000000000000000000000000000000

o
:Z c_

u')
66666666666666666866 _ 8666

z _

"9

_o_

oooo

m _6WWW64
_ooo0

00000000000000000000

_0_0000
0000_0_ 0_ _0_

0000000

_0_
O_

u ***_
_oooo

_ Om _0_0_0_

_u
NNNNNNNNNNNNNNNNNNN*

0
0_0
_-,_

-,
_n 2:0

D 00
 0000000000__000_0000000000000000000000000000

000000000000000000000000000000000000000000000000

O000000000000000000000000000OO000000000000000000

000000000000000000000000000000000000000000000000
000000000000000000000000000000000000000000000000

_666ddd66d_dd66d_6666d666_ddW666_d_d6666666666

_0_0_0__0_0_00_0000 0 00000_0_0_0_
 __oOO
oooooooo

__000

00000000

H L9 _'"

OoOoOOOOoOO _0_0_0_

_o

_4;gd_44S 8_ .o_ _.o_

_o_ i._oi,rl

_o_
'lf_.ll_Nf'.lf'l _)
H _o_
u'l

00000000 i

_A

_o_o_ Hen _o_o_

C_

_ggN_g_, _o _o

_M
_) re..l

_o0 _00
m_m
 I I

oooo

0000

0 ....
_000_
X
 0

o_

:6:"
_o_.

_-_0
 -_ _ ......

_0000000000000000 00000000000 000

6666_66_6_6 66_
_g_ggg
00000

__0__0_0_ __0_0_ 000

_ ...........................
__o_oo_ __o_ ooo

00000
_00000

CJ
00000

_0_0_ 000

in
D9
u

w-i
h ........... 0_ 0 .....

cn ____ _o
h

:::::::::::
.J

: : : : : : : : : : : _g
:::_ _ 0 ._
m
O_ _
o

_ _
i1) '__"
 oO_
0 r_

Z_

_0

O
J<

<D

-,-I OJ

0 0
E_ E_
 X 0000000 0_00 000000000000000

0
U 0

_0_0_0_ _00 _0___0

-,-t .,-I e_

1,4
1.4
Ill 0
,--I

Ul 0_0_ _00 _00_00_0_0_

:J
I1) I1) 0

r"

< < <

0

.,-t 0
" 0

OJ 0 u U ,.o_ _ ._
U r,.

0 r,. ,_ = _ _,_ ._._ _ o

0 0
0 ,4 .,_

0 _ 0 _J __ 0_
_ 00_ _ -_ _ _
0

..g_ 0

_-_
0
O_ _ _
0000000 0000 0
ZZZZZZZ ZZZZ Z

_o_o _ o _o o
o

_ooo_

_o_o_oo

0
.,-.I
A I;) _

U'3
o _
_0

_,'_ 0
'_._U_
_._

_..C _'_0 ._ "_0_0 _

 Appendix D

Approximate Methods of Trajectory Optimization

by

Mark D. Ardema
 Approximate Methods of Trajectory OptimiTation

Mark D. Ardema

Introduction

Application of optimal control theory in the form of the nltaximum' principle to aircraft
_,_wm, w,I.. vo I_ e pr61_le wl
trajectory optimization problems generally results in a two-poin_(2PBVP). The order of
this problem is double the number of state variables and the equations are always "haft

unstable." Many schemeshave been developed to numerically solve this difficult classof

problem, but all are unsuitable in a vehicle synthesiscode. Not only are they

computationaJly expensive,but they are non-robust and not user-fi-iendly.

What is needed in a vehicle synthesis code is a method that optimizes the trajectory in

one pass, that is as an integral part of the trajectory integration. The method must also be

robust and it should be easy to use and to interpret physically. The key to achieving this is

to use judicious approximations to reduce the functional optimization problem to a

function one.

In this report, two approximation techniques are reviewed and developed. The first is

the use of the Energy State Approximation (ESA). This well-known technique substitutes

the total mechanical energy for the speed as a state variable, and then neglects the altitude

and flight path dynamics relative to the energy dynamics. The second technique is the use

of Singular Perturbation Theory (SPT) to time-scale decouple equations of motion. These

two techniques are related, and in fact, the ESA may be viewed as an example of SPT
methods.
Trajecto_ _Optimization: the Maximum Principle

The equations of motion of aircraft flight, no matter what the assumptions (see

Appendix A), are of state equation form:

_x- L(x.__u)

where X__ _ 9_ _ is the state and U _ U cg_m is the control vector. Suitable boundary

conditions on the state vector components are prescribed. It is desired to find the

components of _.U along the trajectory so that

is minimized. It is assumed that time is free. The necessary conditions for optimal control

are provided by the Maximum Principle (NIP).

Theorem (the Maximum Principle): Introduce the variational Hamiltonian function

n

H= - ro+Z ,4
iffil

where the components of the adjoint vector, _,;, satisfy the differential equations

_.i= _H.i l, 17
(_X; '

Then, if __Uis an optimal control,

(a) U = arg max H

r_.r

Co) H=O

(c) Transversality conditions ("natural" boundary conditions on the _,; ) hold.

Thus, we must solve a 2n dimension 2PBVP in the states and adjoints; exactly n
boundary conditions are provided at t = 0 and the other half at t -- tt (due to the

travsversality conditions). The equations are unstable in the sense that if they are

linearized about a nominal trajectory, one-half of the eigenvalues will have positive real

pans and the other half negative (unless they are zero).

_Avpro_ximation Techniques

Methods of reducing the 2PBVP to a simpler problem will now be developed. These

methods focus on order reduction and are motivated by two simple observations.

First, we note that if all componets of_f, except possibly _ and the function fo are

independent of a specific state variable, say X,; and the final value of X,. is not specified,

then the corresponding adjoint is always identically zero and the state equation ._; = _

drops out of the problem (decouples from the other states). To see this, consider the is

adjoint equation and its travsversality condition:

0 f,. ,_, .Ai(tr)= 0

)_i = 8X i

The only solution to this problem for any finite value of 8 fi/SXi, is A; --- O.

Second, we note that if there is only one state equation, then the necessary conditions

can be used to eliminate the adjoint variable and thus the problem reduces from a

functional optimization problem to a function one. To see this, consider

x: :(x,u)

We have

H= -fo+Af
 A 8fo_A 8_._f_f
c_X dX

Applying the MP (assuming for the moment unbounded optimal control exists)

H=-fo+_ f=O

OU 8U dU

Eliminating _ gives

8of _f
+_fo= 0
8U 8U

for the optimal control. Alternatively, a direct approach may be used:

J-- f-}
Thus (fo/f) is to be minimized with respect to Uat constant X; this leads directly to the

equation for optimal control derived just above from the MP.

SPT provides an organized, mathematic, al way to view order reduction of differential

equations. Consider the initial value system

. = r(x, r) x(0)- x0

y"=s(x, v) (o)= ro

where s is a "small" parameter.

Since 6 is small, an approximate system may be expected to be

 x, - r(x,,r_)

o--g(x,,y_)

It can be proved that under certain conditions, the solution of this problem is a good

approximation to the solution of the original problem, except near t = Obecause the

boundary condition Y(O) = Y0 will be generally violated these.

The problem is that Yundergoes a rapid transition from its boundary condition to the

approximate solution at t = 0. To analyze this motion, the time scale is stretched by

T = t / . The resulting equations are called the boundary layer equations

dX=_r(X,r)
dT

dr
dT
=g(x,r)

Setting z = 0 to approximate these equations

dX
= 0 _ X = onst = X o
dT

dY_
=g(Xo,r_)

The solution to this equation approximates the desired solution near t = O. There are

matching techniques to combine these two solutions to give an over-all approximation, if

desired. The key observation is that a second order system has been replaced by two first

order systems, and each of them reduces to a function optimization problem.

 The SPT provides a convenient way to look at the energy state approximation. Define

the aircr_ energy per unit weight by

Differentiate and use the state equations in the Appendix

_ h + v _ v(:r,, - o) = P
g Mg

Where T is the component of thrust along V and P is the specific excess power. Note

that this equation is valid for all three sets of equations given in the Appendix.

Now replace V by E as state variable and use the observation that h and ?' are

capable of rapid change relative to E. This motivates writing

_'=p

_h "-* *

Setting = 0 then gives an order reduction of the equations of motion by two. We will

use this approximation, the ESA, throughout.

This approximation has a long history of successful application in a wide variety of flight

trajectory problems. The main drawback is that the variables h and 2" may now jump

instantaneously at points along the trajectory, as well as at the boundaries. These jumps

could be accounted for by boundary layer analysis, but this is not done in this report.
 ESA equations of motion for the three cases of interest are given in the Appendix.

these equationswillnow be used as the basisfordiscussingspecifictrajectory

optimizationproblems.

Mfinimmn T'unc/Fuel to Climb

Startingfrom equations (I)'

=-cr _(o)=_o

J_ =V

_'=p z(o)=Zo,Z(t,)=z,

L=mg

:=f(K.
+K.c-:),,

Here, the system functions and boundary conditions do not depend on Xand thus the

equation X = V drops out of the problem:

=-cr _(o)=_o

E'=p e(o)=
Eo,e(t,)=
e,

:=f(K.
+K.c_)a,

with P evaluated at L = rag. The system functions and boundary conditions now

depend on both E and M and hence neither state equation uncouples. Thus the MP must

be applied and a 2PBVP solved.

7
 To reducethe problemto oneof functionoptimization,SPTis used to further reduce

the system.

in = -CT

eE=P

Setting e = 0 gives a single state equation

n,- -cr _(0)=

T=D,L=W

The optimization problem is now (h and Eare controls)

:=(x, +K_) t:

With the obvious trivial solution t r = O. Also, because the system functions do not

depend on m and m(t r) is free, >t u =- O.

The boundary layer system for this problem with = 0 is simply

_'=p

with m = const, so that

: =J'(1<,
+ T
and the solution reduces to

8
 This is the well-known
assuming that _/KI + K2CT ) is positive and Eis monotonic.

enrgy climb path.

From now on we will assume "slowly varying" mass, that is that m is on a slower time

scale than E and thus its state equation may be ignored. It is also assumed that the throttle

is fixed.

F'med Range

This problem is the same except that range is fixed

x =v x(0)= Xo,X(t,)= x,

E=P =z0,z(t,)=zr

Here, 2 x = const :/: 0 so that the J_" = V state equation does not uncouple, and we have

a 2PBVP. To effect system order reduction, SPT is used.

.A'=V

s_'--P

H= -KI - K_CT + 2xV + 28P

x = -_=_ H 0 _ /1.x = coast

dX
 _H _(CT) Ax_ V 3. 8P

Setting # = 0 a problem with a single state is obtained

J( =V T--D

H = -K 1 -K_CT+I x V

Applying the MP:

T"D
h,E V / L=W _/rC

This defines a cruise point, characterized by C c, T c, and V c , in the flight envelope. By

proper selection of Kl and K2, this point can be made to closely approximate minimum

direct-operating-cost cruise.

For minimum time (K, = 1, K 2 = 0), the optimum cruise point is given by

Max
h,Z(V)

For minimum fuel consumption (K t = 0, K 2 = 1) it is

which is the classic Brequet cruise point.

10
 The boundary layer with s = 0 is

E=P

H = -K_ - K2CT + gxV + gsP

so that the optimal climb flight path is given by

h K_ +K2CT-_,x
P :)i s=:_

with ,l x as given above.

Maximum Twining W'_.hNo Thrust

We start with (2)' with T---(2

X'=V cosz

I7 = V sin Z

_'=F= VD
mg

L sin
Z=
mV

L cos _ =mg

II
with
L sin ,.

In this case the _', Y_and _, equations all uncouple. Changing to more convenient

variables:

.E = -V(B +C._ 2)

f = -_ _ fdt

wh_e

oJ= , -l<oJ <+I

tanCM

0_ = sec-'[min(C L limit, load factor limit)]

f = g tan C M
V

Thus

_f

J = S V(B +Coj_) dE

and the optimal controls are given by

12
which leadsto

(assuming righthand turn)

(v_:+:g_
(B+c_j-:vl,7-_+7-/_
_" :,_
B ac _j-o

where

D +DL , c=DmV2f2
Mg Mg 3

so that

The search for the optimum h probably should be done numerically.

Next, consider the same problem but using (3)'. Now, the X equation uncouples

but the I_and j_ equations do not. The coupling is of two types. First, through the

Coriolis terms, which are relatively small and can be ignored. Second, through the

centripetal terms, which are large at the start of descent trajectories from orbit.

13
 There are two ways to deal with this problem. First, the Coriolis and centripet_l

terms are ignored. This is justified because what is really sought is turning ability due to

banking and these terms mask this. Second, the _" and Z terms may be decoupled using

SPT.

Maximum Cross=Ran_

Next consider,using (2)'

I = -_ f:dt = -_ V sin Z dt

= Vcosz

I7 = Vsinz

_'=p

Lsin#
mV

L cos# = rng

As before, the .,_and I)" equations uncouple but now the jf equation does not. To

reduce this to a function optimization problem, further time-scale separation is required.

Putting Z on a slower time scale than _" gives the solution V = 0 and _x = 0.

Using equations (3) results in the same problems as for maximum turning.

14
 Appendix

EquationsofMot/on

The following axe the aircraft point-mass equations of motion under various

approximations.

(1) Flight in a vertical plane over a fiat, non-rotating earth; no winds aloft and

thrust aligned with velocity.

r_ = -CT

)C = V cosy

/_ = Vsiny

T-D- mgsiny
177

L -mg cosy
i,=
mV

otherwise the same as (l).

(2) 3-D flight,

rh = -CT

= Vcosy sinz

I_ = Vcosy cosz

15
 /7 = sin?'

T - D -mg sin ?,
m

Lsin_
Z=
rnVcos?,

cos_ -mg cos?"

mV

(3) 3-D flight over a spherical, rotating earth; no winds aloft, thrust not aligned

with velocity, terms in the square of the earth rotation ignored.

T
dT=
gsI_

I/= Tcosp cos (a +_')-D-gsin?,

m

= "rcosfl sin(a+6")+ cos#- cos?' +--cos?' +2mcoszcos Y

mV r R

fTcos,3sin(a + 6")+L 1 ' =" -Tcosrcosz

V Y

+2ca(tan ?, sin2"cosY- sinY 1

16
i_ = V sJn y

= VRcosy cos%
y.
T COS_
R

y= VRcosT" sin %
F

The following are the energy=state approximations of these equations.

(1)'
-- -CT

._'=V

.E _ V(T- D) _ p
mg

L=mg

(:)'
r_ -- -C'T

X = Vcosz

P'= Vsinz

_'=p

z= Lsin
mV

L cos_ = rag

17
(3)' (with T = 0 and m = coast)

L g V Y
0 = _cos_--- +-- +2oJ coszcos--
mV V r R

L V
k = --sin#---cosxtan Y----2m sin--Y
mV r R R

= VRcosz
Y
F COS
R

1_= VRsin Z
F

18