Energy and Buildings 82 (2014) 733–745

Contents lists available at ScienceDirect

Energy and Buildings

journal homepage: www.elsevier.com/locate/enbuild

Balancing energy efficiency and structural performance through

multi-objective shape optimization: Case study of a rapidly
deployable origami-inspired shelter夽
C.P. Quaglia a , N. Yu b , A.P. Thrall a,∗ , S. Paolucci b
a
Department of Civil and Environmental Engineering and Earth Sciences, University of Notre Dame, 159 Fitzpatrick Hall, Notre Dame, IN 46556, USA
b
Department of Aerospace and Mechanical Engineering, University of Notre Dame, 117 Cushing Hall, Notre Dame, IN 46556, USA

a r t i c l e i n f o a b s t r a c t

Article history: For military and disaster relief housing, rapidly deployable shelters must be lightweight, be packaged
Received 16 June 2014 in a small volume for transportability, and be erected without heavy lifting equipment. A critical design
Received in revised form 24 July 2014 criterion is also energy efficiency in heating and cooling. To meet these priorities, the research team
Accepted 26 July 2014
has utilized origami as inspiration for a thermally insulated rigid wall deployable shelter that can be
Available online 7 August 2014
erected manually through counterweighting. To enhance energy efficiency, improvements in the shape of
a structure (i.e., member lengths and angles) at a design stage can lead to savings throughout its lifecycle.
Keywords:
This is magnified in the context of mass-production of deployable shelters, where any improvements are
Optimization
Thermal energy multiplied. Structural efficiency is also critical to achieve lightweight design. This paper presents a multi-
Deployable shelter objective shape optimization methodology which balances the priorities of structural performance (i.e.,
Origami minimum deflections) and energy efficiency (i.e., minimum thermal energy load). This is demonstrated
for the case study of a deployable shelter. Design variables include geometric parameters. Constraints
relate to the package size and capability of interfacing with existing technologies. Structural analysis is
performed using a parametric finite element program. Thermal energy is calculated using EnergyPlus.
An optimized solution is found and compared against existing military solutions.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction and motivation
Standard design priorities for rapidly deployable shelters in
military or disaster relief applications include adequate structural
performance, a low self-weight, a small package volume, and ease
of erection. A rising priority, particularly within a military context,
is reducing fuel consumption for heating and cooling these shelters.
The US military currently spends 66 million USD per day on fuel for
air conditioning (24 billion USD per year; data as of 2011) [1]. This is
due to both the cost of fuel itself and the cost of transportation and
security for fuel missions [2]. More specifically, while the standard
cost of a gallon of generic diesel fuel in 2014 is 3.25 USD [3], the
fully burdened cost of fuel, which includes security and transporta-
tion in a military context, can be up to 600 USD per gallon [2,4]. As
of 2011, one thousand Americans have died during fuel missions in
Iraq and Afghanistan [1]. This data demonstrates the urgent need
夽 This document is a collaborative effort.
∗ Corresponding author. Tel.: +1 5746312533.
E-mail addresses: cquagli1@nd.edu (C.P. Quaglia), nyu@nd.edu (N. Yu),
athrall@nd.edu (A.P. Thrall), paolucci@nd.edu (S. Paolucci).
for sheltering systems that can meet structural and deployability
requirements while reducing fuel consumption.
Existing military sheltering systems (see Fig. 1) include soft wall
(canvas) and rigid wall solutions (see [5] and the review in [6]).
Soft wall shelters feature excellent deployability (including low
self-weight and small package volume), but provide little thermal
insulation. Energy for heating and cooling can be reduced by 80%
in such shelters through the application of spray foam insulation
[1]. However, this prevents such shelters from being re-deployed.
Alternatively, rigid wall shelters provide higher thermal resistance,
but have large self-weights and package volumes. Toward achiev-
ing the standard design priorities and integrating energy efficiency,
Quaglia et al. [6] have proposed an origami-inspired deployable
shelter (see Fig. 2) known as the Lever Shelter Module (LSM). Here,
the art of origami is employed as inspiration for folding a rigid
wall system. It can be packaged small (Fig. 2(A)) to be transported
on a standardized 463L pallet [7] making it transportable by air,
truck, rail, or ship. The folding rigid walls are comprised of sand-
wich panels which have a high strength to weight ratio (resulting
in a low-weight shelter) and offer high thermal insulation [8]. A
novel erection strategy utilizing a lever arm which can rotate the
shelter into a self-supporting position (Fig. 2(A)–(D)) enables the

http://dx.doi.org/10.1016/j.enbuild.2014.07.063
0378-7788/© 2014 Elsevier B.V. All rights reserved.
734 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745
Fig. 1. Existing military shelters, including (A) soft wall TEMPER tent, (B) soft wall
TEMPER Air tent and (C) a rigid wall shelter.
Images courtesy of [5].
system to be erected without the use of heavy lifting equipment.
The shelter is highly modular (Fig. 2(E) and (F)) for versatility of
use and can be interfaced with existing military technologies by
Force Provider [9], including kitchens and latrines, housed in Tricon
shipping containers [10]. This concept provides the deployability
needed to meet transportation goals while also offering thermal
insulation [6]. While the thermal insulation provided through sand-
wich panels increases the energy efficiency of the shelter system,
the design of the shelter itself can also be developed toward this
goal. This paper presents a multi-objective shape optimization of
the LSM for structural performance (i.e., minimum displacement
under service loads) and energy efficiency (i.e., minimum thermal
energy load to maintain the shelter within a temperature range for
one year).
Shape optimization of structures for energy efficiency (i.e., opti-
mization of the global structural form toward reducing the fuel
needed for heating and cooling) is a powerful design tool for
the sustainable building community since any reduction in fuel
consumption made through design is multiplied over the life-
cyle of this structure. This is especially significant for the case
of deployable structures for military and disaster relief applica-
tions where such designs are also mass-produced. A wide body
of research is developing in the study of the effects of different
structural characteristics on the energy efficiency of a structure.
The most relevant examples are briefly reviewed here. Multi-
objective optimization of the shape of a building for minimum
building cost and for minimum yearly heating cost has been inves-
tigated by Adamski and Marks [11] and Jedrzejuk and Marks [12]
for octagonal plans, by Adamski [13] for arbitrary plans defined
by two curves, by Marks [14] for arbitrary and polygonal plans,
and by Adamski [15] for oval plans. Bouchlaghem [16] demon-
strates a procedure for optimizing shading devices for minimum
cooling load of a building. Caldas and Norford [17] optimized win-
dow size for lighting, heating, and cooling. In later work, Caldas
and Norford [18] also optimize (1) building materials for mini-
mum energy consumption and initial building cost, (2) building
form for lighting and heating energy, and (3) Heating, Ventilation,
and Air Conditioning (HVAC) system size, control strategy, and
building envelope for minimum energy cost. Coley and Schukat
[19] have proposed a procedure to optimize the shape (building
perimeter and roof pitch) and construction details (e.g., materials,
window location, orientation) to identify a broad range of low-
energy designs that can then be assessed for their architectural
appeal. Grierson and Khajehpour [20] presented a methodology
for optimization of commercial buildings for minimum capital and
operating cost (including cost of energy, maintenance, and taxes)
and maximum income. Design variables include the structural sys-
tem, the floor system, cladding type, window type, window ratio,
number of bays, and bay width. Here, structural performance is
included as each conceptual design is analyzed for load-path redun-
dancy to prevent against progressive collapse. Michalek et al. [21]
propose a method to optimize the floor layout of a building for
energy, lighting, wasted space, access ways, and hallway objective
functions. Malkawi et al. [22] optimized building shape (i.e., length,
width, and height of an orthogonal geometry, window location and
size, position and size of supply and extract terminals) to maximize
thermal and ventilation priorities. Wang et al. [23] optimized the
building orientation, aspect ratio, window type, window to wall
ratio, and material composition of the roof and walls for minimum
life cycle cost and minimum life cycle environmental impact. This
work is extended for the same objective functions in Wang et al.
[24] with design variables including the plan shape of the build-
ing, the structural system, building envelope configurations, and
overhangs. Shea et al. [25] optimize building panel envelopes for
lighting and for cost. Diakaki et al. [26] proposed a multi-objective
approach to optimizing the window type, wall insulation mate-
rial, and wall insulation thickness for acquisition cost and energy
savings. Yi and Malkawi [27] optimize complex geometries (imple-
mented by agent nodes) for energy performance. Fialho et al. [28]
use a multi-objective evolutionary algorithm to investigate the
effects that changing materials and orientations of the building
have on the construction costs and energy efficiency of the build-
ing. Krem et al. [29] presented a study of the effect of the position
of the structural core/wall and shape of the floor plan on energy
and structural performance of high rise buildings in various climate
zones. Vergauen et al. [30] studied the effects of changing geom-
etry and kinetic parameters of origami-based facade components
on direct solar radiation, daylight, and glare. Martinez-Martin and
Thrall [8] presented a simplified optimization methodology for the
selection of the material properties of sandwich panels (i.e., mate-
rial type and thickness for the face and core) for origami-inspired
deployable shelters based on the competing objective functions
of weight and thermal insulation. Flager et al. [31] employ Pro-
cess Integration Design Optimization (a suite of software products
typically used in the aerospace industry) to optimize the orienta-
tion, length, width, and member section properties of a steel frame
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745
Fig. 2. LSM concept, including (A)–(D) the erection sequence, (E) back-to-back modules with independent geometric variables highlighted, and (F) six modules longitudinally
configured and interfaced with a Tricon container.
building for minimum capital and life-cycle cost. Structural safety is
considered as a constraint in which all members must meet design
code.
Several software packages have been developed aimed at facil-
itating the optimization of buildings for sustainability and energy
efficiency, including BEOpt [32], Opt-E-Plus [33], GENE ARCH [34],
and GenOpt [35]. Although there have been software developments
aimed toward facilitating building energy optimization for archi-
tects in the early design stage, there is limited guidance available
[36]. Primary challenges identified by Bambardeka and Poerschke
[36] include how to integrate the design process with a simula-
tion tool and how to solve the data interoperability between a
Building Information Model (BIM) and an energy model. A simu-
lation tool, CODYRUN, which regroups design and research aspects
and adapts to different types of climate was presented in [37]. It
provides a powerful method of investigation in the thermal analy-
sis of a system, but CODYRUN was designed for professionals in
energy modeling instead of building designers. The MIT Design
Advisor, presented by Urban [38], is a simple and rapid building
simulation tool with a user-friendly interface for non-technical
specialists. However, it requires the users to input the concep-
tual building design details, which demands abundant prior design
knowledge and experience. In addition, merely inputting the con-
ceptual design descriptions does not accurately specify a building
structure. Nielsen [39,40] utilizes a simulation tool to evaluate the
energy demand and thermal indoor environment in the early build-
ing design stage, but the geometric modeler is SketchUp [41], which
is mainly used for draft sketching purposes instead of professional
and detailed building design, and additional efforts are needed to
export the geometric information from SketchUp and input it into
the tool. The Sustain tool developed by Greenberg et al. [42] offers
an interactive graphical interface with EnergyPlus serving as its
primary engine for building energy simulation. However, this tool
currently requires multiple workstations with large computational
power and also uses SketchUp as the primary geometry generator.
In relation to this prior work in the field, the novelty of the
research presented in this paper is (1) an optimization methodol-
ogy for the shape of a building for combined structural performance
and energy efficiency, and (2) the specific application to deployable
735
shelter technology. This research represents the first investigation
of the optimization of the shape of a deployable shelter for energy
efficiency and for combined energy efficiency and structural per-
formance. It will be demonstrated for the case study of the rapidly
deployable shelter discussed above and shown in Fig. 2. First we
present the optimization methodology, including the problem def-
inition, algorithm, and the structural and energy analysis software
with which it is interfaced. Results of single and multi-objective
optimizations are presented. The energy efficiency of the optimized
solution is then compared with that of existing military solutions.
The paper concludes with a discussion of the optimized shelter and
a comparison with these existing military solutions.
2. Optimization methodology
The geometry of a structure (i.e., lengths and angles of mem-
bers) can be optimized for structural performance (measured by
magnitude of deflections under service loads) and energy efficiency
(i.e., thermal energy load provided to maintain a structure within
a comfortable temperature range during a one year period) sub-
ject to desired constraints. This will be demonstrated for the case
study presented in Fig. 2. This system is defined by six indepen-
dent variables (Bh , Wx , Rh , Rw , , and ; see Fig. 2 and Table 1).
These variables are constrained by requiring that the system can
be packaged on a 463L pallet and can be interfaced with a Tricon
container. Since optimization for minimum deflections and mini-
mum thermal energy load would naturally lead to decreasing the
volume, Rw has been selected to be the largest size (2.13 m) while
Table 1
Independent geometric variables for case study.
Variable Description
Bh Back wall height
Wx Wing wall width
Rh Roof height
Rw Roof width
 Back wall angle
 Roof angle
736 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745

still being able to be packed on a 463L pallet and remains con-

stant throughout optimization. The remaining five variables are the
design variables for the optimization procedure. Note that windows
have not been included since the case study demonstrated here
is for a military sheltering application where light pollution from
windows can pose safety problems. For traditional buildings, how-
ever, windows would have a large effect on both thermal energy
load and structural performance and should be considered in the
preliminary design stage. The methodology proposed here could
include windows with the addition of a few design variables.
The following paragraphs will provide the formal problem
definition, design variables, constraints, algorithm, and software
packages that evaluate structural and energy performance.
Fig. 3. Flowchart indicating the parametric model interface with analysis software.
2.1. Problem definition
been discretized in even increments toward the aim of manufac-
Let S ≡ {Bh , Wx , Rh , , }, then the formal optimization problem turability. For each of the length variables, there are 265 discrete
can be defined as follows: options and for each of the angle dimensions, there are 76 options.
⎧
⎪ ı
⎨ self −weight
(S) Overall, this accounts for just over 107 billion different possible
combinations. It takes approximately 30 s to run one combination
min ı(S) = max ıwind (S) (1)
S ⎪
⎩
for structural analysis and about 5 s to run one combination for
ısnow (S) energy analysis. Therefore, 4.32 × 107 days would be necessary to
 exhaustively evaluate every combination for structural and energy
min Fy (S) = fheating (S) + fcooling (S) (2) performance. This justifies the implementation of an optimization
S algorithm for this case study. For example, the longest numerical
such that the following design constraints cj are satisfied: test for structural performance optimization took about 8 h and the
longest numerical test for energy performance took about 2 h which
cj (S) ≤ 0 j = 1, . . ., M. (3) shows the large time saving benefit of implementing an optimiza-
Above, ı is the highest deflection in any direction of the two tion algorithm.
back-to-back modules (Fig. 2(E)) under self-weight, wind (29.1 m/s,
2.3. Constraints
applied as per American Society of Civil Engineers (ASCE) Standard
7-10 [43]), and snow (0.479 kN/m2 ) as prescribed by the US
The constraints of this problem have been formulated to ensure
Army Natick Soldier Research, Development & Engineering Center
that the structure can be packaged on a 463L pallet, that it can
Commanders Smartbook Equipment Catalog [44]. Wind loads are
interface with existing technologies through Tricon containers, and
applied based on the worst case pressure loads for the two-module
also that the wind coefficients remain constant as the form of the
enclosed shelter based on ASCE Standard 7-10 [43]. While the struc-
structure is varied. These equations are also given in [50].
ture would be enclosed for use, only the load bearing panels of the
The shelter must be packaged on a 463L pallet (measuring
modules (i.e., back walls, wing walls, and roofs) are modeled for
2.24 m by 2.74 m [7]) to be transportable by air, truck, rail, or
structural analysis. The total thermal energy load (Fy ) for enclosed
ship. Therefore, the length must be less than that of the pallet
back-to-back modules is calculated as the sum of the loads for heat-
(Py = 2.74 m) with a handling gap (g = 0.0508 m) on each side, or
ing (fheating ) and the loads for cooling (fcooling ) over 12 months to
keep the shelter between 20◦ C and 24◦ C in Chicago, USA. To evalu- Wx − Bh cos 
c1 = Bh + cos( − ) − Py + 2g ≤ 0. (4)
ate these objective functions, a parametric model of the structure cos 
was built in Grasshopper [45] (a plug-in for the computer-aided
Likewise, limits related to the width of the pallet (Px = 2.24 m) with
design package Rhinoceros [46]). This model is interfaced with a
a handling gap must be met, or
finite element (FE) analysis program Karamba [47] which evalu-
ates the deflection of the structure under each service load. It is also c2 = Rw − Px + 2g. (5)
interfaced with EnergyPlus [48] through Gerilla [49] (also embed-
To prevent the wing walls from overlapping when packaged, we
ded in Grasshopper) which calculates the thermal energy load.
take
Fig. 3 provides a flowchart of the analysis procedure which is also
discussed in more detail below. The constraints ci (see Eqs. (4)–(11)) Rw
c3 = Wx sin  + t + e − ≤ 0, (6)
relate to design priorities, including that the system can be pack- 2
aged on a 463L pallet for transportation, that in the erected form where t is the panel thickness (t = 0.0685 m) and e is additional dis-
the structure can interface with existing technologies which are tance included for tolerance (e = 0.0254 m). The wing wall is inset
packaged in Tricon containers, and that wind coefficients remain from the back wall by the amount of its own thickness. To ensure
constant as the shape is varied. that the roof can fold onto the back panel, the roof is constrained
as follows:
2.2. Design variables
c4 = Rh + e − Bh ≤ 0, (7)
Five design variables are considered (see Fig. 2 and Table 1). Note where e is added for tolerance. These equations constrain the geom-
that Rw has been selected to be the largest value possible to ensure etry for packaging on a standardized 463L pallet.
sufficient volume is maintained and is therefore not a design vari- Existing technologies (e.g., kitchens, latrines) such as those pro-
able. The range of length dimensions (i.e., Bh , Rh , and Wx ) is 0 m vided by Force Provider are housed in Tricon containers (2.44 m by
to 2.64 m (increments of 0.01 m) and the range for angles (i.e.,  2.44 m by 1.96 m [10]). For the erected structure to interface with
and ) is 15◦ to 90◦ (increments of 1◦ ). These design variables have these containers (see Fig. 2(F)) the horizontal distance between
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745
opposing wing walls must exceed the width of the Tricon container
(Cx = 1.96 m) with an additional distance for tolerance, or
Cx
c5 = Wx + + e − Bh cos  − Rh cos  ≤ 0. (8)
2
The roof above the Tricon must be higher than the height of the
container (Cy = 2.44 m) with tolerance, or
C  of SA was used first to minimize deflections and thermal energy
x
c6 = Cy + e + tan  + e − Bh sin  − Rh sin  ≤ 0. (9)
2
The wing walls are necessary to support the roof and therefore it is
required that they intersect:
c7 = Bh cos  − Wx < 0. (10)
Note that this constraint is a strict inequality. This is because the
wing wall width must be greater than 0 m if the angle of the back
wall is 90◦ . These three constraints ensure that the shelter can
interface with a Tricon container housing existing technologies.
Lastly, the uniform velocity exposure coefficient utilized in the
calculation of wind loads increases for heights above 4.57 m. As a
result, the shelter is limited to be less than this height (H = 4.57 m),
or
c8 = Bh sin  + Rh sin  − H ≤ 0. (11)
2.4. Optimization algorithm
Since this optimization problem includes nonlinear objective
functions, nonlinear constraints, and discrete design variables, the
heuristic algorithm of Simulated Annealing (SA) has been selected.
This iterative improvement algorithm searches the design space
[51] and has been successfully implemented in structural engi-
neering applications (e.g., [52–54]) and specifically for deployable
structures (e.g., [8,55–57]).
The SA algorithm is based on an analogy to the natural process
of solidification when a high temperature mass is cooled slowly.
2
During this controlled cooling process, lower energy configurations
typically result. However, there is a probability that higher energy
configurations that are functions of energy and temperature occur
[58]. Kirkpatrick et al. [51] first related this natural process to opti-
mization, where the objective function is analogous to the energy
and this probability of higher value solutions enables the algorithm
to escape local minima.
The SA algorithm, as implemented in this research, begins by
finding an initial solution by randomly assigning each of the design
variables a value from user defined databases. The algorithm finds
a second solution by randomly perturbing a number of the vari-
ables of this initial solution. If the new solution has a lower value
for the objective function than the initial solution, the new solution
is taken as the current working solution upon which the algorithm
continues to iterate. If not, the algorithm calculates the probability
of accepting this higher value solution as the current solution based
on P = e−Ef /T , where Ef is the value of the objective function and T
is the temperature (a parameter controlled by the algorithm). This
enables the algorithm to escape local minima. The temperature is
initially set to a high value so that the algorithm explores many
higher value solutions initially. In this research, an initial tempera-
ture which accepts between 20% and 40% of higher value solutions
is found [59]. To minimize the computational effort in determin-
ing an initial temperature, three numerical tests were performed
to find a temperature that accepts solutions within this range and
the average of the results was used. The algorithm iterates using
this initial temperature for a certain number of iterations defined
as the length of a cooling cycle. The temperature is then reduced by
a user defined factor. The algorithm reaches convergence when a
certain number of cooling cycles have been obtained in which there
737
is no improvement on the solution [51]. The user must define sev-
eral parameters related to this algorithm, including the maximum
number of variables to be varied in a solution (v), the maximum
amount of perturbation within variable databases (pm), the tem-
perature reduction factor (r), the number of iterations in a cooling
cycle (m), and the number of cooling cycles in which there has been
no improvement for convergence (n). This single objective version
loads separately.
To optimize deflections and thermal energy loads simulta-
neously, a multi-objective version of SA (MOSA) was employed
based on the approach proposed by Suppapitnarm et al. [60] and
utilized for the optimization of deployable structures previously
[8,55,57]. In the version of the MOSA algorithm implemented in
this research, the algorithm searches the feasible solution space as
in the single objective SA case. However, in MOSA, a set of Pareto-
optimal solutions (i.e., solutions which are not overshadowed by
another solution) is formed. As the algorithm evaluates new solu-
tions, each solution is compared against this Pareto-optimal set. If
the new solution is Pareto-optimal, it is accepted and the Pareto-
optimal set is appropriately updated. If not, the algorithm calculates
the probability of continuing to iterate on the solution (P) by
Q
P= e−[Efi (s1 )−Efi (s2 )]/Ti , (12)
i=1
where Q is the number of objective functions, Efi is the value of each
objective function, s1 is the current working solution, s2 is the new
solution under consideration, and T is the temperature related to
each objective function (determined as discussed for single objec-
tive SA). At the end of a cooling cycle, an intelligent return-to-base
strategy is implemented to select a solution to be iterated on. This
enables the algorithm to select progressively more isolated solu-
tions to explore the extreme ends of the Pareto-optimal set. The
degree of isolation for each point is calculated by
Q

As
 Ef 1 (sk ) − Efi (sj )
I(sj ) = , (13)
Efimax − Efimin
k=1 i=1
k=
/ =j
where As is the total number of solutions in the Pareto-optimal
set and Efmax and Efmin are the maximum and minimum values of
each objective function in the Pareto-optimal set. The solutions are
ordered by decreasing amount of isolation. The algorithm starts
by selecting any number of As solutions, but as the algorithm pro-
gresses, it selects a smaller and smaller percentage (decreasing this
percentage by a factor of 0.9 after each chain; this factor is typi-
cally used in the literature) of the most isolated points in As . This
enables the algorithm to select the most isolated solutions as those
upon which it iterates. The algorithm converges when no additional
Pareto-optimal solutions are found after a certain number of spec-
ified cooling cycles [60]. The final result is a Pareto-optimal set of
solutions from which a designer can select an optimized solution.
2.5. Analysis: finite element modeling
To evaluate the structural performance of each solution (i.e.,
determine the highest deflection of any element in the structure
under service loads), the optimization algorithm is interfaced with
the finite element package Karamba [47] through Grasshopper [45],
a plug-in for the computer aided three-dimensional drafting appli-
cation Rhinoceros [46] (see Fig. 3). More specifically, a parametric
model was built in Grasshopper which, when given the values of
the design variables, generates the shelter geometry as surfaces.
This geometry is input into Karamba along with loads, boundary
738 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745

conditions and material properties. Karamba is a pre-design para- 40

metric finite element tool which can determine the linear elastic
deflection response of a structure. It has been previously imple- 30
mented for the evaluation of deploying structures (e.g., [61–63]).
The case study presented here is modeled in Karamba as shell
20
elements. Only the load bearing panels of the structure are modeled

Temperature [°C]
(i.e., back wall, wing, wall, and roof) since any enclosure panels
would not be expected to perform structurally. To ensure conver- 10
gence of results, a mesh refinement study was performed leading
to a mesh size of approximately 60 mm (see [64]). The shelter is 0
restrained by pinned (free rotation, no translation) boundary con-
ditions along the base of each load bearing panel. Connections −10 High set−point
between panels are hinges which are approximated by reducing Low set−point
the thickness of elements (to approximately 15% as thick as the Summer day
−20 Winter day
element) until pin-like behavior is achieved (see [64]). Material
properties for this model are discussed further below.
−30
0 2 4 6 8 10 12 14 16 18 20 22 24
2.6. Analysis: thermal modeling Hours

To determine the energy efficiency of a solution (i.e., thermal
energy load to maintain the shelter temperature between 20 ◦ C
and 24 ◦ C in Chicago, USA for 1 year), the optimization algorithm
is interfaced with the freely available EnergyPlus [48] software
package which is interfaced with Grasshopper through Gerilla [49].
EnergyPlus is an energy analysis and thermal load simulation pro-
gram for buildings that is based on a lumped capacitance model.
By utilizing the user’s inputs of the building’s detailed descrip-
tion including location, geometry, thermophysical properties of the
construction materials, and type of HVAC system, EnergyPlus cal-
culates heating and cooling loads required to maintain the desired
comfort temperatures given the exterior weather conditions. Note
that the outputs calculated by EnergyPlus are ideal heating/cooling
loads that are needed to maintain the desired temperature range
(without considering the efficiency of the HVAC system). The same
parametric model that was built for the evaluation of structural per-
formance can therefore also be used for thermal energy load (see
Fig. 3). The Grasshopper/Rhinoceros interface has been previously
used by Vergauen et al. [30] to manually evaluate the geometry and
kinematic parameters of origami-based facade shade elements.
For this case study, the fully enclosed shelter is modeled. Mate-
rial properties are assigned to each of the surfaces of the shelter,
including the roughness, thickness, conductivity, density, and spe-
cific heat. Roughness was taken as medium roughness for all
materials; other material properties are discussed further in the
next subsection. The simulation location used in this optimization is
Chicago, USA, because it exhibits a large variation of temperature in
a year. The temperature information is obtained from Typical Mete-
orological Year 3 (TMY3) [65], which is commonly used to represent
local climatic conditions. This file provides all climate information
that is important for energy analysis, including temperature, solar
Fig. 4. Temperature during the day in winter and summer, and set-points inside the
structure in Chicago, USA.
Data courtesy of [65].
computed by the Stefan-Boltzmann law [66]. Two types of radiation
are included: (1) direct sunlight which is a combination of bright
light and radiant heat un-obstructed by clouds, and (2) diffused
light which is radiation blocked by clouds or reflected by objects
and particles in the air [67]. Diffused radiation is predetermined
by EnergyPlus, but the direct radiation impinging on a surface is
dependent upon the changing geometry of the shelter and the shel-
ter’s orientation. For this research, the shelter is oriented so that
the wing walls are facing the north and south directions and cal-
culations by EnergyPlus are based on a time-dependent solar angle
[68].
On a hot day, solar radiation hitting the exterior wall surfaces of
the tent can increase the surface temperature. Heat is transferred to
the walls of the shelter through convection due to the temperature
difference between the outside air and the external wall surface.
External wind can usually cause the convection to be stronger.
Also, heat from an external surface of the wall is transferred by
conduction to the internal surface due to the temperature differ-
ence between the external and internal surfaces. Heat is transferred
40
30
20
Temperature [°C]

radiation, wind speed, wind direction, relative humidity and so

on. Temperature setpoints are 20 ◦ C and 24 ◦ C. Figs. 4 and 5 show 10
the temperature distribution on the summer day, winter day and
also over a complete year in comparison with the desired indoor 0
temperature set-points: within 20 ◦ C and 24 ◦ C. When the indoor
temperature is below the low set-point 20 ◦ C, heating is needed, −10
High set−point
while when it is above the high set-point 24 ◦ C, cooling is needed. Low set−point
Additional default settings are in Table 2 for the reader’s reference. Outside temperature
The energy loads in EnergyPlus are based on the three modes −20
of heat transfer that take place in the shelter: conduction, con-
vection, and radiation [66]. Conduction, which is characterized by −30
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
thermal conductivity, occurs through the building envelopes by
Months
direct contact. Convection, which is characterized by a heat con-
vection coefficient, takes place between a wall surface and the Fig. 5. Temperature distribution during a complete year and set-points inside the
ambient surrounding air. Radiation is caused by electromagnetic structure in Chicago, USA.
waves between two surfaces that have different temperatures, and Data courtesy of [65].
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745 739

Table 2 behavior of the entire sandwich panel. More specifically, the flex-
Settings in EnergyPlus for energy simulation.
ural stiffness of the panel (D) is given by
Objectives Description 
Ec tc3 b tf3 tf (tf + tc )2
Terrain Country D= + Ef b + , (14)
Shadow calculation frequency 20 12 6 2
Surface convection algorithm: inside TARP
Surface convection algorithm: outside TARP where E is the Young modulus for the core (c) and face (f), t is the
Heat balance algorithm Conduction transfer function
thickness, and b is the width [71]. Since Ec Ef , this stiffness can be
Time step 4
Run period 01/01–12/31 approximated as [71]
Site: ground temperature 18.3, 18.2, 18.3, 18.4, 20.1, 22, 
(January–December) 22.3, 22.5, 22.5, 20.7, 18.9, 18.5 tf3 tf (tf + tc )2
Schedule type limits Any number D ≈ Ef b + . (15)
6 2
Heating setpoint temperature Always 20 (◦ C)
Cooling setpoint temperature Always 24 (◦ C)
Zone HVAC Ideal loads air system If the panel is approximated as a beam, then this stiffness can be
Output: variable Zone air system sensible further approximated as D ≈ Ef I, where I is the moment of inertia
cooling energy, monthly; Zone 3
(I = bt es /12, for a rectangular beam of width b and thickness tes ).
air system sensible heating
energy, monthly Using these relationships, the equivalent thickness for the struc-
tural model is then calculated by
1/3
tes = [2tf3 + 6tf (tf + tc )2 ] . (16)
to the indoor air from the interior surfaces by natural convection
An equivalent density (es ) is also used for the structural model:
if there is no forced convection in the room. Also, the HVAC and
lighting system act as internal heating sources. On a cool day the te
es = e , (17)
opposite occurs where heat is transferred from the indoor to the tes
outdoor. In order to maintain indoor air within a comfortable tem- where te is the total sandwich panel thickness (te = 2tf +tc ) and e is
perature range, the HVAC system provides the required heating the effective density of the sandwich panel which is calculated by
or cooling loads to the room based on the direction of heat trans-
2f tf + c tc
fer at any given moment throughout the year. The thermal energy e = , (18)
from the HVAC system that is required to maintain the desired 2tf + tc
thermal conditions in the shelter in a year is what is quantified where  is the density.
by EnergyPlus and is the focus of the energy optimization [48]. For the thermal model, the entire panel thickness is used and the
material properties are approximated using core and face thermal
properties. More specifically, the effective density (e ), as defined
2.7. Material properties above is used. The effective specific heat (Cpe ) is given by

The shelter in this case study is comprised of sandwich panels, 2Cpf f tf + Cpc c tc
Cpe = , (19)
made up of two rigid faces and a lightweight core, which provides a e (2tf + tc )
high strength to weight ratio and good thermal insulation. The type
where C is the specific heat. The effective thermal conductivity (ke )
of material and thickness of each component can be optimized for
is given by
similar design priorities (i.e., minimum self-weight and maximum
thermal resistance) [8]. However, since the focus of this research is 2tf + tc
ke = , (20)
on the global geometry of the system, the material properties will 2(tf /kf ) + (tc /kc )
be assumed to remain constant. More specifically, the faces are fiber
reinforced polymer (FRP) comprised of two sheets of E-LT 2400-P where k is the thermal conductivity.
and two sheets of E-BX 2400 manufactured by Vectorply [69]. The Individual and effective values for these properties and their
core is Corecell M Foam by Gurit (M80) [70]; see Table 3 for the sources are listed in Table 3. The face conductivity was calculated by
relevant properties. These materials were selected based on rec- adding 30% of the value of the specific heat for resin plus 70% of the
ommendations from Lyman-Morse Boatbuilding Co. (Thomaston, value for the specific heat of the fibers (based on the fiber content
ME) and thicknesses of each component were designed to meet percentage). A value of 0.22 W/m K was used for the resin based on
the current design code under prescribed loads (see [6]). personal communication with Rick Pauer of CCP Composites US (on
The software programs utilized in this optimization procedure 1/20/2014). A value of 1.3 W/m K was provided by Vectorply for the
are not capable of modeling multi-layer material such as these fibers [72].
sandwich panels. As a result, equivalent structural and thermal
properties as discussed below and as catalogued in Table 3 are used. 3. Results
For the structural model, the panel is approximated as an FRP
material with an equivalent thickness (tes ) which accounts for the 3.1. Parameter study

Before implementing single and multi-objective optimizations,

the effects of each variable on structural and thermal performance
Table 3 was investigated (see Fig. 6). For this parameter study, all variables
Material properties for case study.
were held constant while changing just one variable within the
Material t (mm)  (kg/m3 ) E (MPa) C (J/kg K) k (W/m K) limits of the constraints. The variables are highly constrained (see
Face 2.49 1880.0 [64] 17500 [64] 1000 [78] 0.976 Section 2.3), but this study provides an indication of the influence
Core 63.5 85 [70] 52.5 [70] 1500 [79] 0.04 [70] of the variables on the objective functions.
Structural model 40.2 3.60 NA NA NA Variables Bh and  are too highly constrained and show little
Thermal model 68.5 216 NA 532 0.043
effect on the objective functions. The effect of the wing wall width
740 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745

Fig. 6. Results of parameter study for deflection (left) and for thermal load (right).
Deflection data reprinted from Quaglia et al. [50] with permission.

(Wx ), the roof height (Rh ), and the roof angle () are shown in Fig. 6.
Fig. 6 shows that an increase in wing wall width (Wx ) results in
lower deflections. This is expected as a wider wing wall reduces the
unsupported length of the roof, thereby improving performance.
The slight upward trend when Wx exceeds 0.9 m can be attributed
to the greater longitudinal loads applied to the wing walls as it
is widened. This is a result of simplifications made in modeling
and should be ignored. This variable had no effect on thermal per-
formance, however, since the structure is fully enclosed for the
thermal model. The roof height (Rh ) had the greatest overall effect
on both structural and thermal performance, with both objective
functions favoring a smaller value. From a structural perspec-
tive, decreasing the roof height decreases the unsupported length
between modules and also decreases the total wind load. From a
thermal perspective, the volume and surface area are both smaller,
leading to energy savings. Structural performance favors a lower
roof angle (), while thermal performance favors a higher value.
Deflections are reduced with the smaller angle since the projected
area over which the wind acts is reduced. Conversely, the larger
roof angle, reduces the volume of shelter to be heated or cooled.
This parameter study shows that the objective functions are
competing for some design variables. However, since only one
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745 741

Table 4
Single-objective optimization numerical tests. Data for minimum deflection reprinted from conference proceedings [77] with permission.

Parameters Minimum deflections Minimum thermal energy load

Name v pm r m n  (mm)  (mm) cv  (J/yr)  (J/yr) cv

SA 1 1 10 0.9 50 2 1.24 0.483 39.1 4.61 × 109 3.31 × 108 7.18

SA 2 1 10 0.9 100 2 1.10 0.168 15.3 4.67 × 109 2.19 × 108 4.70
SA 3 1 10 0.9 250 2 1.07 0.137 12.8 4.52 × 109 2.81 × 108 6.21
SA 4 1 15 0.9 50 2 1.15 0.247 21.5 4.69 × 109 3.17 × 108 6.76
SA 5 1 15 0.9 100 2 1.06 0.0358 3.37 4.64 × 109 2.24 × 108 4.82
SA 6 1 15 0.9 250 2 1.03 0.0373 3.63 4.57 × 109 2.58 × 108 5.65

parameter is varied at a time and these parameters are very
constrained, further study is required through single and multi-
objective optimizations.
3.2. Single objective results
Single objective optimization for both minimum deflection and
minimum thermal energy load was performed to (1) gain an under-
standing of the best extremes for each objective function, and
(2) evaluate the algorithmic parameters employed in SA (v, pm,
r, m, and n) to determine a robust combination for MOSA. Six
combinations of algorithmic parameters were studied. For each
combination, ten numerical simulations were performed (a total
of 60 simulations for each objective function). Table 4 includes the
statistical data for these numerical tests, including the average (),
standard deviation (), and coefficient of variation (cv , dimension-
less). Note that the temperature reduction factor (r) is pre-selected
to be 0.9 based on values typically chosen in the literature ([59,73]).
The most robust combination of algorithmic parameters (where
robustness is defined as having both a low average and a low
standard deviation) for both objective functions of structural and

Fig. 7. Representative convergence plot for (A) minimum deflection optimization

and (B) minimum thermal energy load optimization. Renderings of the initial and Fig. 8. Renderings of optimized designs for (A) single objective minimum deflection,
final solutions are shown. (B) single objective minimum thermal energy load, and (C) MOSA.
(A) Reprinted from conference proceedings [77] with permission. (A) Reprinted from conference proceedings [77] with permission.
742 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745

Table 5 energy load. Adversely, deflections increase significantly with the

Comparison between optimized designs.
decrease of wing wall width since the unsupported length of the
Min ı Min F MOSA roof increases. Although the wing wall width is not the only geo-
Back wall height (Bh , m) 2.13 2.45 2.3 metric parameter changing, it plays the most significant role in the
Wing wall width (Wx , m) 0.69 0.01 0.32 range of solutions because of the competing effects of the objective
Roof height (Rh , m) 1.85 1.12 1.49 function. Note that while the wing wall did not effect the thermal
Back wall angle () 84 90 90 performance in the parameter study, it does impact this objec-
Roof angle () 30 22 27
tive function when all variables are varied. A smaller roof height
Highest deflection (mm) 0.973 169.4 1.55
Thermal energy load (GJ/yr) 5.09 3.99 4.54 (Rh ) results in lower deflections due to a decrease in unsupported
length which is also seen in the parameter study. This also influ-
ences a decrease in the thermal energy load by limiting the total
energy performance is SA 6 (v = 1, pm = 15, r = 0.9, m = 250, and n = 2; surface area of the shelter in contact with the ambient and helping
see Table 4). to decrease the overall volume of the shelter (less thermal energy to
Fig. 7 shows convergence plots of representative single objec- cool/heat a smaller volume). Varying the back wall and roof angles
tive optimization numerical tests for deflection and thermal energy ( and ) has a much more significant effect on thermal energy load
load. Lines indicate the current solution upon which the algorithm than on deflections. As the back wall angle increases towards ver-
iterates and the best solution found as a function of iteration num- tical, it decreases the amount of direct sunlight on the back walls
ber. Renderings of the first solution and the final best solution are of the shelter during the day. Lastly, a larger back wall angle, along
shown. These plots demonstrate the efficacy of the algorithm to with a smaller roof angle, creates a smaller boxier volume which
search the solution space. They also demonstrate the general trend takes less energy to cool or heat.
in shape for each objective function. Optimization for structural The final optimized design is selected from the Pareto-optimal
performance moves toward wider wing walls and inclined back set at the knee of the curve which is considered a good compromise
walls in order to improve behavior. Optimization for thermal per- between two competing objective functions (see Figs. 8 and 9, and
formance moves toward a boxier shape which minimizes volume Table 5). The filled diamond indicates the final optimized design.
and surface area while just accommodating the Tricon container. Table 5 shows a comparison of the multi-objective and single objec-
This trend is further shown in the best solutions found over all tive results. The multi-objective result shows a 12% improvement
numerical tests for each objective function (see Fig. 8 and Table 5). in thermal energy load while negligibly increasing deflections com-
Since the two objective functions are clearly competing, multi- pared to the minimum deflection result. It also shows a significant
objective optimization is warranted. reduction in deflections while only increasing the thermal energy
load by 12% compared to the minimum thermal energy load result.
3.3. Multi-objective results

Using the most robust combination of parameters from the sin-
gle objective optimization (SA 6), MOSA was performed. Fig. 9
shows the Pareto-optimal solutions (indicated by the “+” symbols).
Changes in geometry affect both deflections and thermal energy
load of the shelter. The major trend in this data is that as the wing
wall width is decreased (moving left on the Pareto-optimal plot),
the shelter volume decreases resulting in a much lower thermal
4. Comparison with existing shelter energy efficiency
Most conventional soft wall shelters consume a great deal of
energy in heating and cooling due to their poor insulation and
inefficient design. This section compares the thermal energy loads
of the LSM and existing soft wall shelters including the Tent,
Extendable, Modular, Personel (TEMPER, Fig. 1(A)) and TEMPER

Fig. 9. Pareto-optimal solutions. The “+” symbols indicate the Pareto-optimal set of solutions. The filled diamond represents the final optimized design.
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745 743

1500 1500
Heating energy
Heating energy
Cooling energy
Cooling energy

1000 1000
Fw [W⋅h/m2]

Fs [W⋅h/m ]
2
500 500

0 0
TEMPER TEMPER Air LSM TEMPER TEMPER Air LSM

Fig. 10. Energy consumption per unit usable floor area (W h/m2 ) for a day in winter Fig. 11. Energy consumption per unit usable floor area (W h/m2 ) for a day in summer
(01/21) in Chicago, USA. (07/30) in Chicago, USA.

Air-Supported tents (TEMPER Air, Fig. 1(B)). Both the TEMPER and
TEMPER Air tents have a width of 6.09 m and length of 9.75 m at
their bases. The TEMPER tent has a height of 2.06 m at its eaves and
3.07 m at its ridge, while the TEMPER Air tent has a half cylinder
shape with a radius of 3.05 m [74,75]. The material used in a typ-
ical TEMPER tent is polyester duck, and that in a typical TEMPER
Air is polyvinyl chloride (PVC) coated polyester [5]. To compare
these systems, this investigation assumes that both TEMPER and
TEMPER Air tents use polyester as the construction material with
a thickness of 7.6 mm. To make an equivalent comparison with
the LSM, 20 optimized modules are configured longitudinally to
achieve a comparable usable floor area of 59.5 m2 . The material
properties of polyester include a density  = 1260 kg/m3 , a specific
heat C = 1200 J/kg K, and thermal conductivity k = 0.05 W/m K [76].
Table 3 summarizes the thermal properties of the LSM. For a point
of practical comparison, the reader can consider the heat transfer
coefficient [66], or U-value, for both the LSM and existing shelters.
Computing an overall skin U-value requires the use of an exter-
nal convection heat transfer coefficient which depends on wind
speeds and exterior wall roughness. Nevertheless, we assume the
same environmental external conditions for the LSM and existing
shelters in the calculations of U-values. If average values of 18.0
and 3.10 W/m2 K are chosen for exterior and interior convection
coefficients (based on the equations provided in [68]; a medium
rough surface and wind speed of 2 m/s are assumed for exterior
convection calculation; a vertical wall is assumed for interior con-
vection calculation) respectively, then the U-value for the LSM is
0.507 W/m2 K and for the existing shelters is 1.86 W/m2 K. This
shows that the LSM features a significantly lower U-value compared
to these existing systems.
summer time, Fig. 4 shows that the weather is hot during the day
and cool during the night, indicating that ideally cooling is needed
during the day while heating is needed during the night so as to
maintain the desired temperature range. We note that from a prac-
tical standpoint, heating and cooling loads are seldom applied on
the same day. Fig. 11 demonstrates that the LSM can save energy up
to 80%. The cooling load of the LSM is reduced compared with the
other two existing tents due to its good insulation. Also, it can be
seen that no heating is needed in the LSM, which can be explained
as follows. The large thermal capacitance of the material enables
storage of heat absorbed during the day and released during the
night when it is required. Fig. 12 shows that the annual energy
demand of LSM is 70% less compared with the other two tents.
While the main focus of this section is to demonstrate the sig-
nificant energy savings for the LSM compared to the two existing
shelters, it is also interesting to notice the slight difference in the
thermal energy load between the two TEMPER tents. Fig. 11 shows
that TEMPER tent has a relatively larger cooling load than the TEM-
PER Air tent in the summer. This is because, with equivalent usable
floor surface, the TEMPER tent has a larger exterior surface area
than TEMPER Air tent, therefore, TEMPER tent receives much more
solar radiation resulting in the need for higher cooling loads in the
summer. Fig. 10 demonstrates that the TEMPER Air tent requires
500
Heating energy
450 Cooling energy
400
350
Fy [kW⋅h/m ⋅ year]

EnergyPlus [48] is used to perform the energy simulation com- 300

parisons for the three shelters: TEMPER tent, TEMPER Air tent,
2

and LSM. The parameter settings are the same as those used in 250
the optimization process (Section 2.6 and Table 2). This compar-
200
ison evaluates the energy consumptions for a day in the winter
(January 21), a day in the summer (July 30), and over a complete 150
year, denoted as Fw , Fs , and Fy , respectively. Figs. 10–12 are the
100
energy consumption comparisons among the three structures. They
all show that LSM provides better insulation compared with the 50
other two existing tents, which have similar energy use to one
another. Fig. 10 shows that the existing tents do not provide good 0
TEMPER TEMPER Air LSM
insulation when there is a large difference in temperature between
the inside and outside of the tent in winter (see the temperature Fig. 12. Annual energy consumption per unit usable floor area (kW h/m2 year) in
variation in Fig. 4) while the LSM can save energy up to 60%. In Chicago, USA.
744 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745
more heating loads than the TEMPER tent. This is because TEMPER
Air tent has a relatively larger indoor air volume than TEMPER tent,
thus, it requires more loads to heat the indoor air to maintain the
desired temperature conditions in the winter when only heating
is needed. But overall, the TEMPER Air tent has a slightly smaller
annual energy consumption than the TEMPER tent.
5. Conclusion
This paper has presented an optimization methodology which
balances the priorities of structural performance and energy effi-
ciency. This is a very useful strategy in the preliminary design stages
to better understand the effects that differing geometries have on
the structural and energy performance of a building. This allows the
designer to be better informed before detailed design and analysis
is undertaken and valuable time is lost.
The methodology has been demonstrated for the case study of
an origami-inspired deployable shelter where good structural per-
formance and energy savings are particularly critical. The paper
has shown that the methodology results in a reasonable choice
which balances the objectives of minimizing deflection and thermal
energy load. It was determined that both a low deflection as well as
a low thermal energy load could be achieved. The final optimized
design still fits within a 463L pallet and will interface with a Tricon
container satisfying the main constraints imposed by this study.
Fig. 8 compares the geometries for the single and multi-objective
optimizations. The multi-objective result clearly shows the com-
promise between the two single objective solutions. A comparison
between the optimized LSM form and the energy efficiency of exist-
ing soft wall shelters for a comparable floor area shows significant
energy savings.
This multi-objective optimization methodology, in combination
with structural and energy analysis software packages, is a valu-
able preliminary design tool to search for a geometry that exhibits
both low overall deflections and low thermal energy load. In order
to move forward with the final optimized design, an extensive 3-
dimensional finite element structural analysis must be completed
similar to that performed in Quaglia et al. [6]. This would include
material design (i.e., the thickness of the face and core, which were
assumed constant in the optimization process) to meet the require-
ments of a design code.
Acknowledgements
This material is based upon work supported by the US Army
Natick Soldier Research, Development and Engineering Center
(NSRDEC) under Contract W911QY-12-C-0128 (PAO #U14-258).
The authors are grateful for the assistance of Postdoctoral Research
Associates Zach Ballard and Saran Salakij and graduate student
Casey Casias. The authors would like to thank Lyman-Morse Boat-
building Co. (Thomaston, ME) for their contribution related to
material selection for the core and face of the sandwich panels. The
authors would also like to acknowledge the other principal inves-
tigators on this project, including Mihir Sen, Panos Antsaklis, and
Ann-Marie Conrado and project coordinator, Patrick Murphy.
References
[1] S.M. Anderson, Save Energy, Save Our Troops, 2011 http://www.nytimes.
com/2011/01/13/opinion/13anderson.html? r=0 (05.06.14).
[2] B. Severson, A. St. Leger, Feasibility study of photovoltaic panels in military
temporary housing structures, in: Proceedings of the IEEE Green Technologies
Conference, Denver, CO, 2013, pp. 78–84.
[3] Standard Fuel Prices in Dollars FY 2014, Presidents Budget FY 2014, 2014
http://www.energy.dla.mil/dla finance energy/documents/fy14 oct 01 2013
standard prices v3.pdf (16.05.14).
[4] P. Dimotakis, R. Grober, N. Lewis, Reducing DoD Fossil-Fuel Dependence, JSR-
06-135, 2006 http://www.fas.org/irp/agency/dod/jason/fossil.pdf (16.05.14).
[5] Joint Committee on Tactical Shelters, Department of Defense Standard Fam-
ily of Tactical Shelters (Rigid/Soft/Hybrid), US Army Natick Soldier Research,
Development, and Engineering Center, Natick, MA, 2012.
[6] C.P. Quaglia, A.J. Dascanio, A.P. Thrall, Bascule shelters: a novel erection strategy
for origami-inspired structures, Engineering Structures 75 (2014) 276–287.
[7] Compliance Packaging International Ltd., 463L Pallets, 2013
http://www.463lpallet.com/ (05.06.14).
[8] F.J. Martinez-Martin, A.P. Thrall, Honeycomb core sandwich panels for origami-
inspired deployable shelters: multi-objective optimization for minimum
weight and maximum energy efficiency, Engineering Structures 69 (2014)
158–167.
[9] United States Army Integrated Logistics Support Center, Natick Force Provider,
2013 http://ilsc.natick.army.mil/forceprovider.htm (17.11.13).
[10] Charleston Marine Containers Inc., Tricons, 2011 http://www.cmci.com/
Tricons.aspx (17.11.13).
[11] M. Adamski, W. Marks, Multicriteria optimization of shapes and structures of
external walls of energy conservation buildings, Archives of Civil Engineering
39 (1) (1993) 77–91.
[12] H. Jedrzejuk, W. Marks, Analysis of the influence of the service life and shape of
buildings on the cost of their construction and maintenance, Archives of Civil
Engineering 40 (3-4) (1994) 507–518.
[13] M. Adamski, Optimization of the form of a building with an arbitrary base,
Engineering Transactions 42 (4) (1994) 359–376.
[14] W. Marks, Multicriteria optimisation of shape of energy-saving buildings,
Building and Environment 32 (4) (1997) 331–339.
[15] M. Adamski, Optimization of the form of a bulding on an oval base, Building
and Environment 42 (4) (2007) 1632–1643.
[16] N. Bouchlaghem, Optimising the design of building envelopes for thermal per-
formance, Automation in Construction 10 (1) (2000) 101–112.
[17] L.G. Caldas, L.K. Norford, A design optimization tool based on a genetic algo-
rithm, Automation in Construction 11 (2) (2002) 173–184.
[18] L.G. Caldas, L.K. Norford, Genetic algorithms for optimization of building
envelopes and the design and control of HVAC systems, Journal of Solar Energy
Engineering 125 (3) (2003) 343–351.
[19] D.A. Coley, S. Schukat, Low-energy design: combining computer-based opti-
misation and human judgement, Building and Environment 37 (12) (2002)
1241–1247.
[20] D.E. Grierson, S. Khajehpour, Method for conceputal design applied to
office buildings, Journal of Computing in Civil Engineering 16 (2) (2002)
83–103.
[21] J.J. Michalek, R. Choudhary, P.Y. Papalambros, Architectural layout design opti-
mization, Engineering Optimization 34 (5) (2002) 461–484.
[22] A.M. Malkawi, R.S. Srinivasan, Y.K. Yi, R. Choudhary, Decision support and
design evolution: integrating genetic algorithms, CFD and visualization,
Automation in Construction 14 (1) (2005) 33–44.
[23] W. Wang, R. Zmeureanu, H. Rivard, Applying multi-objective genetic algo-
rithms in green building design optimization, Building and Environment 40
(11) (2005) 1512–1525.
[24] W. Wang, H. Rivard, R. Zmeureanu, Floor shape optimization for green building
design, Advanced Engineering Informatics 20 (4) (2006) 363–378.
[25] K. Shea, A. Sedgwick, G. Antonuntto, Multicriteria optimization of paneled
building envelopes using ant colony optimization, in: Intelligent Computing
in Engineering and Architecture, Lecture Notes in Computer Science, vol. 4200,
Ascona, Switzerland, 2006, pp. 627–636.
[26] C. Diakaki, E. Grigoroudis, D. Kolokotsa, Towards a multi-objective optimization
approach for improving energy efficiency in buildings, Energy and Buildings 40
(9) (2008) 1747–1754.
[27] Y.K. Yi, A.M. Malkawi, Optimizing building form for energy performance based
on hierarchical geometry relation, Automation in Construction 18 (6) (2009)
825–833.
[28] A. Fialho, Y. Hamadi, M. Schoenauer, Optimizing architectural and struc-
tural aspects of buildings towards higher energy efficiency, in: Proceedings
of the 13th Annual Genetic and Evolutionary Computation Conference, Dublin,
Ireland, 2011, pp. 727–732.
[29] M. Krem, S.T. Hoque, S.R. Arwade, S.F. Brena, Structural configuration and build-
ing energy performance, Journal of Architectural Engineering 19 (1) (2013)
29–40.
[30] K. Vergauwen, L. Alegria Mira, K. Roovers, N. De Temmerman, Parametric design
of adaptive shading elements based on curved-line folding, in: Proceedings of
the 1st Conference Transformables, Seville, Spain, 2013, pp. 337–342.
[31] F. Flager, B. Welle, P. Bansal, G. Soremekun, J. Haymaker, Multidisciplinary pro-
cess integration and design optimization of a classroom building, Journal of
Information Technology in Construction 14 (2009) 595–612.
[32] National Renewable Energy Laboratory, BEopt, 2014 https://beopt.nrel.gov/
(03.06.14).
[33] National Renewable Energy Laboratory, Opt-E-Plus Software for Commer-
cial Building Optimization, 2010 http://www.nrel.gov/tech deployment/pdfs/
45620.pdf (03.06.14).
[34] L. Caldas, GENE ARCH: an evolution-based generative design system for
sustainable architecture, in: Intelligent Computing in Engineering and Archi-
tecture, Lecture Notes in Computer Science, vol. 4200, Ascona, Switzerland,
2006, pp. 109–118.
[35] Lawrence Berkeley National Laboratory, GenOpt: Generic Optimization Pro-
gram, 2011 http://simulationresearch.lbl.gov/GO/ (03.06.14).
[36] S. Bambardekar, U. Poerschke, The architect as performer of energy simulation
in the early design stage, in: Proceedings of the Eleventh International Building
 C.P. Quaglia et al. / Energy and Buildings 82 (2014) 733–745 745

Performance Simulation Association Conference, Glasgow, Scotland, 2009, pp. [58] N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, E. Teller, Equation
1306–1313. of state calculation by fast computing machines, Journal of Chemical Physics
[37] F. Lucas, T.A. Mara, F. Garde, H. Boyer, A comparison between CODYRUN and 21 (6) (1953) 1087–1092.
TRNSYS, simulation models for thermal buildings behaviour, CoRR (2012), [59] J.R. Medina, Estimation of incident and reflected waves using simulated anneal-
2012;abs/1212.5255. ing, Journal of Waterway, Port, Coastal, and Ocean Engineering 127 (4) (2001)
[38] B.J. Urban, The MIT design advisor: simple and rapid energy simulation of early- 213–221.
stage building designs (Master’s thesis), Massachusetts Institute of Technology, [60] A. Suppapitnarm, K.A. Seffen, G.T. Parks, P.J. Clarkson, Simulated annealing algo-
Cambridge, MA, 2007. rithm for multi-objective optimization, Engineering Optimization 33 (1) (2000)
[39] T.R. Nielsen, Simple tool to evaluate energy demand and indoor envi- 59–85.
ronment in the early stages of building design, Solar Energy 78 (2005) [61] L. Alegria Mira, N. De Temmerman, C. Preisinger, Structural optimization of
73–83. deployable scissor structures using new computational methods, WIT Trans-
[40] T.R. Nielsen, C.A. Hviid, S. Petersen, BuildingCalc LightCalc User Guide, Technical actions on the Built Environment 124 (2012) 469–480.
Report, Danmarks Tekniske University, 2008. [62] L. Alegria Mira, N. De Temmerman, R. Filomeno Coelho, Integrated framework
[41] Sketchup, 2013 http://www.sketchup.com/ (18.06.14). for the sensitivity analysis of deployable planar scissor arches, in: Proceedings
[42] D. Greenberg, K. Pratt, B. Hencey, N. Jones, L. Schumann, J. Dobbs, Z. Dhong, of the IASS Symposium Wroclaw, Beyond the limits of man, Wroclaw, Poland,
D. Bosworth, B. Walter, Sustain: an experimental test bed for buildng energy 2013.
simulation, Energy and Buildings 58 (2013) 44–57. [63] L. Alegria Mira, A.P. Thrall, N. De Temmerman, Deployable scissor arch for
[43] Minimum Design Loads for Buildings and Other Structures: ASCE Standard 7- transitional shelters, Automation in Construction 43 (2014) 123–131.
10, Structural Engineering Institute, USA, 2010. [64] C.P. Quaglia, Novel, deployable origami-inspired shelters for forward operating
[44] US Army Natick Soldier Research, Development and Engineering Cen- bases: design and optimization (Master’s thesis), University of Notre Dame,
ter, Commander’s Smartbook Equipment Catalog, 2009 http://nsrdec.natick. Notre Dame, IN, 2014.
army.mil/media/print/Smartbook Web.pdf (07.11.12). [65] Lawrence Berkeley National Laboratory, Auxiliary EnergyPlus Programs, 2013.
[45] D. Rutten, Robert McNeel & Associates, Grasshopper, 2012 http://www. [66] F.P. Incropera, D.P. Dewitt, T.L. Berman, A.S. Lavine, Fundamentals of Heat and
grasshopper3d.com/ (05.06.13). Mass Transfer, John Wiley and Sons, New York, 2007.
[46] Robert McNeel & Associates, Rhinoceros, 2012 http://rhino3d. [67] J.H. Lienhard IV, J.H. Lienhard V, A Heat Transfer Textbook, Phlogiston Press,
nl/pythposter/pyth3dm-eng.html (05.06.13). Cambridge, MA, 2012.
[47] C. Preisinger, M. Heimrath, M. Tam, R. Vierlinger, C. Zimmel, Karamba, 2014 [68] EnergyPlus Engineering Manual, 2013 http://apps1.eere.energy.gov/buildings/
http://www.karamba3d.com/ (05.06.14). energyplus/pdfs/engineeringreference.pdf (19.05.14).
[48] US Department of Energy, EnergyPlus Energy Simulation Software, 2013 [69] Vectorply – Performance Composite Reinforcements, Laminated Properties and
http://apps1.eere.energy.gov/buildings/energyplus/energyplus about.cfm Packaging, 2002 http://www.vectorply.com/ri-laminateprop.html (09.10.13).
(05.06.14). [70] Gurit, Corecell M-Foam – The Marine Foam, 2013 http://www.gurit.com/files/
[49] M. Marvin, D. Orvieto, B. Silverman, B. Albano, Gerilla, 2014 http:// documents/corecell-m-foamv5pdf.pdf (09.10.13).
www.food4rhino.com/project/gerilla (05.06.14). [71] H.G. Allen, Analysis and Design of Structural Sandwich Panels, Pergamon Press,
[50] C.P. Quaglia, Z.C. Ballard, A.P. Thrall, Parametric modelling of an air-liftable Oxford, UK, 1969.
origami-inspired deployable shelter with a novel erection strategy, in: WIT [72] Owens Corning, Composite Solutions Reinforcement Guide, 2011 http://www.
Transactions on the Built Environment, vol. 136, Proceedings of the 4th Inter- ocvreinforcements.com/pdf/library/Composite Solutions E finalprintable.pdf
national Conference on Mobile, Adaptable and Rapidly Assembled Structures, (19.05.14).
Ostend, Belgium, 2014, pp. 23–34. [73] I. Paya-Zaforteza, V. Yepes, A. Hospitaler, F. Gonzalez-Vidosa, CO2 -optimization
[51] S. Kirkpatrick, C.D. Gelatt, M.P. Vecchi, Optimization by simulated annealing, of reinforced concrete frames by simulated annealing, Engineering Structures
Science 220 (4598) (1983) 671–680. 31 (7) (2009) 1501–1508.
[52] K. Shea, I.F.C. Smith, Improving full-scale transmission tower design through [74] Department of the Army, Air Force, and Navy, Operator, Unit, and Direct
topology and shape optimization, Journal of Structural Engineering 132 (5) Support Maintenance Manual for Tent, Extendable, Modular, Personnel
(2006) 781–790. (TEMPER): TM 10-8340-224-13, 1993 http://www.liberatedmanuals.com/
[53] I. Paya, V. Yepes, F. Gonzalez-Vidosa, A. Hospitaler, Multi-objective optimiza- TM-10-8340-224-13.pdf (04.06.14).
tion of concrete frames by simulated annealing, Computer-Aided Civil and [75] US Army Natick Soldier Research, Development & Engineering Center, Guide
Infrastructure Engineering 23 (8) (2008) 596–610. for Tactical Training Bases, Shelters Handbook, 2008 http://nsrdec.natick.
[54] M. Ohsaki, H. Tagawa, P. Pan, Shape optimization of reduced beam section army.mil/media/print/ShelterGuide.pdf (0506.14).
under cyclic loads, Journal of Constructional Steel Research 65 (7) (2009) [76] Ogden Manufacturing Co., Physical Properties of Materials, 1997
1511–1519. http://www.ogdenmfg.com/pdf/tech9.pdf (07.05.14).
[55] A.P. Thrall, S. Adriaenssens, I. Paya-Zaforteza, T.P. Zoli, Linkage-based movable [77] C.P. Quaglia, A.P. Thrall, Shape optimization of an origami-inspired deployable
bridges: design methodology and three novel forms, Engineering Structures 37 shelter for minimum deflections, in: Proceedings of the International Associ-
(2012) 214–223. ation Shells for Spatial Structures – Latin American Symposium on Tension
[56] A.P. Thrall, M. Zhu, J.K. Guest, I. Paya-Zaforteza, S. Adriaenssens, Structural opti- Structures Symposium, Brasilia, Brazil, 2014.
mization of deploying structures composed of linkages, Journal of Computing [78] T. Pepper, Polyester resins, ASM Handbook 21 (2001) 90–96.
in Civil Engineering 28 (3) (2014) 04014010. [79] H. Winterling, N. Sonntag, Rigid polystyrene foam (eps, xps), Kunstoff Interna-
[57] B.R. Russell, A.P. Thrall, J.A. Padula, J.E. Fowler, Reconceptualization and tional 10 (2011) 18–21.
optimization of a rapidly deployable floating causeway, Journal of Bridge Engi-
neering 19 (4) (2014) 04013013.