2016 - An Integrated Approach To The Preliminary Weight Sizing

Aerospace Science and Technology 58 (2016) 134–149
Contents lists available at ScienceDirect
Aerospace Science and Technology

www.elsevier.com/locate/aescte
An integrated approach to the preliminary weight sizing

of small electric aircraft
C.E.D. Riboldi ∗ , F. Gualdoni
Department of Aerospace Science and Technology, Politecnico di Milano, Via La Masa 34, Milano 20156, Italy
a r t i c l e i n f o a b s t r a c t
Article history: Electric propulsion has received attention in aviation as witnessed by studies in hybrid designs and by
Received 7 April 2016 the production of aircraft with support electric motors to be used in limited parts of the mission with
Received in revised form 27 June 2016 ancillary roles. Until the recent past, the main limit to a wider adoption of electric propulsion, which
Accepted 28 July 2016
besides having a lower environmental impact with respect to internal combustion engines (ICE) in terms
Available online 3 August 2016
of noise and emissions, can also improve reliability and on-board comfort, was the need for mass and
Keywords: volume-inefficient battery packs as devices for energy storage. However, thanks to the level of technology
Electric aircraft now reached by batteries, it is becoming possible to design and build electrically propelled aircraft at
Integrated design least in the category of light or general aviation. Due to the relative novelty of this technology, only few
Preliminary weight sizing examples of similar aircraft exist today, mainly modifications of more traditional concepts, and thinking
Optimal design of a completely new electric aircraft is made difficult by the lack of a consolidated design framework,
differently from the case of traditional ICE-powered models. This paper tries to cope with some basic
aspects typical to electrically propelled aircraft, to the aim of setting up a stable and reliable preliminary
sizing procedure allowing designers and aircraft companies to quickly size up and compare all-electric
designs. To this aim, a statistical analysis of the basic characteristics of existing aircraft is presented first,
showing a good correlation level between some of them. Next a method for the preliminary sizing of
weights is shown, obtained starting from a more usual step-by-step procedure typically adopted for ICE-
propelled aircraft. Due to the peculiar characteristics of electrically powered aircraft, the new procedure
involves an integrated use of the case-specific mission profile and sizing matrix. The validity of the
proposed procedure is testified by example analyses on two realistic designs of lightweight aircraft.
© 2016 Elsevier Masson SAS. All rights reserved.
1. Introduction tution. The main limit to the usability of the electric alternative
for propulsion has been bound to the limits of energy storage
Electric and hybrid propulsion systems have received a great systems, i.e. batteries, which especially for aircraft did not of-
deal of attention in recent years in various branches of transporta- fer until recently sufficient energy-to-mass and energy-to-volume
tion including aviation. This is due not only to the unpredictabil- densities [4] to be accommodated on board an aircraft without a
ity of oil price in this era, making a less oil-dependent source relevant negative impact on payload or aircraft size. Today, as a
of power more attractive for owners and operators in terms of result of many research efforts towards the improvement of such
cost and budget planning, but also to the better level of reliabil- performance indices, it is possible to design and fly an electrically
ity and economy attained by electric motors [1,2], as well as to propelled aircraft, as testified by some existing examples, both pro-
the improved comfort generated by less noisy electric motors with totypical and production models, in the categories of ultra-light
respect to internal combustion engines (ICE) [3]. and general aviation [5].
Nowadays, electric motors are generally more reliable than in- Among the factors limiting the diffusion of the existing models
ternal combustion engines (ICE), and their efficiency in converting
of electric aircraft is their relatively high production cost, which
stored energy into mechanical energy is much higher by consti-
will be only recursively lowered by the spreading of this technol-
ogy, through know-how consolidation and scale economy effects.
Also the cold perception by the potential customers plays a role
* Corresponding author. Fax: +39 02 2399 8334.
in the lingering diffusion of such systems. Especially private pilots
E-mail addresses: [email protected] (C.E.D. Riboldi),
[email protected] (F. Gualdoni). and flight training organization tend to be very cautious with re-
http://dx.doi.org/10.1016/j.ast.2016.07.014
1270-9638/© 2016 Elsevier Masson SAS. All rights reserved.
C.E.D. Riboldi, F. Gualdoni / Aerospace Science and Technology 58 (2016) 134–149 135
Nomenclature
AR Aspect ratio T Time of flight

C AC Purchase cost of aircraft TTC Time to climb
C bat Cost of batteries USD United States Dollars
CD Drag coefficient W bat Battery weight
CL Lift coefficient We Empty weight
E bat Energy of batteries Wm Motor weight
Er Energy required (propulsive) W pl Payload weight
FAR Federal Aviation Regulations
W to Take-off weight
ICE Internal combustion engine
e Battery-specific energy, Oswald coefficient
J acro Merit function for acrobatic class
g Gravitational acceleration
J tour Merit function for touristic class
h Altitude
K Drag-due-to-lift polar coefficient
p Battery-specific power
P bat Battery power
Pm Power of electric motor q C , q R , q RC Optimization weights
Pr Power required (propulsive) γ Climb angle
P SL Power required at sea level ηP Propeller efficiency
R Range of flight λ Energy-specific cost of batteries
RC Rate of climb ξ Power loss coefficient
S Wing reference surface ρ Density of air
SMP Sizing matrix plot τ Power ratio
spect to radically new technologies and prototype aircraft, when it existing – both prototype and production aircraft – clearly show
comes to risking a relevant capital. a statistical correlation on some key design parameters. This fact,
Also on the side of researchers and designers, the tendency to that will be suitably documented in the present paper, suggests a
treat electric aircraft as prototypes is testified by the lack of liter- design approach starting from the statistical analysis of what has
ature illustrating a common framework for preliminarily designing been done up to the present time. Subsequently, in the proposed
such aircraft, which in most cases are obtained through a mod- sizing procedure, differently from the well-known procedures for
ification of existing machines, originally gliders or ICE-propelled non-electric aircraft where the analysis of the sizing matrix and
designs. This is also due to some peculiar features of electric air- the sizing of the aircraft weights are basically independent pro-
craft, requiring an analysis of the design point which will not be cesses, these two areas of the design are intimately linked for
limited to those variables – basically empty and take-off weights, the case of electric aircraft, due to the peculiar construction of
wing loading and power loading – usually considered for sizing the mission-specific correlation between weights for such design
a traditionally propelled aircraft. Some works exist in the liter- case. This yields and integrated procedure for sizing the new air-
ature about the similar issue of the sizing of hybrid propulsion craft, where weights, wing loading and power loading influence
aircraft [6,7], but the futuristic assumptions concerning the per- each other, hence they need to be sized together.
formance of batteries and motors typically made in such analyses In a first stage, the paper presents a statistical database of exist-
does not allow to consider these works for an immediate practical ing electric aircraft types and illustrates the sizing procedure at a
outcome. Some more realistic past research efforts are focused on theoretical level. In a subsequent section two examples of quanti-
parametric studies for all-electric designs, starting from a design tative analyses are presented, showing the ability of the procedure
point which is already well characterized, and from well proven to produce designs matching acceptable requirements typical to
existing aircraft in the same weight class, thus proving its signif-
data concerning the power-plant and the general sizing of the air-
icance. In a final stage, it will be shown again through practical
craft obtained from an existing testbed [8,9]. Being based on pre-
examples how the design method bends itself to an optimal anal-
cise measurements from an assigned aircraft example, these works
ysis, thus making the potentially complex scenario of integrated
do not present methods sufficiently general to allow application to
design easier to deal with by means of a computationally-intensive
a generic aircraft with given desired specifications. Another part
approach. Finally, the approach and the results are recalled and
of the literature on all-electric aircraft is devoted to the design
critically discussed in the paragraph devoted to the conclusions.
and optimization of sub-systems on an assigned testbed, hence it
is more centered on a later stage of the design, where the de-
2. Preliminary sizing of electric aircraft
sign point for the aircraft has been already determined [10]. Very
little exists about the preliminary design of electric aircraft with 2.1. Database of existing aircraft
respect to assigned mission requirements, trying to bend the pro-
cedures typically adopted for ICE-propelled aircraft to this new Similarly to what is usually done for the preliminary sizing of
field of aeronautics. The lack of standard best practices among de- ICE aircraft [11,12], also for electric aircraft it is possible to set up a
signers and aircraft builders in turn fuels the lack of confidence by statistical analysis of the values assumed by some key parameters
customers, leaving the sector of electrically propelled aircraft in a for some designs already existing. To this aim, in a first stage of
condition of stagnation. the research the characteristics of some such aircraft have been
Trying to fill this lack in the existing literature, this script con- collected and analyzed. The fact that an aircraft has been flown
centrates on the existing technology and introduces a possible and the completeness of the available data on the electric plant
simple way to preliminarily size electric aircraft, borrowing much have been considered as criteria for the inclusion of a model in
from the preliminary design technique typically studied and imple- the database.
mented for ICE aircraft, but with some substantial modifications. Table 1 presents the values of some key construction parame-
A first fact that was noted is that the few electric aircraft already ters for the aircraft included in the database. For the sake of clarity,
136 C.E.D. Riboldi, F. Gualdoni / Aerospace Science and Technology 58 (2016) 134–149
Table 1
Database of electrically propelled aircraft already flying.
# Model W to / g [kg] W e / g [kg] W bat / g [kg] W m / g [kg] Pm [kW] e [Wh/kg] p [W/kg]

1 ElectroLight 2 [13] 315 188 34 7.0 19.4 163.2 795.6
2 LAK-17B FES [14] 550 246 32 7.3 35.3 131.3 910.9
3 Lange Aviation Antares 20E [15] 660 440 77 29.1 42.0 136.0 794.0
4 Lange Aviation Antares 23E [15] 850 496 77 29.1 42.0 136.0 794.0
5 Pipistrel Taurus Electro G2 [16] 550 253 42 11.0 40.0 113.1 952.4
6 UAV Factory Penguin BE [17] 21.5 9.83 4.41 0.650 2.7 145.0 807.1
7 Yuneec International E430 [18] 470 157 74 19.0 40.0 153.7 801.0
8 Silent 2 [19] 300 200 36 8.5 13.0 113.9 792.4
Fig. 1. Example aircraft in the database in Table 1.
note that in this paper all weights are considered as a forces,

hence they are measured in Newtons (N). When the corresponding
masses are considered in plots or tables, they are properly referred
to using the ratio W / g, where g is gravitational acceleration, as
in Table 1. This expression for mass stems from the usual rela-
tionship between mass and weight force. Masses are expressed
in kilograms (kg). Both mass and weight quantities are frequently
used in aeronautics, sometimes with slight inaccuracies in their
respective definitions. Here a distinction is made to preserve sci-
entific rigor as much as possible.
The considered aircraft models fall in the category of general
aviation or ultra-light aviation, whereas in terms of mission profile
they are designed mostly for tourism and sporting activities (see
Fig. 1).
Due to the particular nature of electric aircraft, it is necessary
to introduce some analytical relationships between the various
weights, which do not always have a counterpart in the design Fig. 2. Take-off (W to ) vs. empty (W e ) weight. Original data and regression curve.
of ICE aircraft.
The take-off weight W to and the empty weight W e are related
by the definition ies, thought to reach over three times the reported eav , D B by 2020,
have not been considered relevant in this work, which is focused
W to = W e + W pl + W bat + W m (1) on realistic design of aircraft at the current technological level. Fur-
thermore, the cost behind the technological effort needed to reach
where W pl is the weight of payload, W bat the weight of batteries,
the predicted figure is hard to quantify, hence accounting for it
and W m the weight of the electric motor.
would make further analyses less safe in terms of cost assessment.
Other quantities relevant to the analysis reported in Table 1 are
From Table 1 it is apparent that the number of aircraft entries is
the energy density and power density of the batteries for each
not very high, and this is obviously a potential limit for the present
of the mentioned existing aircraft. These quantities are defined
E g P g analysis. On the other hand, from Fig. 2 it is possible to notice that
respectively as e = Wbat and p = Wbat , where g is gravitational ac-
bat bat the empty and take-off weights for all considered entries in the
celeration, E bat the stored energy and P bat the peak output power
database fall in a limited range and are also well correlated. This
of the battery. These parameters are specific to the technology
highlights an existing relationship between W e and W to for air-
of the battery [5]. In this regard, it is noteworthy that the aver-
craft in the considered weight category, which comes in the usual
age value of the specific energy for the aircraft in the database is
logarithmic form (see Part I of the book by Roskam [11])
eav , D B = 139.75 Wh/kg, with a small relative standard deviation
σ (e )
of e D B = 11.6%. These values are typical to Li-ion batteries cur- log( W to ) = A + B log( W e ) (2)
av , D B
rently manufactured for general purpose applications, and eav , D B thus appearing linear on a logarithmic plot as in Fig. 2.
is much less than what is usually assumed in the literature when Other parameters for which a statistical correlation appears to
dealing with future designs [20,21]. Information concerning futur- exist are weight and power of the electric motors. In this case, the
istic values of energy-related quantities characterizing the batter- best match is with as semi-logarithmic curve, in the form
the case of electric aircraft, when the weight of the aircraft is the
same during all the mission.
In order to retain most of the approach used in the traditional
non-electric design scenario, it is necessary to obtain a relationship
between the weights in Eq. (1) and the mission requirements. In
this section we show a possible way of computing the effect of
some key mission parameters on the weights and especially on
W bat , passing through the definition of the amount of energy and
power required for each phase of the flight.
2.2.1. Mission analysis through flight mechanics

Going back to the considered mission profile, the most pro-
longed phases of the flight are climb, cruise and loiter. The per-
formance can be characterized for each of them by referring to
basic static equilibrium equations from flight mechanics [22].
Concerning climb, a typical requirement can be defined in
Fig. 3. Motor weight (W m ) vs. power ( P m ). Original data and regression curve. terms of a target constant rate of climb RC (see [11] Part I). The
equilibrium equation in presence of a non-null vertical speed can
be written as in Eq. (4)
log W m = C + D P m , (3)
where P m is the power of the electric motor. 1 3
P rclimb = W to RC + ρ climb V climb SC climb
D (4)
Fig. 3 shows the data and the correlation corresponding to 2
Eq. (3) on a semi-logarithmic plane.
It can be noticed that the similarity between the values of where the average air density ρ climb during climb, the constant
installed power of most entries in the database does expose fur- airspeed V climb and the value of the drag coefficient C climb D need
ther analyses to the need to extrapolate for a motor weight or to be known in order to compute the power required for climb
power higher than those considered. The correlation tends to di- P rclimb .
verge rapidly to unrealistic values of motor weight for power val- Concerning the value of the drag coefficient, it can be esti-
ues slightly higher than those considered in the database. To avoid mated by firstly guessing the polar of the aircraft. In preliminary
this, the correlation in Eq. (3) will be used only on the domain design where very little of the aircraft is known, the procedure
consisting of the values between minimum and maximum installed proposed by Roskam can be applied for estimating the coefficients
power in the database, whereas out of this domain another realis- of the polar in the usual parabolic form C D = C D ,0 + K C L2 (see [11],
tic relationship will be introduced later for sizing. Part I). The procedure calls for a guessed value of aspect ratio AR,
which can be estimated from similar aircraft, and a possible level
2.2. Integrated analysis of the mission profile and design constraints of refinement of the aerodynamic design, influencing the value of
parasite drag C D ,0 through empirical correlation. Estimates for the
1
In order to correctly exploit the data in Table 1, the present values of C D ,0 and K = π ·AR ·e , where e is here the Oswald coeffi-
work is mainly focused on small aircraft, ideally for cross-country cient, can be corrected making use of the same method accounting
flight or training, thus matching the type of aircraft considered also for flap deployment and extended landing gears, thus ob-
in the database. With this in mind, taking into account the typi- taining polars for various aircraft configurations, including clean,
cal mission of existing aircraft in the database, it was decided to take-off and landing.
consider in the mission profile five usual phases: take-off, climb, In the case of climb, cruise and loiter a same polar referring to
cruise, loiter and landing. a clean configuration can be assumed. Provided C clean D ,0 and K
clean
For ICE-propelled aircraft, the well-known sizing method by have been estimated somehow, it is possible to compute the value
Raymer [12] (see Chapters 3 and 6) makes use of the definition of the lift coefficient from vertical equilibrium for an assigned air-
of take-off weight as a function of fuel weight and empty weight speed, as
plus payload, and tries to link the decrease of weight from take-off
to landing to the flight mechanics parameters of each phase of the 2W to
CL = (5)
flight. To this aim the same method makes use of fuel fractions, ρV 2S
i.e. ratios between the weight of the aircraft at the beginning and
at the end of each phase of the flight. Fuel fractions are defined thus yielding a value for C D from the analytic polar of the aircraft
as analytical functions of range and endurance as well as other in clean configuration.
aerodynamic and engine-specific parameters for cruise and loiter Alternatively, once the polar is known, a given C L can be im-
respectively, or explicitly guessed for shorter phases like take-off posed, theoretically capable of optimizing some performance in-
and landing. By solving a 2-by-2 system between the regression of dex, like time-to-climb or distance-to-climb. In such case it is al-
historical W to vs. W e data (Eq. (2)) and the analytical relationship ways possible to compute C D from the analytic polar, using Eq. (5)
between the same quantities obtained from the mission profile, it to find the corresponding V for equilibrium, and finally feeding the
is possible to find in a standard design scenario values for W to and equation for required power (e.g. Eq. (4) for climb) to find P r for
W e representing a feasible design solution in terms of weights that the considered phase.
also accomplishes the mission goals. It should be remarked that using the take-off weight W to in
This procedure is clearly not applicable to the case of electric Eq. (5) for every phase of the flight is not an approximation, dif-
aircraft, due to the time-varying fuel weight being ideally sub- ferently from the case of ICE-propelled aircraft, due to the fact
stituted by the fixed battery weight in the definition of take-off that this quantity is constant for the whole duration of the flight
weight (Eq. (1)). Furthermore, the requirements for the various for electric aircraft. On the contrary, ρ climb is representative of all
phases of the mission cannot be translated into the assignment altitudes crossed during climb, hence using this value brings in
of fuel fractions, which are themselves meaningless variables in some approximation. However, for the considered class of aircraft,
typically unpressurized and flying at relatively low levels, this ap- In analytical terms, the value of battery weight from the mis-
proximation is still acceptable in this preliminary phase, provided sion profile W bat, M P can be defined as
a proper value is assumed for ρ climb .

The energy required for the climb phase can be computed g E climb + E cruise + E loiter
through an estimate of the time-to-climb TTC, which can be de- W bat, M P = max ,
ηP e
fined as
max{ P rclimb P rcruise P rloiter }
hcruise . (12)
TTC = (6) p
RC
where hcruise is the cruising altitude. Therefore, the energy required Concerning the electric motor, the historical regression pre-
for the climb phase can be obtained by definition as sented in Eq. (3) can be used. As previously stated, due to the
peculiar shape of such regression curve, yielding extreme weight
E climb = P rclimb TTC (7) values for required power slightly above the top database en-
The cruise and loiter phases can be treated in a similar way to tries, the following procedure has been adopted to cope with
get the corresponding required energy and power values. For both higher
values of required power, based on the value of P r =
cruise and loiter, the power required comes from static equilibrium max P rclimb P rcruise P rloiter :
in the longitudinal direction as
Pr Pr
1 if ∈ P min, D B , P max, D B : W m = exp C + D
Pr = ρ V 3 SC D (8) ηP ηP
2
Pr
where ρ , V , C D and hence P r will be typically different for the if > P max, D B : W̃ m = exp C + D P max, D B
ηP
two considered flight phases. For cruise the usual requirements
come in the form of an assigned airspeed V cruise and range R. The Pr
τ= / P max, D B
resulting time for cruise will be ηP
R W m = W̃ m τ (13)
T cruise = . (9)
V cruise
The procedure in Eq. (13) is of course arbitrary in its analytic
It should be noted that differently from the climb phase, where form, but it tries to model the increase in W m with P r more
an approximation was made in the computation of TTC assuming realistically than with a simple extrapolation. The proposed proce-
a single value of density ρ climb for all altitudes during climb in dure can be interpreted hypothesizing that an increase of required
Eq. (6), the definition of the value of T cruise in Eq. (9) is exact. power over the maximum recorded in the database for a single
As weight does not change during the flight, differently from ICE- engine can be coped with through a milder increase in weight,
propelled aircraft, for electric aircraft the value of cruise speed can bound to power through a linear relationship instead of an expo-
be maintained constant without power adjustments for all the du- nential one. This is in accordance with the fact that over a certain
ration of the cruise. amount of required power it is typical to increase the number
For a given V cruise , using vertical equilibrium (Eq. (5)) it is pos-
of smaller engines instead of implementing an overloaded single
sible to obtain C Lcruise and from the clean polar of the aircraft the
one.
value of C cruise
D . The energy required for cruise is defined as As previously pointed out, the proposed procedure allows to
compute explicitly the values of the weights for the propulsion
E cruise = P rcruise T cruise . (10)
system, i.e. W bat and W m . The relationship between empty weight
The loiter phase can be treated formally in the same way as W e and take-off weight W to accounting for W bat and W m comes
cruise. As usually the loiter time T loiter is part of the requirements, from the definition of take-off weight in Eq. (1). A very relevant
it is easy to compute the value of the energy required for this part difference with respect to ICE-propelled aircraft is that Eq. (1) for
of the flight as electric aircraft is explicit in both W e and W to , whereas according
to the usual sizing procedure for ICE-propelled aircraft (see Chap-
E loiter = P rloiter T loiter . (11) ter 3 in [12]) it should be possible to express W to as a function of
We
From the computation of the required energy and power for the the ratio W , i.e. the product of the fuel fractions for each phase
to
various phases of the flight it is possible to estimate the weight of of the flight, in turn obtained from the analysis of the mission pro-
the batteries and motor that need to be installed in order to allow file. In other words, the weight of fuel is not present explicitly in
flying the intended mission profile. In order to do so, values of the solution of the design point in terms of weights for ICE-aircraft,
energy density e and power density p need to be assumed. It is whereas the weight of batteries is in the case of electric aircraft.
possible to set these quantities to realistic average values obtained Therefore, in the case of electric aircraft it is necessary to guess a
from the database of existing aircraft. Batteries will be designed on value of W to in order to compute W e or vice versa from Eq. (1),
the most stringent requirement among those bound to energy and and the difference between these values is due to W bat , W m , com-
power, whereas the weight of the motor will be sized using the puted as illustrated above, and an assigned W pl .
statistical regression previously shown (Eq. (3)) on the basis of the The take-off and landing phases have not been accounted for
value of required power. explicitly in the proposed analysis of the mission profile. This is
It should be observed that Eqs. (4), (7), (8), (10) and (11) re- due to their usually very low incidence with respect to the other
fer to energy and power required from the viewpoint of flight phases of the flight in the sizing of battery weight. Such incidence
mechanics performance. In order to translate these desired perfor- has been investigated in this work with simplified but very conser-
mance figures into energy and power requirements for the electric vative energetic models, and found to be responsible in any case
motor and batteries, it is necessary to account for the propulsive for less than 1% of the total W bat . Therefore, take-off and land-
efficiency η P < 1 of the propeller. This yields a higher requirement ing may be safely accounted for by simply increasing the value of
on batteries and motor than what is obtained from pure flight me- W bat computed based on the analysis of the other three phases by
chanics analyses. a factor of 1.02.
2.2.2. Design constraints from sizing matrix plot

Two further remarks concern the equations involved in the
analysis of the mission profile. Firstly, Eq. (5), which is instru-
mental in computing the power required for the various phases
of the mission profile, depends on wing loading WSto . Secondar-
ily, the equations for required power (Eqs. (4) and (8)) depend
explicitly on the reference surface S, but they can also be rewrit-
ten as relationships between power loading WPto and wing load-
W to
ing S
.
Similarly to the case of ICE-propelled aircraft, the explicit so-
lution of the design point in terms of W e and W to can be ob-
tained by solving the system constituted by the historical regres-
sion Eq. (2) and the definition of weight Eq. (1). Differently from
ICE-propelled aircraft, wing loading WSto must be specified in order
to complete the computation of weight using the mission equa-
tions typical to the mission profile. In the conventional scenario
Fig. 4. Sizing matrix plot (SMP) and historical data. Blue solid vertical lines: landing
of ICE-propelled aircraft a target value for the wing loading can for various C L values. Red dashed lines: take-off run for various choices of C to
L . Black
be defined together with power loading WPto based on the analysis dash-dotted lines: climb (two in take-off configuration with assigned climb angle γ
of the sizing matrix. The target power loading is defined by con- and rate RC , one in landing configuration with assigned angle γ as specified by FAR
straints pertaining both to typical maneuvers encountered during Part 23). Black dotted line: climb with assigned speed and rate RC from mission
profile. Cyan dotted lines: cruise and loiter. Magenta dashed line: envelope. Red
the mission, including those considered in the mission profile, and circles: historical data. (For interpretation of the references to color in this figure
to regulations (i.e. certification constraints). legend, the reader is referred to the web version of this article.)
Due to the fact that wing loading is a significant parameter
for both the sizing matrix analysis and for the weight estimation
procedure described above, which is typical to electric aircraft, it to be assigned. Furthermore, altitudes for all maneuvers need to
makes sense to consider an integrated sizing approach where the be guessed.
value of the wing loading specified in order to comply with sizing Fig. 4 presents a possible SMP for a prototype mission for which
matrix constraints is used as a constraint for the explicit computa- the constraints have been specified based on the investigation of
tion of the design weight. Once the wing loading has been defined, the characteristics of the aircraft in the database. In particular, the
for a given take-off weight W to it is possible to obtain the cor- take-off ground run distance has been set to 200 m at sea level,
and cruising and loiter altitude has been set to 1500 m. Polar
responding reference surface S, so that the computation of power
coefficients have been obtained imposing an AR = 30 and a cer-
and hence of energy required for the mission profile can be carried
tain level of cleanliness of the aerodynamic construction, that is
out explicitly, following the procedure presented in the previous
typical to gliders or lightly powered aircraft. Furthermore, no re-
sub-section, leading to an estimation of all weights. The result of
tractable landing gear have been hypothesized, and no deployment
the weight sizing procedure will be implicitly compliant with re-
of flaps on take-off, that is typical to smaller aircraft. For land-
spect to the constraints expressed in terms of power loading and landing
wing loading from the analysis of the sizing matrix. ing, stall speed has been set to V stall = 40 kn (= 20.4 m/s) at
sea level. Some curves in Fig. 4 are parameterized, in particular
The analysis of the sizing matrix can be setup considering the
for different values of C L ,max for the landing constraint and for
potential certification framework for the aircraft of interest here.
various C to
L for take-off. Climb, cruise and loiter speeds V
climb
=
Once again, it is possible to refer to the database to see that a
potential certification class for the aircraft in it is that of general 48 kn (= 24.7 m/s), V cruise
= 90 kn (= 46.3 m/s) and V loiter =
aviation aircraft. Therefore, FAR Part 23 rules can be used as guide- 81 kn (= 41.7 m/s) respectively are assigned for the correspond-
lines to setup the equations of the sizing matrix. ing phases. Further quantities assigned are the vertical speed in
climb in clean configuration, RC = 400 ft/min (= 2.02 m/s) and
Seven constraints should be considered, namely take-off, land-
the average density during climb.
ing, climb in take-off and landing configuration, climb at a speci-
The curves representing the constraints on the sizing matrix
fied speed and rate of climb (from the mission profile), cruise and
plot define an area of compliance with respect to all constraints.
loiter. The analytic expressions of such constraints are the same
With the information available at this stage, a possible rule for
for electric aircraft and conventional aircraft, hence they can be
choosing a design point on the SMP is that of maximizing wing
found in the literature [11] (see Part I) for those coming from cer-
loading, thus reducing the size of the reference surface, and hence
tification procedures, or by manipulation of Eqs. (4), (5), (8) for
of the aircraft, for a given W to . Secondarily, it is possible to set
those coming from the mission profile. The curves corresponding
power loading to the highest possible value, in order to assure
to each constraint can be plot on the plane WPto vs. WSto (sizing
the lowest power required with respect to a same take-off weight,
matrix plot, SMP, see [12] Chapter 19 and [11] Part I).
hence reducing motor weight and cost.
Some requirements for the aircraft need to be guessed in order
The values of WPto and WSto for all aircraft in the database
to set some parameters in the equations of the constraints coming
have been reported on the plot in Fig. 4, showing that all aircraft
from regulations as well as from the mission profile. For take-off, fall in the space of solutions defined by the adopted constraints.
the ground run distance needs to be assigned. For climb in take- A propulsive efficiency η P = 0.86 has been hypothesized for all
off and landing configurations the respective polars of the aircraft database aircraft. Some of those aircraft are actually far from the
must be assigned, and they can be obtained at this stage using borders of the space. This only means that the numerical specifica-
the procedure proposed by Roskam and already mentioned above tions used for the plot of the constraints are less demanding than
for obtaining C clean
D ,0 and K
clean
. For the constraint represented by those considered for the actual design of some of the considered
cruise speed, it is necessary to know V cruise . For the constraints aircraft.
coming from the mission profile, the clean polar, climb, cruise A more thorough study of the SMP will be presented through
and loiter speeds, and the vertical speed RC during climb need some examples in the section devoted to the results.
Fig. 5. Work-flow of integrated preliminary sizing procedure. (For interpretation of the references to color in this figure, the reader is referred to the web version of this
article.)
2.2.3. Selection of the design point through integrated analysis has still to be compared with the energy requirement in order to
As explained above, in order to close the problem of finding check what is the governing value in the computation of W bat .
W to and W e it is possible to specify a value of WSto , providing a The work-flow diagram in Fig. 5 summarizes how the inte-
way to explicitly write the equations for the mission profile and grated procedure for preliminary sizing shown in this section can
computing W bat . Such value can be obtained from the analysis of be setup.
the SMP, so that all constraints on the plot be implicitly satisfied In particular, it should be noticed that there are two main op-
in terms of wing loading. In practice, the assigned value of wing erational blocks (in dark blue in the figure), one representing the
loading can be used to find the value of the reference wing area sizing matrix plot and producing design wing loading and power
W to
S = W to / S
, thus allowing to compute W e for a given W to loading values, the other representing the explicit weight sizing
through the power equations of the mission profile. procedure, resulting in design values for all weights in Eq. (1).
Similarly, a value of power loading WPto can be obtained from The equations for the mission profile are present in both main
the analysis of the SMP that satisfies all constraints from the mis- blocks, providing constraints on the SMP and on one of the sub-
sion profile and from regulation or required standard maneuvers. procedures in the weight sizing block. The latter is centered on
From the viewpoint of the procedure presented above for sizing assembling the general definition of weights for an electric aircraft
W bat , the power loading information coming from the analysis of (Eq. (1)), which is possible computing the battery weight and mo-
the SMP may be included as a constraint to the maximization writ- tor weight from the requirements of the mission profile. A 2-by-2
ten in Eq. (12), which consequently can be extended yielding system in the unknowns W to and W e is setup next, composed of
the weight definition (Eq. (1)) and of the historical regression of
g E climb + E cruise + E loiter W to and W e data from the database (Eq. (2)).
W bat = max ,
ηP e
3. Quality of the design point
max P rclimb P rcruise P rloiter W to / W to
P
(14)
p The choice of the design point is often the result of a pa-
rameterized analysis, intended also to assess the sensibility of a
W to
where the latter term W to / P
brings the level of power load- possible design solution with respect to changes to the mission
ing chosen to satisfy all constraint on the SMP as a constraint in requirements and constraints. From the computational viewpoint,
the computation of necessary power from the mission profile. It is the design point is obtained having the requirements as basic in-
clear from Eq. (14) that in case the top required power is due to put. Thus altering the requirements causes the need to re-run the
the requirements coming from the mission profile there will be no design procedure correspondingly. Luckily, the procedure presented
numerical difference between this maximum power and that com- above is based on light computations, which can be performed in
puted in the previous Eq. (12). In any case, the power requirement multiple instances in a short time on an average processor. For this
reason, the design procedure presented above bends itself to a pa- increase in weight, specifically W to and W bat in the models just
rameterized analysis. introduced (Eq. (15) and (16)). In order to help defining a design
In order to better assess the effect of some design parameters point in the space of the possible solutions, it is possible to setup
on quantities not directly related to flight mechanics, yet interest- an optimality problem based on a merit function J tour including
ing for the design, a possible performance index to be considered both range and cost, or analytically
is cost. This can be split into aircraft and battery cost.
A basic model for the manufacturing cost of small aircraft is J tour ( W to , R ) = q C (C AC + C bat ) − q R R (17)
constituted by a linear-quadratic function of the take-off weight where q C and q R are weights of cost and range respectively.
alone W to [23]. For the aircraft in the database this model may This means that if a change in the range requirement is accept-
be not highly accurate, due to its partly experimental nature. able, this new degree of freedom in the sizing problem can be used
Nonetheless, the values obtained are comparable with those ob- to steer the design point in order to obtain the maximum of J tour .
tained for other ICE-propelled aircraft of the same size and level of Similarly, for aircraft for which maneuvering performance is
complexity [24]. The equation for the aircraft cost yields more relevant than range, vertical speed can be used instead of
range to form a merit function J acro together with aircraft cost.
C AC = 4.649 · 101 W to + 2.85 · 10−2 W to
2
(15)
Hence for an acrobatic aircraft
with W to in N and the cost C AC in USD. Similarly to the case of
J acro ( W to , RC ) = q C (C AC + C bat ) − q RC RC . (18)
ICE-propelled aircraft, C AC does not include fuel – and hence bat-
tery – cost. where q C and q RC are weights of cost and rate of climb respec-
Concerning the cost of batteries, several statistical relationships tively.
can be found in the literature bearing somewhat scattered results In this case RC will represent a new degree of freedom in the
depending on the considered database [5,23,27–29]. In this work, problem, i.e. a tunable design requirement.
reasonable results have been obtained using the data of an anal- For both considered cases an optimal problem can be posed as
ysis by Kromer and Heywood on Li-ion batteries employed on the minimization of the respective J subject to some constraints,
hybrid vehicles [5]. This analysis links the specific cost of batter- or analytically
ies λ in USD/(Wh) to the ratio of power-to-energy of the battery, ⎧
P bat ⎪
⎪ Certification requirements
E bat
in 1/s. The latter can be computed in the procedure proposed ⎪
⎪
in the present work at the level of the mission profile analysis, af- ⎨ Standard maneuvers (take-off, landing)
ter having estimated the required energy and power for the whole min J s.t. : (19)
p ⎪
⎪ Mission profile
P
mission. The correlation between E bat and λ can be obtained based ⎪
⎪
bat ⎩
on a linear regression on the data proposed by Kromer and Hey- W to vs. W e regression of historical data
P bat
wood, such that λ = M + N E . In this expression the coefficients where in case J = J tour then the array of optimization parameters
bat
are M = 0.2362 USD/(Wh) and N = 0.0138 USD/W, whereas the is p = { W to , R }, whereas in case J = J acro then p = { W to , RC }.
average absolute normalized error between the data and the linear In Eq. (19) the first three constraints have an effect at the level
correlation is 1.76% – a very low value, raising confidence in the of the SMP, which translate into a choice of wing and power load-
adopted linear regression function. Therefore, an expression for the ing. The equations of the mission profile are further used together
cost of batteries C bat in USD can be written as with the historical regression W to vs. W e to compute the value of
C bat = e λ W bat / g . (16) all weights including W to , other flight mechanics parameters and
cost which appear in the merit function J .
As shown on the work-flow diagram in Fig. 5 the evaluation of It should be remarked that various other performance param-
aircraft and battery cost can be completed aside from the main eters could be included in the merit function as well, like for
computations, after the weights necessary for running the cost instance the cruising speed. The choice of the variables of interest
model have been found. is bound to the intended mission and specific design requirements.
A possible way to set up a parametric analysis is constituted by Furthermore, more degrees of freedom might be freed up together,
simultaneously studying more measures of performance, wrapped making the optimal analysis more comprehensive, but less intu-
together in a merit function. The value of this measure of merit itive. In the results section some hints will be presented about
can be mapped with respect to parameters relevant to the de- possible optimal analyses, with an accent on studying the behav-
sign, having an impact on the selected measures of performance. ior of the respective merit functions instead of concentrating on
Furthermore, provided the merit function features a suitable de- the search of the optimum – which thanks to the regularity and
pendence on its parameters, it is possible to seek for its optimum. the size of the problem is not a numerically demanding task. This
This will correspond to a stationary condition of the merit func- is made possible by choosing merit functions depending on two
tion, obtained with particular, optimal values of the parameters. scalar optimization variables at most, keeping the complexity of
If the parameters considered in the analysis can be considered the problem low and allowing a graphical analysis of the problem.
as tunable variables, i.e. they are not fixed and can be governed This in turn helps in the assessment of the validity, and usefulness
by the designer, it is possible to exploit the information obtained of the proposed optimal approach to sizing.
from the map of the merit function to select an optimal design
condition, corresponding to optimal values of the parameters. Such 4. Examples
criterion of optimality is often adopted for the choice of the design
point. The aim of the proposed results is firstly that of showing how
The approach to parametric analysis just described will be ap- the design procedures presented above can be profitably used to
plied to the case of two example designs by first building-up suit- carry out the preliminary design of small electrically propelled air-
able merit functions. Flight mechanics quantities that are usually craft. This will be demonstrated through two case studies, based
interesting to maximize for an aircraft in the touristic category are on different design requirements. Secondarily, through purpose-
those related to the cruise and loiter phases, in particular range. built parametric analyses it will be shown what is the behavior
An increase in this performance index usually comes together with of some key design parameters with respect to changes in some of
an increased take-off weight. Cost is a measure that penalizes an the constraints and mission specifications.
Table 2 altitude changes, whereas clearly no loss due to motor conversion

Specifications for sizing matrix plot for both considered example efficiency should intervene with altitude. However, due to the rel-
designs.
atively small span of working altitudes for the proposed aircraft,
Motor-glider Acrobatic this change is not expected to be dramatic, hence justifying the
landing
V stall [kn|m/s] 40|20.6 50|25.7 high value assumed for this coefficient.
Take-off run [m] 200 350 As the effect on available power due to changing altitude is
Take-off altitude [m] 3000 1500 accounted for separately, propulsive efficiency η P has been con-
hcruise [m] 3000 1500
sidered constant, neglecting the dependence on speed, as typical
V cruise [kn|m/s] 90|46.3 110|56.6
in the preliminary stage of the design. This choice can be further
hloiter [m] 3000 1500
V loiter [m/s] 0.9 Vcruise 0.9 Vcruise justified with low expected changes over the limited speed enve-
RC [ft/min|m/s] 400|2.02 2500|12.7 lope where the proposed example aircraft are intended to operate,
V climb [kn|m/s] 1.2 Vstall 1.2 Vstall and also with the chance to include variable pitch propellers in a
Power loss exponent ξ [–] 0.98 0.98 later stage of the design in case this was needed.
η P [–] 0.85 0.85
The preliminary values for the polar coefficients in Table 3 have
been obtained using the procedure presented by Roskam [11] (see
4.1. Specifications Part I). For sizing the parasite share in the total drag, assumptions
need to be made concerning the wet surface, based on the weight
The first example is that of an electric motor-glider. This cate- class of the aircraft, and on the level of refinement of the aerody-
gory is of interest for electric applications at the current technolog- namic design, according to the acceptable limits presented for the
ical level, due to the relatively low power and mild propelled-range corresponding aircraft categories. For both examples, flap deploy-
required. The second aircraft considered is a more maneuverable, ment has been considered for landing only. Furthermore, parasite
higher-power, faster, lower-range aircraft, intended for over-field drag for the acrobatic example will be increased due to a non-
hour-building in a semi-acrobatic role. retractable landing gear, typical to this category and not to motor-
Tables 2 and 3 show the requirements and polar coefficients gliders. The drag-due-to-lift coefficient is determined according to
1
respectively for both aircraft, necessary for preparing the respec- the usual formula K = π ·AR ·e . Possible values for the Oswald co-
tive sizing matrix plots. Most requirements have been set looking efficients are suggested by Roskam [11] (see Part I), whereas the
at the performance of existing ICE-propelled aircraft in pertinent aspect ratios have been hypothesized for both examples from ex-
categories [25,26]. isting ICE-propelled aircraft in the respective categories.
The power loss exponent ξ refers to the usual equation mod-
eling the loss of engine performance with altitude [22], yielding 4.2. Sizing matrix plot
ξ
P (h)
P SL
= ρρ(ShL) , where P S L and ρ S L are respectively power and air The SMPs for the motor-glider and for the semi-acrobatic air-
density at sea level. In the case of electric propulsion, such loss craft are presented respectively to the left and to the right on
has been hypothesized to be bound mainly to the performance of Fig. 6. The standard and certification constraints included in the
the propeller, which may be drifting from the design conditions as analysis are those for take-off run, climb and landing, whereas
Table 3
Polar coefficients assumed for both considered example designs.
Configuration Motor-glider Acrobatic

Clean Take-off Landing Clean Take-off Landing
C D ,0 0.0110 0.0310 0.1010 0.0230 0.0230 0.0930
C L ,max 1.5 1.5 2.2 1.5 1.5 2.0
e 0.83 0.83 0.75 0.81 0.81 0.71
AR 30 8
Fig. 6. Sizing matrix plot (SMP). Left: motor-glider. Right: acrobatic. Blue solid lines: landing. Red dashed lines: take-off run. Black dotted lines: climb from mission profile.
Cyan dotted line: cruising or loiter speed. Magenta dashed line: envelope. Blue star: chosen design point.
Table 4 Table 5
Design values of wing loading and power loading Performance requirements for weight computa-
for both considered aircraft. tion based on the analysis of the mission profile
for both considered example designs.
Motor-glider Acrobatic
W to Motor-glider Acrobatic
S
[N/m2 ] 600 800
W to R [km] 300 100
P
[s/m] 0.2 0.055
T loiter [min] 15 15
W pl / g [kg] 150 100
from the mission profile further constraints for climb as well
as cruise and loiter have been added as shown in the previous
methodological section (see Fig. 4 and the corresponding descrip-
tion). For take-off run and standard climb, excerpts from FAR
Part 23 for the case of single-engine aircraft have been considered.
For take-off they suggest a semi-empirical way to define the power
loading necessary for an assigned take-off run based on basic po-
lar coefficients, making use of a very simplistic energy method. For
climb, they suggest analyzing the power loading necessary to ob-
tain an assigned RC and climb angle in take-off configuration, and
an assigned climb angle in landing configuration (aborted land-
ing).
The curves presented on the SMPs of Fig. 6 have been obtained
for assigned values of the polars as in Table 3, and for assigned
design values of the lift coefficients for the various configurations,
especially for the curves representing take-off, climb and landing,
as specified on the plots. Such values are themselves the result
of a parametric analysis of the SMP for changing values of the Fig. 7. Result of weight sizing for both example aircraft. Blue: mission curve for
guessed polar values, similarly to Fig. 4. For the sake of clarity, motor-glider. Red: mission curve for acrobatic aircraft. Green: regression of historical
the plots presented in Fig. 6 feature only those curves correspond- data. Blue star: design point. (For interpretation of the references to color in this
ing to the data presented as specifications in Tables 2 and 3 having figure legend, the reader is referred to the web version of this article.)
an impact (or a near impact) on the resulting space of possible so-

lutions. Table 6
Weight components, required power and energy
As specified in the description of the sizing procedure, the
from the integrated sizing procedure for both
choice of the design points, shown as blue stars in the plots, considered examples.
is aimed at increasing wing loading while simultaneously rising
Motor-glider Acrobatic
power loading. This would reduce the size of the aircraft and the
power required. The two considered scenarios bear rather differ- W to / g [kg] 793 784
W e / g [kg] 379 375
ent results in terms of space of the solutions. The envelope for the W m / g [kg] 23 84
motor-glider extends to higher values of power loading, and for in- W bat / g [kg] 241 225
creasing values of wing loading it is limited by cruise, by the climb W pl / g [kg] 150 100
performance requirement from the mission profile, and by take-off P r /η P [kW] 45.3 162.6
E r /η P [MJ] 116.1 108.6
distance. On the other hand, the envelope for the acrobatic aircraft
is dominated by vertical speed performance coming from the mis-
sion profile, limiting the value of power loading to a much lower analysis of the SMP, it is possible to initialize the weight sizing pro-
level with respect to the motor-glider design. cedure based on the equations for the mission profile, illustrated in
It should be noted that these relative differences between the the methodological Section 2.2, and the historical regression curve
SMPs for the two designs are clearly in line with the respective in Eq. (2). The additional performance requirements needed for
roles of the two aircraft categories, and show that the required running the weight computation based on the mission profile are
specifications included in Table 2 are compatible with a realistic listed in Table 5.
solution in terms of wing loading and power loading. Table 4 sum- The plots reported in Fig. 7 show the result of the weight siz-
marizes the characteristics of the design points for both aircraft. ing procedure in terms of W e and W to . The curves corresponding
The design point for the motor-glider is close to the right wing to the missions of both considered aircraft are reported as blue
loading limit. This choice rises the requirement on engine power and red dash-dotted lines respectively for the motor-glider and
to some extent (the top value for power loading can be attained acrobatic aircraft, whereas the green dashed line represents the re-
for wing loading values less than WSto = 450 N/m2 corresponding gression on historical data. The design points, marked with stars,
to the climb limit of the envelope), but the decrease in wing sur- are on the respective intersections of the mission curves with the
face S thanks to a higher wing loading will be advantageous also regression of historical data. Notwithstanding the differences be-
in terms of drag forces and power required, besides reducing the tween the two mission requirements the weight sizing procedure
size of the aircraft. Furthermore, this choice is in line with respect bears remarkably similar results in terms of W to and W e . Table 6
to similarly designed ICE-aircraft. Differently from the motor-glider summarizes the design values for W to and its various components,
case, the design point for the acrobatic aircraft tends to simultane- as well as the power and energy required for flying the intended
ously maximize both power loading and wing loading. mission.
The power and motor weight values included in Table 6 show
4.3. Mission profile and weight sizing that the motors required for the two designs are very different. The
procedure for sizing the electric motor described in Eq. (13) bears
Following the work-flow diagram in Fig. 5, once the design a rather high value of W m for the acrobatic design. Such weight
values for wing loading and power loading are known from the value is not very realistic for a single electric motor. As the aircraft
Fig. 8. Battery energy, power and weight as functions of W to . Top plots: motor-glider. Bottom plots: acrobatic aircraft. Blue vertical line: design W to . (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.)
is intended to feature a single propeller – otherwise the sizing and from the design power loading from the SMP. The quality of the
cost evaluation procedures would be somewhat different, and the results for both aircraft is very similar, with the climb requirement
historical regression data less correct –, a scenario can be envis- being by far the most stringent of those from the mission profile,
aged where more motors, with individual weights summing up to and also in absolute terms – i.e. considering all constraints – as
the value of W m , transfer power to the same shaft. can be noticed also from the respective SMPs. The choice of the
The plots in Fig. 8 show the required energy and power for design SMP leaves a margin with respect to the climb requirement
climb, range and loiter as functions of take-off weight. Similarly from the mission profile. Contrarily to energy, required power is
to the mission curves presented in Fig. 7, these plots are obtained very different for the two aircraft.
running the mission profile analysis for different assigned values From the plot of the battery weight corresponding to varying
of W to . values of take-off weight in Fig. 8 it can be noticed that the most
From the energy picture, it can be noticed that the largest share stringent requirement on batteries for both designs comes from
of required energy comes from the cruise phase for both aircraft. energy and not from power, even though the acrobatic design re-
On the other hand, it is apparent that the amount of energy per- quires much more power and the two requirements come very
taining to the various phases of the flight is different for the two close to one another in this case. The prevalence of the energy
designs, the climb phase being particularly requiring for the motor- requirement is mainly due to the relatively low energy-to-mass
glider, due to its longer duration in the profile of that aircraft density eav of current Li-ion batteries, whereas the corresponding
because of the higher cruising altitude and lower required rate of power performance seems to be currently somewhat less critical.
climb. As can be noticed also from Table 6, the overall required It cannot be excluded that further technological developments may
energy is comparable for the two designs. change the picture, but as specified at the beginning of the paper,
Concerning power, the plots in Fig. 8 present the values pertain- this work is focused on the applicability of the presented method-
ing to each phase of the mission, as well as the constraint coming ology to currently feasible designs.
Fig. 9. Effect of power loading and wing loading on W to (left) and W bat (right). Top row: motor-glider. Bottom row: acrobatic aircraft. Magenta dashed line: SMP envelope.
Blue star: design point.
4.4. Quality of the design solution weight (Eqs. (15) and (16)) [24]. The dependence of battery cost
on battery weight expressed by Eq. (16) calls for the knowledge
The quality of the selected solution can be analyzed by intro- of the required power-to-energy ratio of the battery. Fig. 10 shows
ducing some metrics bound to quantities characterizing the design the total value of cost, defined as C AC + C bat , as a function of wing
solution and studying the effect on such metrics obtained when loading and power loading. The resemblance with the behavior of
changing the design point in terms of power loading and wing take-off weight is apparent, due to the fact the cost of the bat-
loading. The first interesting metrics are take-off weight and its tery is a low fraction of the cost of the aircraft for the considered
components. In Fig. 9 the behavior of take-off (left) and battery designs.
(right) weights is shown for the motor-glider (top row) and the
It should be noted that the scale of the plots for the two pro-
acrobatic aircraft (bottom row). The diagrams are contour plots of
posed aircraft both in Fig. 9 and 10 is largely different. The area
W to and W bat obtained running the central block of the procedure
of the respective envelopes is henceforth rather different in abso-
in Fig. 5 for various choices of power loading and wing loading.
lute terms. Due to the flatness of the plotted functions over the
Fig. 9 supports the choice of the design condition to the top-
SMP envelopes, this indicates that albeit a similar cost is encoun-
right corner of the envelope, corresponding to lower take-off and
battery weights. On the other hand, it can be noticed that for tered over the full span of the envelope in absolute terms for both
both aircraft the same weight results can be obtained for various aircraft, this corresponds to many different choices of wing and
choices of wing and power loading, i.e. for every point along a power loading for the motor-glider and to comparatively less fea-
line corresponding to an assigned performance level. Furthermore, sible design points in the case of the acrobatic aircraft.
in a relatively large area of the envelope around the selected de- Further information about the quality of the solution come from
sign points, the gradient of the weight solution is not very intense, a parameterized analysis where weight and its components are
showing that the area of choice of the design point is more rele- kept constant to the obtained design values, and some required
vant than the exact value of wing and power loading. performance of interest is changed. Fig. 11 shows the result of such
An often important quantity for the design is cost, which can analysis for the case of the motor-glider, considering range (left)
be hypothesized to be proportional to take-off weight and battery and cruising speed (right) as changing performance parameters.
Fig. 10. Effect of power loading and wing loading on total cost. Left: motor-glider. Right: acrobatic aircraft. Magenta dashed line: SMP envelope. Blue star: design point.
Fig. 11. Effect of power loading and wing loading on range R (left) and V cruise (right) for the case of the motor-glider, for fixed W to as in Table 6. Magenta dashed line: SMP
envelope. Blue star: design point.
Once again, these result show that a change in the requirements sures of performance. The analysis presented here has been based
aimed at increasing range or cruising speed would drive the so- on the merit functions J tour and J acro , which can be expressed as
lution towards higher wing and power loading. Even considering functions of the two couples of parameters ( W to , R ) and ( W to , RC )
these metrics, it is possible to say that the top-right part of the respectively, thus allowing showing their behavior through usual
envelope is the most attractive. contour plots and visually checking it.
The most relevant information which can be obtained from The value assumed by J tour for the case of the motor-glider
Fig. 11 with respect to the next analysis is the fact that both range can be seen to the left of Fig. 12, whereas the overall cost for the
and cruising speed show a gradient of relevant intensity with re- same design can be seen on the right plot. These quantities have
spect to wing and power loading. been evaluated for the selected value of wing loading and power
As for the previous analysis (Figs. 9 and 10), similar results can loading for the motor glider, hence all solutions here correspond
be obtained for the acrobatic aircraft with respect to range and to a feasible point from the viewpoint of the constraints included
vertical speed. in the SMP. The black line represents the locus of the solutions
for which the equality constraint constituted by the regression on
4.5. Optimal design historical data W to vs. W e in Eq. (2) has been satisfied.
As far as the optimal problem in Eq. (19) is considered, where
After having highlighted the degree of mutual dependency of the historical regression Eq. (2) is imposed as an equality con-
the design parameters, it makes sense to study the behavior of the straint, the black line is the actual space of the solutions of the
cost functions presented in Eqs. (17) and (18) to the aim of un- optimal design problem. For this reason, it is also clear that the
derstanding whether it is possible to check the level of optimality solution computed previously by applying the integrated design
of the selected design point with respect to a more comprehensive procedure and represented by the blue star in Fig. 12 should (and
variable, in the form of a merit function depending on more mea- does) lie on the black envelope line. In order to get a wider pic-
Fig. 12. Merit function J tour (left) and cost (right) as functions of W to and R for the case of the motor-glider. Black dotted line: W to vs. W e historical regression. Blue star:
design point. Red circle: optimum point.
ture of how the design solution would change in case the historical
regression constraint were turned into an inequality constraint,
stating from Eq. (2) that log( W e ) ≥ 1B (log( W to ) − A ), the plots in
Fig. 12 are extended considering those cases for which the empty
weight corresponding to a given take-off weight were greater than
what is obtained from the historical regression. For this extended
space of solutions the limit line is represented by the same black
line described above. Conditions for which W e is lower than what
is defined by Eq. (2) for an assigned W to are not considered, on
the basis of their lower safety – they would require an excessively
lightweight construction, with respect to the lightly powered air-
craft included in the database for which the ratio of structural and
power-plant weight to W to is already a critical issue.
In the actual computation of the merit function, both range and
cost weights q C and q R in the function are unitary, so as to ob-
tain a similar share of both components in the overall value of the
function. As can be seen to the left of Fig. 12, the value of the
merit function tends to decrease for lower values of W to , and for Fig. 13. Payload as a function of W to and R for the case of the motor-glider. Black
dotted line: W to vs. W e historical regression. Blue star: design point. Red circle:
values of R between 100 and 400 km for decreasing W to . A same
optimum point.
level of optimality can be obtained for a lower weight and a higher
range or vice versa. From the cost plot to the right of Fig. 12, again
Considering the plot in Fig. 13, the black envelope line corresponds
it can be noticed that this quantity increases mainly with W to ,
to the W pl / g = 150 kg previously assigned as a design constraint.
whereas it shows little sensitivity to R. This is due to the fact that
As already pointed, going down vertically from this line on the
the cost of the aircraft excluding batteries, which is bound to W to ,
same plot corresponds to violating the equality constraint repre-
is by far the highest fraction of the overall cost.
The position of the design solution, selected independently of sented by the historical regression W to vs. W e in Eq. (2). For each
the optimal analysis on J tour (i.e. in the previous paragraphs), is W to the difference between the value of W e obtained from histor-
shown as a blue star on the plots of Fig. 12. Considering the ical data and that computed from the mission profile (i.e. from the
adopted setup, metrics and tuning of the optimal problem it can be top sub-procedure in the central block of the work-flow diagram
said that the solution previously selected with the integrated pro- in Fig. 5) can be interpreted as a change to the weight of trans-
cedure for the motor-glider is not optimal, yet it is not far from the ported payload. A contour plot of the so-obtained W pl for every
optimum. The position corresponding to the optimum is marked condition on the plane R vs. W to is presented in Fig. 13, under the
with a red circle. If range is turned into a variable of the design same hypotheses on the values of W to of interest considered for
problem, i.e. a design specification that it is possible to tune where Fig. 12, again for the case of the motor-glider. Considering a fixed
necessary, then it will be possible to obtain a cost-range-optimal W to it can be noticed that a decrease in range R can be converted
design solution – in other words, if R is a tunable requirement, it into an increase in payload weight W pl . This is true especially for
can be chosen to obtain a design solution which optimizes simulta- higher values of the overall weight W to , whereas for smaller val-
neously cost and range, according to the definition of the adopted ues, possibly more typical to the considered aircraft category, the
cost function J tour , graphically pushing the blue star to the posi- advantage may be lower, as can be noticed by the lower gradient
tion of the red circle. of the contoured surface with respect to range in the leftmost area
Qualitatively similar remarks concern payload weight W pl , of the considered space of solutions.
which is a relevant design specification that may be considered as A similar analysis can be carried out for the acrobatic exam-
a tunable parameter for steering the solution towards an optimum. ple, considering in this case the merit function J acro . Being based
Fig. 14. Merit function J acro (left) and cost (right) as functions of W to and R for the case of the acrobatic aircraft. Black dotted line: W to vs. W e historical regression. Blue
star: design point. Red circle: optimum point.
on cost and RC , this measure is more meaningful for this kind of

aircraft. The following results have been obtained with unitary op-
timization weights q C and q RC . In Fig. 14 once again the black line
represents the locus of the design solutions satisfying the equal-
ity constraint on W to and W e obtained from the regression curve
of historical data. As for the previous case, the result of the inte-
grated design procedure (blue star on the plots) turns out to be not
optimal under the selected metrics, i.e. it is not cost-rate-of-climb-
optimal, but also in this case it is not far from the optimum, which
can be found for nearly the same W to but for an RC increased by
about 500 ft/min – see the red circle marking the optimum on
the plots of Fig. 14. Total cost is once again dominated by the air-
craft cost excluding batteries. For this reason, the contour lines on
the corresponding plot to the right are almost vertical, highlighting
the fact that according to this cost model the RC performance has
not a great influence on total cost. It has indeed an effect on the
weight and hence cost of the batteries, whose share in the total
cost is however limited for aircraft in this weight class. Fig. 15. Payload as a function of W to and R for the case of acrobatic aircraft. Black
Similarly to the motor-glider case, by changing the values of dotted line: W to vs. W e historical regression. Blue star: design point. Red circle:
optimum point.
one of the design parameters, it will be possible to steer the de-
sign point towards the optimum. Looking at the plot to the left of
Fig. 14, considering to allow a change in the requirement on RC , grated procedure mixing both design tools – SMP on one hand,
the position of the actual optimum (red circle) can be reached by and analysis of historical data and mission profile on the other –
pushing RC to a higher value than the one corresponding to the is introduced for the specific case of electrically powered aircraft.
design solution previously adopted (blue star), while keeping on The core of the sizing method is a new definition of the partial
the black envelope line representing the satisfaction of historical weights summing up to give the take-off weight, in turn allowing
constraints. to study the mission profile without making use of fuel fractions,
The analysis of payload (Fig. 15) bears results qualitatively simi- which are meaningless quantities for an electrically powered air-
lar to those previously obtained for the motor-glider, showing that craft. The new procedure couples the results of the analysis of the
an increase in W to for the level of RC of the solution selected sizing matrix plot with the weight sizing procedure, yielding an
through the integrated analysis, graphically marked with a blue integrated sizing procedure.
star, can be translated into a rapidly increasing payload. This is not Furthermore, the paper shows that an approach to weight siz-
the case for an RC in a range above 4000 ft/min, for which the ing based on statistical analysis of historical data can be still
gradient of W pl with W to is not so high as for lower RC values. exploited also for electrically propelled aircraft, due to the good
correlation between data pertaining to existing aircraft. Statistics
5. Conclusions are used not only for linking take-off and empty weight values as
mostly usual for ICE-propelled aircraft, but also for relating power
The present paper faces the need to establish a standard pro- and weight of electric motors, and again in the battery sizing
cedure for the preliminary sizing of electric aircraft. Starting from phase, when it is necessary to translate a power or energy re-
the well-known design methods based on the analysis of the sizing quirement into the corresponding weight of batteries. The database
matrix plot, and on weight sizing resulting from the simultaneous extends to a rather restricted aircraft category, implicitly limiting
analysis of the mission profile and of historical data – both tools the analysis to smaller aircraft. This has been accepted as a limit,
typically used in the case of ICE-propelled aircraft –, a new inte- with the advantage of making the results of the analysis safer and
attainable with current technologies – i.e. free from futuristic tech- [7] C. Pornet, C. Gologan, P.C. Vratny, A. Seitz, O. Schmitz, A.T. Isikveren, M. Hor-
nological projections. nung, Methodology for sizing and performance assessment of hybrid energy
aircraft, J. Aircr. 52 (2015) 341–352.
In order to show a possible way to fully exploit the pro-
[8] T.P. Choi, D.S. Soban, D.N. Mavris, Creation of a design framework for all-electric
posed procedure, which is straightforward and computationally aircraft propulsion architectures, in: 3rd International Energy Conversion Engi-
light, some computational-intensive analysis have been proposed, neering Conference, San Francisco, CA, August 15–18, 2005.
where the design is optimized by preliminarily defining a cost [9] S.M. Batill, M.A. Stelmack, X.Q. Yu, Multidisciplinary design optimization of an
function including cost and key performance of the aircraft. Cost electric-powered unmanned air vehicle, Aircr. Des. 2 (1999) 1–18.
is accounted for by considering cost models specific to small air- [10] B.A. Moffitt, T.H. Bradley, D.E. Parekh, D.N. Mavris, Design and performance val-
idation of a fuel cell unmanned aerial vehicle, in: 4th AIAA Aerospace Sciences
craft and to batteries at the current level of technology.
Meeting and Exhibit, Reno, NV, January 9–12, 2006.
A complete preliminary sizing is illustrated for two possible air- [11] J. Roskam, Airplane Design, Part I–VII, second edition, DARcorporation, 2003.
craft in the weight category of the considered database but with [12] D.P. Raymer, Aircraft Design: A Conceptual Approach, fifth edition, AIAA Edu-
different mission requirements, showing that standard analyses cation Series, 2012.
also typical to ICE-propelled aircraft, albeit modified to encompass [13] Electravia, ZA Aérodrome, 04200 Vaumeilh, France. Website: www.electravia.fr.
the specific procedures presented in this paper, can be carried out [14] JSC Sportine Aviacija ir Ko, Pociunai LT-59327, Prienai, Lithuania. Website:
www.lak.lt.
bearing realistic results. The reasonable results obtained with such
[15] Lange Aviation GmbH, Brüsseler Straße 30, D-66482 Zweibrücken, Germany.
analyses, based on battery data pertaining to existing batteries and Website: www.lange-aviation.com.
specifically not to futuristic designs, tend to confirm the validity of [16] Pipistrel d.o.o., Ajdovšcina, Goriska Cesta 50a, SI–5270 Ajdovšcina, Slovenia.
the design procedure. Website: www.pipistrel.si.
A limit of the proposed procedure is that of considering a rela- [17] UAV Factory USA LLC, 50 South Buckhout Street, Irvington, NY 10533, USA.
Website: www.uavfactory.com.
tively small database of aircraft, making the exploration of aircraft
[18] Yuneec Americas (USA), 5555 Ontario Mills Parkway, Ontario, CA 91764, USA.
in other weight categories more difficult. Scaling problems con- Website: www.yuneec.com.
nected with the preliminary design of electrically powered aircraft [19] Alisport srl., via Confalonieri 22, 23894 Cremella (LC), Italy. Website: www.
will be the subject of a research work to follow. alisport.com.
[20] M. Hagen, S. Dörfler, P. Fanz, T. Berger, R. Speck, J. Tübke, H. Althues, M.J. Hoff-
Funding mann, C. Scherr, S. Kaskel, Development and costs calculation of lithium–sulfur
cells with high sulfur load and binder free electrodes, J. Power Sources 224
(2013) 260–268.
This research received no specific grant from any funding
[21] M. Hagen, D. Hanselmann, K. Ahlbrecht, R. Maça, D. Gerber, J. Tübke, Lithium–
agency in the public, commercial, or not-for-profit sectors. sulfur cells: the gap between the state-of-the-art and the requirements for high
energy battery cells, Adv. Energy Mater. 5 (2015).
Conflict of interest statement [22] G.J.J. Ruijgrok, Elements of Airplane Performance, Delft University Press, 2009.
[23] E.A. Estrada Rodas, J.H. Lewe, D. Mavris, Feasibility focused design of electric
The Authors declare that there is no conflict of interest. on-demand aircraft concepts, in: 14th AIAA Aviation Technology, Integration,
and Operations Conference, Atlanta, GA, June 16–20, 2014.
[24] S. Gudmundsson, General Aviation Aircraft Design, first edition, Butterworth–
References
Heinemann, 2013.
[25] R. Simpson, Airlife’s World Aircraft, Crowood Press, 2001.
[1] W. Cao, B.C. Mecrow, G.J. Atkinson, J.W. Bennett, D.J. Atkinson, Overview of
[26] L.R. Jenkinson, J.F. Marchman III, Aircraft Design Projects, first edition,
electric motor technologies used for More Electric Aircraft (MEA), IEEE Trans.
Butterworth–Heinemann, 2003.
Ind. Electron. 59 (2012) 3523–3531.
[2] B. Bilgin, A. Emadi, Electric motors in electrified transportation, IEEE Power [27] M.A. Kromer, J.B. Heywood, Electric powertrains: opportunities and challenges
Electron. Mag. 1 (2014) 10–17. in the U.S. light-duty vehicle fleet, Technical report LFEE2007-03RP, Sloan Au-
[3] R.H. Lyon, Machinery Noise and Diagnostics, first edition, Butterworths, 1987. tomotive Laboratory, Massachusetts Institute of Technology, Cambridge, MA,
[4] K. Ozawa, Lithium-ion Rechargeable Batteries, first edition, Wiley–VCH, 2009. 2007.
[5] G. Pistoia, Electric and Hybrid Vehicles, first edition, Elsevier, 2010. [28] C. Pornet, A.T. Isikveren, Conceptual design of hybrid–electric transport aircraft,
[6] C. Pornet, S. Kaiser, A.T. Isikveren, M. Hornung, Integrated fuel–battery hy- Prog. Aerosp. Sci. 79 (2015) 114–135.
brid for a narrow-body sized transport aircraft, Aircr. Eng. Aerosp. Technol. 86 [29] L.W. Traub, Range and endurance estimates for battery-powered aircraft,
(2014) 568–574. J. Aircr. 48 (2011) 703–707.

2016 - An Integrated Approach To The Preliminary Weight Sizing

Uploaded by

Copyright:

Available Formats

2016 - An Integrated Approach To The Preliminary Weight Sizing

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2016 - An Integrated Approach To The Preliminary Weight Sizing

Uploaded by

Copyright:

Available Formats

Aerospace Science and Technology 58 (2016) 134–149

Contents lists available at ScienceDirect

Aerospace Science and Technology

An integrated approach to the preliminary weight sizing

AR Aspect ratio T Time of ﬂight

# Model W to / g [kg] W e / g [kg] W bat / g [kg] W m / g [kg] Pm [kW] e [Wh/kg] p [W/kg]

Fig. 1. Example aircraft in the database in Table 1.

note that in this paper all weights are considered as a forces,

2.2.1. Mission analysis through ﬂight mechanics

2.2.2. Design constraints from sizing matrix plot

Table 2 altitude changes, whereas clearly no loss due to motor conversion

Conﬁguration Motor-glider Acrobatic

an impact (or a near impact) on the resulting space of possible so-

on cost and RC , this measure is more meaningful for this kind of

You might also like