You are on page 1of 24

Res Eng Design

DOI 10.1007/s00163-017-0269-0

ORIGINAL PAPER

A digital thread approach to support manufacturing-influenced


conceptual aircraft design
Dennis J. L. Siedlak1 • Olivia J. Pinon1 • Paul R. Schlais1 • Todd M. Schmidt1 •

Dimitri N. Mavris1

Received: 6 July 2016 / Revised: 27 August 2017 / Accepted: 31 August 2017


 Springer-Verlag London Ltd. 2017

Abstract This paper proposes and demonstrates the inte- Keywords Design process  Aerospace design 
gration of manufacturing and production considerations Manufacturing  Production  Affordability  Digital thread
with traditional aircraft design metrics to support afford-
ability-based design. To enable the necessary multi-disci-
plinary trades, a digital thread approach is proposed that 1 Introduction
integrates detailed models and analyses. The digital thread
refers to linking models from various disciplines through This paper proposes and demonstrates a design approach
common inputs and data flows with the goal of speeding that incorporates manufacturing and production consider-
design time and enabling trades across traditionally iso- ations into the early aircraft design process. This is
lated disciplines. When used within an overarching design accomplished by integrating detailed aircraft wing design,
process, the production cost, rate, and efficiencies of non- process-based costing, and production simulation tools.
conventional designs in variable demand environments can Surrogate modeling is leveraged to enable detailed results
be quantified and traded early in the design process. In to be used for early design, optimization, and decision
particular, the methodology is demonstrated using a support. These capabilities support the aerospace and
wingbox design problem such that aircraft performance defense industries in their shift towards affordability-based
considerations, production rate, manufacturing cost, and design (Air 2016; LCA 2015) and the design of resilient
financial planning metrics can be traded within a para- systems, which entail systems that ‘‘need to be manufac-
metric, visual trade-off environment. The environment, turable, readily deployable, sustainable, easily modifiable,
combined with a multi-objective optimization routine, and cost-effective’’ (Goerger et al. 2014).
facilitates effective affordability-based tradespace explo- Affordability can be thought of as the balance between
ration during the early stages of the design of non-tradi- cost and performance throughout the product’s life-cycle
tional aircraft (e.g., those utilizing composite structures) (Defense Acquisition University 2012; Blair and Hartong
under demand variability. An F-86 Sabre redesigned 2000; Schaffner et al. 2013; Tuttle and Bobinis 2013), so a
wingbox using three separate manufacturing concepts is more costly system that provides better capabilities may be
used as a proof-of-concept for this research. more affordable than a less expensive option. Recent
designs, however, have experienced schedule delays and
cost overruns (Captain 2009). Two of the many issues that
contribute to the lack of aircraft affordability, namely (1)
Research conducted while attending the Georgia Institute of
Technology.
the expanded use of composite materials and (2) a large
and increasing amount of variability and uncertainty in
& Dennis J. L. Siedlak product demand, are addressed through the presented
dsiedlak3@gatech.edu approach.
1 In the traditional aircraft design paradigm, up to 2/3 of a
Aerospace Systems Design Laboratory, The Daniel
Guggenheim School of Aerospace Engineering, Georgia system’s cost is committed before Detailed design (Mavris
Institute of Technology, Atlanta, GA, USA et al. 1998; Stebbins 2015; Asiedu and Gu 1998) and

123
Res Eng Design

70–90% before production begins (Marx and Thut 2012). 2 Literature review
However, as seen in the literature, traditional design
methods struggle to quantify the impact of design choices The aerospace and defense industry is faced with the need
when designing unconventional vehicles in new physical to deliver products with superior performance (e.g., higher
and economic environments (Briceno et al. 2013; Rush and speed, longer range, reduced fuel burn, etc.) at a reduced
Roy 2000; Eaglesham 1998; Peters et al. 1994; Sundaresan cost (Raj 1998; Usher et al. 1998; Curran et al. 2001;
et al. 2012; La Trobe-Bateman and Wild 2003; F35 2015; Siedlak et al. 2015; Air 2016; LCA 2015). Many design
Curran et al. 2001; Campbell 2003; Siedlak et al. techniques have been developed to support this objective.
2015, 2014; Werner-Westphal et al. 2008). In other words, Such techniques include Product Lifecycle Management
designers are losing the ability to capture and understand (PLM) (Stark 2005), Integrated Product and Process
the cost-capability trades before costs are committed. Development (IPPD) (Magrab et al. 2009; Usher et al.
The proposed non-weight-based methodology integrates 1998), and the Lean Aircraft Initiative (LAI) (Usher et al.
parametric, physics-based structural design and aircraft 1998). These techniques have been successfully imple-
performance models alongside process-based costing and mented on programs such as the Boeing 777 (Usher et al.
discrete-event simulation into a single digital thread. The 1998) and F/A-18 E/F (Haggerty 2004), both of which
digital thread refers to the ‘‘communication framework that were ahead of schedule and under budget. However,
allows a connected data flow and integrated view of the despite the successes of the 90’s, the early twenty-first
asset’s data throughout its life-cycle across traditionally century has seen multiple aerospace programs, such as the
siloed functional perspectives (Leiva 2016).’’ By providing F-35 Joint Strike Fighter (Oppenhein 2011; F35 2015),
better cost estimates during early design and better col- Boeing 787, and Airbus A380 (Michaels 2012), plagued
laborative decision-support capabilities, the methodology with cost overruns and delivery delays (Wu 2014). While
and design environment developed throughout this research IPPD and PLM have had some successes, these approaches
help support the primary needs identified by the Model- tend to focus too often on reducing costs and not improving
based Systems Engineering (MBSE) and Engineered affordability (Usher et al. 1998; F35 2015). In other words,
Resilient Systems (ERS) communities (Holland 2015; while these and similar techniques do analyze cost during
Neches 2011; Spero et al. 2014b; Ramos et al. 2012; design, the analysis generally is conducted in a siloed
Goerger et al. 2014). In particular, by enabling the inte- manner (Spero et al. 2014b). To reverse this trend,
gration of various design tools early in the design process, methodologies are being developed and implemented that
this methodology represents a first significant step towards integrate quantitative analysis from a myriad of disciplines
achieving the goals of the digital thread, namely ‘‘more to investigate many aspects of a system’s affordability and
producible, serviceable, usable, sustainable, safe and resiliency during design (Goerger et al. 2014; Spero et al.
lower-cost designs with shorter product development 2014a, b; Ramos et al. 2012). The following section
cycles and fewer design iterations (Stebbins 2015).’’ reviews such methodologies and discusses research needs
The paper is structured as follows: Sect. 2 reviews as they pertain to improving early system design.
existing approaches and discusses research opportunities
and methodology needs. From this review, the core prin- 2.1 Multi-disciplinary design methodologies
ciples of the proposed methodology are presented in Sect.
3. Section 4 then discusses how the general design Model-based systems engineering has arisen as an impor-
methodology is applied to the wing design problem. Sec- tant design tool for complex, interdisciplinary systems.
tion 5 provides a description of the analytical capabilities MBSE seeks an ‘‘integrated holistic approach (Ramos et al.
that this methodology enables and further discusses the 2012)’’ to help the ever-increasing number of stakeholders
parametric environment developed to support designers. A understand and design a new system (Ramos et al. 2012;
particular emphasis is brought on affordability-based trades Estefan et al. 2007). It can be described as ‘‘the formalized
between cost and capability that can help designers make application of modeling principles, methods, languages,
decisions over the product’s life-cycle. Section 6 discusses and tools to the entire life-cycle of large, complex, inter-
optimization results, hence highlighting the trades and disciplinary, sociotechnical systems (Ramos et al. 2012).’’
sensitivities that are now quantifiable during the early Hence, MBSE ultimately hopes to provide a common
design stages. The paper concludes by discussing potential interface between various models and disciplines to enable
implications of implementing this multi-disciplinary design a system-level understanding of the design’s performance
environment and proposed areas of future research. across domains.
The Department of Defense’s push towards the devel-
opment of engineered resilient systems has similar goals,

123
Res Eng Design

namely ‘‘to transform the engineering of complex systems on the relationship between design parameters and cost to
to make the resulting systems more affordable, effective, either assess the manufacturability of a design or provide
and adaptable (Spero et al. 2014b).’’ ERS seeks to improve estimates of the manufacturing costs. Das and Kan-
the effectiveness of engineering by advancing following chanapiboon, for example, developed a model for manu-
areas (Lucero 2011): (1) systems representation and mod- facturability analysis that estimates unit production cost. It
eling, (2) characterizing changing operational contexts, (3) assumes that the bill of materials part lists, assembly plan
cross-domain coupling, (4) data-driven tradespace explo- and initial costs estimates for part procurement and
ration and analysis, and (5) collaborative design and assembly rates are available (Das and Kanchanapiboon
decision support. 2011). Bao and Samareh (2000) developed a methodology
that integrates process-based manufacturing and assembly
2.2 Research needs cost models in a traditional performance-focused multi-
disciplinary design and optimization process. Curran et al.
In support of these strategic thrusts, many research (2007) discuss the development of a digital design and
opportunities have been suggested that are relevant to the manufacturing modeling platform with integrated cost
current work. First, the ‘‘-ilities’’ of a design should be modeling capabilities.
characterized and quantified (Spero et al. 2014b). Model- Past approaches that focus on the link between design
ing life-cycle costs and manufacturing has been specifically and manufacturing suffer from the following shortcomings.
identified as a technical challenge (Holland 2015). It is Many of these approaches are completely dependent on
ultimately desired to ‘‘build cross-impact models and existing data. As such, their applicability is limited to the
relationships that quantitatively determine the impact of range of available data. While they capture the relationship
changing one ility on the other ilities (Spero et al. 2014b).’’ between design parameters and manufacturing-related
Spero et al. specifically mentions the need to consider costs, most approaches do not allow for trades between cost
affordability during design (instead of the traditional sep- and performance to be conducted. This limits their ability
aration of performance and cost trades). Second, an to identify a feasible design and cost reduction opportuni-
approach to link the tradespace during conceptual design is ties. Many of these approaches also limit themselves to a
identified as a research need (Spero et al. 2014b; Neches manual exploration of the design/process space. As such,
2011). Especially during conceptual design, where freedom they do not support any sensitivity and requirements
is the highest, it is desired to have (1) a ‘‘quick assessment analyses and provide no information as to the interactions
of ‘goodness’ (Spero et al. 2014b),’’ (2) the ability to link that exist between the different disciplinary variables. The
to more detailed tools, and (3) the quantification of how approaches that do mention some optimization capabilities
domain-specific choices propagate to system-level perfor- either limit themselves to a discussion of the benefit they
mance (Spero et al. 2014b). could bring or only optimize a single criterion (cost).
Finally, the need to support collaborative decision- Finally, most approaches do not account for the pro-
making across disciplines through tradespace exploration is ducibility of a design, making it impossible to evaluate
identified (Spero et al. 2014a, b; Neches 2011). Relevant to whether the chosen design can be produced at the required
the current work is the desire to understand the system’s rate without incurring any additional costs. The ones who
performance across multiple disciplines for uncertain do account for producibility lack the flexibility to para-
futures (Spero et al. 2014b). Indeed, in addition to cost- metrically modify the production model and do not capture
capability trades, the resilience and affordability initiatives changes in demand, hence limiting their ability to identify
also contain a time element. As such, there is a need to design that are robust to demand variability.
assess the system’s performance against multiple, uncertain The following section discusses how the proposed
program futures or scenarios (Siedlak et al. 2015; Vascik methodology helps address the shortcomings and needs
et al. 2016; Bobinis et al. 2013; Wu et al. 2014; Tuttle and identified in the current literature:
Bobinis 2013; Libby et al. 2017).
1. Linkage of cross-disciplinary models and analyses
The general theme presented throughout these design
2. Quantification of the ‘‘-ilities’’ of a novel design
approaches and methodologies (PLM, IPPD, MBSE, and
3. Quick propagation of design choices through each
ERS) is the desire for better communication (among tools,
discipline
disciplines, and decision-makers) to support early design.
4. Improved design space exploration across disciplines
Most of past multi-disciplinary efforts have focused on
5. Support for collaborative decision making and data
the integration of design and manufacturing. Such meth-
analysis
ods, as discussed in (Das and Kanchanapiboon 2011), focus

123
Res Eng Design

3 Proposed methodology modeling. The model, depending on its level of fidelity,


may run slowly and may also encapsulate a large number
The goal of the proposed methodology is to provide an of variables. These two characteristics, in turn, tend to
integrated framework and collaborative design environ- significantly limit the amount of information that can be
ment capable of quickly assessing and visualizing the abstracted and rapidly visualized (Mavris et al. 2010). A
impact that changes in system design and factory config- successful approach that addresses these challenges, while
uration have on aircraft performance, production rate, providing an all-encompassing model for exploring a
backlog, and profitability. The developed environment, complex design space, is surrogate modeling (De Baets
thus, represents an integrated digital thread of analyses and Mavris 2000; Kamdar et al. 2003). Surrogate modeling
from aircraft performance through structural analysis and techniques, by constructing mathematical approximations
manufacturing cost to production rate over time. Through of analysis codes, supports the integration of discipline-
the integration of these analyses, the environment is cap- dependent and often organization-dependent codes while
able of comparing the affordability of proposed designs also representing an efficient way to lessen the time
under changing production conditions without a library of required for an integrated environment to run. These
relevant historical data. techniques also yield insight into the relationships between
Table 1 illustrates how the aircraft design and produc- design variables (inputs) and responses (outputs), hence
tion planning exercises can be aligned. It further summa- facilitating concept exploration (Simpson et al. 2001). By
rizes the kind of trades and decisions that can be made allowing analyses to run faster, such techniques also allow
during each phase (Sundaresan et al. 2012). The concep- for more detailed models and codes to be brought forward
tual design phase is mainly concerned with the identifica- in the design process and for integrative visual environ-
tion of a feasible design space and the selection of a ments to become parametric and interactive (Ligetti et al.
candidate design concept within that space. At this stage of 2003).
the design process, the design space is primarily con- The core principles of the proposed methodology, thus,
strained by performance requirements. Bringing production include:
considerations and, consequently, production-driven
• Replacing cost estimating relationships (CERs) and
requirements in the early design stages requires detailed
other historical data-based producibility estimates with
information about key production parameters. To be of any
bottoms–up, process-based costing and production
value, this information, which traditionally becomes
simulation tools
available much later in the design process, needs to be
• Integrating inputs, outputs, and models into a single
brought forward in a parametric fashion so as to enable the
digital thread to speed data collection and design space
quick evaluation of multiple design concepts.
exploration
For unconventional designs that lack reliable historical
data, such information is often generated through

Table 1 Characteristics and decisions made at each aircraft and production planning phase (Sundaresan et al. 2012)
Design phase Aircraft design Production phase Production planning

Conceptual Requirements analysis Strategic


Multiple concepts Production rate
Low fidelity Production cost
Rapid assessment Production capacity
Trade-off analysis
Preliminary Frozen configuration Tactical
Medium fidelity Preliminary production plan at work-station level
Sub-component validation and design Higher fidelity cost and production rate estimation
Test and analytical database development
Detailed High fidelity Operational
Design tooling and fabrication process
Design actual pieces Detailed work plan on the factory floor
Test major items
Performance engineering completed

123
Res Eng Design

• Generating surrogate models for all disciplines cap- off/landing field length) could potentially be relaxed to
tured within the design environment. This enables: reduce costs. Finally, the significant amount of information
available about such fighters (Fli 1960; Wagner 1998) can
• The creation of a real-time, collaborative tradespace
be used to support the development of the performance and
exploration environment
structural models.
• The use of analysis results from detailed tools for
The remainder of this section discusses the spiral
Conceptual design. For example, a surrogate model
development approach followed and the integration of
of production simulation results can replace the
models to estimate the performance, cost, and producibility
utility function commonly used in epoch-era anal-
of a wing design. Because of the impact that structural
ysis (EEA) (Vascik et al. 2016; Wu et al. 2014).
components have on manufacturing and production rates,
• Affordability-based, multi-objective optimization.
the focus will mainly be on the development and integra-
Surrogate models make objective function evalua-
tion of structural design, process-based costing, and pro-
tions essentially instantaneous; thus, metaheuristic,
duction simulation tools.
multi-objective optimization runs can be feasibly
completed.
4.1 Leveraging past efforts
The following section discusses the application of this
methodology to support the design of an F-86 Sabre’s This work builds on and realizes the Manufacturing-In-
wingbox. fluenced Design (MInD) (Ceisel et al. 2012; Briceno et al.
2013) and initial iterations of the Manufacturing-Influ-
enced Design Production Optimization (MInDPRO)
4 Methodology application frameworks (Siedlak et al. 2014, 2015) developed by the
Aerospace Systems Design Laboratory (ASDL) at the
The design of an F-86 Sabre’s wingbox (illustrated in Georgia Institute of Technology. Figure 2 presents an
Fig. 1) is used to demonstrate this methodology. The wing overview of the models and tools implemented within this
is designed such that the entire vehicle meets its perfor- framework. The original MInD framework encompasses all
mance requirements. The wing is also designed as to analyses presented in Fig. 2 except for the production flow
improve producibility. The reasons for selecting a wingbox model. The performance model is calibrated to the original
as a proof-of-concept are as follows: (1) there is a direct F-86’s performance, and the structural models contain
link between wing design and vehicle performance, (2) similar components (e.g. ribs, skins, spars, and stringers) as
there are a variety of structural components in a wingbox seen in the original F-86.
with many potential manufacturing processes, and (3) The primary contribution of the MInD framework is to
composite wing design is a current issue in the aerospace allow for process-based costing information to be brought
industry. forward in the design process. SEER for Manufacturing
Furthermore, a military aircraft is chosen because mis- (SEER-MFG) is a commercially available process-based
sion capability/performance versus cost versus producibil- time and costing tool that utilizes empirical data collected
ity trades are of great importance to the affordability-based under the U.S. Governments Composite Affordability Ini-
acquisition paradigm. While some survivability require- tiative to estimate manufacturing cycle times and costs
ments (such as turn rate or payload capacity) may be non- (Russell 2006). In particular, it provides the capability to
negotiable, others that may not directly impact the pilot’s estimate costs for a variety of structural components and
probability of surviving the mission (e.g., range and take- manufacturing process options (e.g., hand layup

Fig. 1 Example wingbox Wing Skin


structure
Stringers

Spars Ribs

123
Res Eng Design

Fig. 2 MInDPRO framework

vs. automated tape laying). It is also widely used within the for the production run, and (4) a multi-objective opti-
aerospace industry for cost estimation (Ceisel et al. 2012; mization routine. These capabilities are intended to tie the
Kaufmann et al. 2010). SEER-MFG requires detailed part previous efforts together into an approachable, collabora-
geometry information (e.g. lengths, thicknesses, ply tive design and optimization environment. The additional
tables for composite parts, etc.) to develop the cost esti- costing models are incorporated to better capture the
mations. Hence, a structural design model is included to infrastructure, equipment, and tooling costs associated with
translate outer mold line geometry (e.g., wing area, aspect the design (Heckwolf 2014) to support more meaningful
ratio, etc.) into this detailed information. The Detailed production cost analysis. By looking beyond a single year
Manufacturing Translator then supplements this data with of production, the framework enables the analysis of pro-
rules of thumb identified through interactions with subject duction runs across multiple time-spans. In doing so, the
matter experts. An example of such rules includes using the method can now consider the flexibility of a production
length of a joint to calculate the number of fasteners system and its ability to respond to changes in demand.
required. The reader is invited to consult (Ceisel et al. Furthermore, as detailed costs are broken down across a
2012; Lee et al. 2012; Sundaresan et al. 2012) for a full number of cost categories, cash-flows can be generated for
description of the MInD framework’s development and its a variety of designs, demand scenarios, and economic
capabilities. assumptions. Finally, because the product and process
The ensuing sections discuss how the authors leveraged design space spanning the entire production run is too large
and modified the MInD framework to address the chal- for effective manual exploration, a multi-objective opti-
lenges resulting from demand variability and to enable mization routine is developed. The optimization supports
more informed, affordability-based design decisions. sensitivity and requirements analyses to understand the
interactions between wing design, demand, requirements,
4.2 Manufacturing-influenced design production and factory setup and evolution modeled during previous
optimization (MInDPRO) multidisciplinary iterations of the methodology. The following sections
framework further describe the implementation of these capabilities.

The MInDPRO framework expanded upon the MInD 4.3 Production flow modeling
framework by including a module to provide production
rate and efficiency information. This is accomplished A major portion of the research methodology is concerned
through the development of a discrete-event simulation in with the development of a Preliminary production model.
Simio. The work presented herein further expands upon the After a literature review of various composite manufac-
MInDPRO framework by including: (1) additional costing turing and production methods (Anderson et al. 2004; Cole
models, (2) expansion from a yearly production rate to 2001; Sloan 2012), along with discussions with subject
production over time, (3) economic and cash-flow analysis

123
Res Eng Design

matter experts, a discrete-event model is developed in The skin/stringer and spar/spar stiffener lines follow a
Simio. similar process; with the only difference being that the
Simio is chosen as the modeling platform for the pro- components are co-cured together before proceeding to
duction component because it is capable of tracking key NDI.
production statistics, it has an interface that lowers the After each component passes through NDI, the final
learning curve compared to other discrete event simulation wingbox assembly begins. The steps for final assembly are:
programs, and it easily links to excel and can be run in
1. Rib sub-components (intercostals and C section) are
server mode to facilitate running many cases for a Design
paste bonded
of Experiments. Figure 2 illustrates how the Production
2. Leading edge spar and lower skin are paste bonded
Planning Optimization (PRO) component (in green on the
3. Ribs are paste bonded to the leading edge spar and
figure) is integrated into the existing MInD framework. The
lower skin sub-assembly
following section discusses the logic and flow between the
4. Trailing edge spar is paste bonded to the sub-assembly
geometry, manufacturing, and production elements before
5. Upper skin is paste bonded to the sub-assembly to form
describing in more detail the modeling of the production
the final wingbox assembly
flow itself.
Finally, the following assumptions are made. Because
the goal of this methodology is to support decision making
4.3.1 Framework overview during early design phases, the factory floor is assumed to
be spatially unconstrained. Additionally, simplifying
Outer mold line (OML) dimensions (e.g., wing area, aspect assumptions about the structures are made. It is assumed
ratio, etc.) are fed into MInDPRO’s structural optimization that each rib is the same size, and, therefore, requires the
models to generate more detailed part dimensions same amount of processing time. The processing times for
(e.g., spar length, ply count, rib quantity, etc.). These the upper and lower skins are also assumed to be the same.
dimensions are then used as inputs to SEER-MFG process- The leading and trailing edge spars, however, are modeled
based models to calculate processing times for each man- separately since there is a more significant difference
ufacturing process. The calculated times from SEER are between their structures then that of the skins.
used to define the process time at each station within the The following section describes the manufacturing line
production flow model. The authors are not aware of pre- balancing methodology that is implemented to ensure that
vious design methods that use production times from a the process proceeds in a realistic fashion for a wide range
process-based costing tool to gain insight into the choice of of DoE inputs.
a design concept through DES. Finally, inputs specific to
the production module (e.g., number of technicians, 4.3.2 Manufacturing line balancing
workstations, shifts, etc.) are entered, and the simulation is
run to estimate the production rate of the current wing Manufacturing line balancing methodologies are applied to
design and factory setup. the production flow model to ensure that the total work in
The model contains three lines accounting for the major process remains reasonable even when the DoE provides
components of the wingbox: spars, skin/stringers, and the poorly balanced inputs. This is accomplished by utilizing
ribs. A high-level view of the factory layout is illustrated in real-time process logic to dynamically balance the pro-
Fig. 3. Within the rib line, the tool goes through the fol- duction lines. The logic is based on concepts from surplus
lowing production steps: capacity and theory of constraints to achieve a low Work-
In-Process (WIP) inventory (Lee 2000; Rahman 1998).
1. Tool cleaning and preparation One of the main goals of this logic is to most efficiently use
2. Composite hand layup the production resources to maximize throughput while
3. Layup inspection reducing excessive WIP.
4. Bagging/prepare for cure The logic stops subassembly lines (rib, skin, or spar line)
5. Vacuum-assisted Resin Transfer Molding (VaRTM) when a line creates a significant excess WIP inventory.
resin infusion and curing This logic effectively mimics the decisions that a produc-
6. Debag and inspect cure tion manager would make and allows the model, which
7. Water-Jet trimming would otherwise be extremely sensitive to unbalanced
Following this, the tool and rib component are separated. production times, to accept a wide range of DoE inputs. For
The component goes on to a non-destructive inspection example, if the number of skins produced is currently at 20
(NDI) station, while the tool is sent back to the beginning (enough to create 10 wings) and the spar line just finished
of the process to begin a new cycle. producing the 40th spar (enough for 20 wings), the model

123
Res Eng Design

Rib C Section and


Intercostals Hand Rib C Section and Rib C Section and Rib Subassembly
Layup Intercostals VaRTM Intercostals NDI

Spar Stiffener Hand Ribs to Bottom


Layup Skin and LE Spar
Spar VaRTM Co-
Bond
TE Spar to Ribs,
Spar C Section Bottom Skin to LE
Bottom Skin, and
Hand Layup Spar Pastebond
LE Spar Pastebond
Spar and Skin NDI

Stringer Hand Pastebond Top Skin


Layup to Assembly
Skin VaRTM Co-
Bond

Skin Hand Layup

Fig. 3 Production model factory layout

stops the spar line to allow for the skin line to catch up. The person. With three workers, it is assumed that the pro-
threshold of 10 wings was chosen to maintain a balance cessing time is now 40% of the original time.
between low WIP and low schedule uncertainty. The par- With these assumptions, process logic is implemented
allel lines (e.g., Spar C Section and Spar Stiffener parallel to balance the workers based on the queues forming at
lines) within each sub-assembly line are balanced follow- each sub-component station. To simplify the model,
ing the same logic. workers are assigned only at the beginning of each pro-
Since the component hand layup stations account for a cessing step and are required to remain for the duration of
significant amount of each line’s processing time, and the process. At least one worker is assigned to each sta-
given that the processing time is directly impacted by the tion that has a tool ready for layup, with the remaining
number of workers laying up a part, it is important that the workers being assigned to workstations that have longer
workers at these stations be modeled to help balance the queues. The model also tracks the station that processes
lines. Using a floating worker balance (Lee 2000; Zavadlav the most parts (i.e., the fastest line) and only allows one
et al. 1996), workers are allowed to move between each worker to be assigned there to ensure that processes that
layup station within their line when needed. For instance, a are lagging behind have access to the additional resour-
rib hand layup worker is allowed to work on upper inter- ces. It is important to note that these line balancing
costals, lower intercostals, or rib C sections. Because hand methodologies do not decrease the final throughput of the
layup is a manual process that can be sped up by increasing factory floor because stations on the critical path are
the number of workers, up to three workers are allowed to never shut down. In other words, the model is able to
work on a single component at one time. The estimated maintain a low WIP without negatively impacting
processing time that is taken from SEER assumes one throughput. This enables the model to mimic a real-life
worker, but it does not allow the user to change this factory floor throughout the wide range of DoE inputs
assumption to modify the processing time. After discus- without producing massive queues that would skew effi-
sions with subject matter experts, approximations were ciency and waiting time results.
made about the percent improvement to the processing The following section discusses how Design of Exper-
time that can be seen when adding workers. With two iments (DoE) and surrogate modeling techniques were
workers at a workstation, it is assumed that the process implemented to speed up the simulation runs, provide a
takes only 60% of the original processing time with a parametric representation of the model, and facilitate the
single worker. Hence, while having two workers would integration of the results into an interactive decision sup-
theoretically cut the processing time in half, some parts of port environment.
the process do not benefit from having an additional

123
Res Eng Design

4.3.3 Design of experiments and surrogate modeling In addition to the advanced composite wingbox design,
two additional parametric production and scheduling
In its current state, the model requires approximately two models are developed in Simio (one for a machined alu-
minutes to run a single, year-long production simulation, minum design and one for a conventional composite
which is too long to be an effective early design tool design). A summary of the three concepts modeled can be
(Mavris et al. 2010). Because the middle of the design found in Table 4. These three models expand the capabil-
space is of interest and there is little previous knowledge ities to trade between production concepts, materials, and
about the space, a Latin Hypercube Sampling DoE is processes alongside factory settings and wingbox design
chosen (Keane and Nair 2005; Myers et al. 2009). parameters. In particular, by considering both aluminum
ModelCenter by Phoenix Integration is used to link the and composites designs, trades between improved perfor-
structures model, manufacturing cost and time model, and mance of a composite wingbox against the possible
production flow model to automatically run the DoE cases. decreased cost and improved producibility of the aluminum
The model for an advanced composite design and process design are now available.
has 19 inputs variables (5 wing and vehicle design and 14 The production inputs and outputs for the three design
discrete factory design variables) as shown in Table 2. For concepts considered are summarized in Table 5. These
this experiment, 3500 cases were run to capture the com- inputs and outputs are for the Production Flow Model
plex nature of the design space. Indeed, the discrete inputs block in Fig. 2 only; additional responses related to
have complex interactions that are much more difficult to manufacturing cost, backlog, and aircraft performance are
model than continuous variables. calculated within the visualization environment using
As seen in Table 2, the DoE explores both wing outer previously generated surrogate models (Ceisel et al.
mold line (OML) geometry and production facility vari- 2012).
ables within defined ranges. Surrogate models are gener-
ated using Artificial Neural Networks (ANNs) (Myers et al. 4.4 Development of a parametric cash flow
2009). Equations are generated for the responses shown in formulation
Table 3. Upon completion of the DoE runs and the creation
of the surrogate models, or Neural Networks, the compu- A parametric cash flow formulation, in the form of cost
tation time of the entire model is decreased from approx- equations for equipment acquisition, labor rates, material
imately two minutes to less than a second. costs, loan payments, and taxes, is also implemented and
The surrogate models generated are then integrated with integrated into the MInDPRO framework. The cost esti-
JMP visual analytics capabilities to provide the designer mating relationships are developed from a combination of
with an interactive visual environment for rapid trade publicly available cost data (such as aerospace technician
studies between performance and production. rates) and benefits information, surveys of equipment and
material providers, and process-based costing models
developed within SEER-MFG. With this information, an
estimate of total production cost per wingbox produced for
Table 2 Design of experiments input variables for advanced com- a selected wing geometry and factory layout is available.
posite model The production cost includes labor, material, equipment,
Input variables Min. Max. facility, recurring (e.g., energy, management overhead,
maintenance, etc. costs) and tooling costs. The visualiza-
Wing area ðft2 Þ 240 410
tion environment (described in Sect. 5) combines these
Aspect ratio 3.84 6.24
calculations with user defined market assumptions (such as
Range (nmi) 220 1040
interest rate, tax rate, facility depreciation rate, and salvage
Taper ratio 0.36 0.66
value) and the production rate information to produce a
Sweep angle () 28 44
cumulative cash flow. The user may get a sense of the
Tool sets (rib, spar, and skin) 2 5 proposed program’s robustness to a range of market con-
Shifts (rib, spar, and skin) 2 3 ditions by varying the economic assumptions and quickly
Workers/station (rib, spar, and skin) 1 3 visualizing how the cash flow and other economic factors
Workstations (rib and spar) 1 3 are affected. The reader is invited to consult (Heckwolf
Skin/stringer workstations 1 4 2014) for a full description of how the cost estimations
Rib NDI stations 1 5 were developed.
Spar and skin NDI workstations 1 4 The integration of the capabilities described throughout
this section enables the simultaneous evaluation of vehicle
performance, production rate, and profitability. These three

123
Res Eng Design

Table 3 Design of experiments responses and justification


Responses Justification

Flow time Provides meaningful comparisons between components’ fabrication times


When integrated into a Gantt chart helps visualize critical path in the production flow
Process time When used in conjunction with flow time, provides insight to waiting times, queues, and efficiencies throughout the
factory
Workstation utilization Enables user to quickly see constraining production steps
rates
Yearly factory throughput Major response to be traded with performance and cost
Worker utilization rates Enables user to predict the number of workers required to support the factory layup stations

Table 4 Summary of manufacturing concepts modeled


Aluminum design Conventional composite design Advanced composite design
Automated machining Advanced process Advanced process

Facilities and CNC machining Automated fiber placement Hand lay-up


tools Autoclave cure Vacuum-Assisted Resin Transfer Molding
Integration Very high integration Medium integration High integration
Structural Industry practices and first order Industry practices and first order Industry practices and structural Optimization
data structural analyses structural analyses using Hypersizer
Joining Fasteners (assembly only) Co-bonding Co-bonding/Paste-bonding
Fasteners Fasteners (assembly only)

Table 5 Production model inputs and outputs


Inputs for aluminum concept Inputs for conv. composite concept Inputs for adv. composite concept Model outputs

7 wing geometry variables 7 wing geometry variables 5 wing geometry variables Throughput
Stretch forming station capacities AFPa station capacities HLUb station capacities Workstation utilization rates
Machining capacities NDIc station capacities NDIc station capacities Component/batch flow times
Number of shifts Number of shifts Number of shifts
Number of workers Number of tools available Number of workers
Number of tools available
a
Automated fiber placement
b
Hand lay-up
c
Non-destructive inspection

categories of information are parametrically linked to 4.5 Development of multi-objective optimization


inputs relating to the vehicle design, production facility capabilities
layout over time, and economic assumptions. As such, this
digital thread enables information to be brought forward Much research has been devoted to developing optimiza-
into the conceptual design phase. This represents a critical tion frameworks to bring forward multiple disciplines
capability, especially when investigating concepts that (structures, weights, manufacturing costs, and life-cycle
depart from more conventional (e.g., aluminum) designs. costs) forward in the design process and then using the
However, the manual exploration of this high dimensional resulting models to perform multi-disciplinary optimiza-
design space is not practical. To address these limitations, tion (Rais-Rohani and Dean 1996). A methodology is
multi-objective optimization capabilities are developed that proposed to optimize both composite and aluminum wing
support the effective exploration of the multi-disciplinary skin/stringer components by minimizing a combination of
design space. aircraft performance and manufacturing cost (Kaufmann
et al. 2010). Similarly, a study for an aluminum concept

123
Res Eng Design

was conducted by integrating performance, operating costs, Table 6 Summary of optimization problem design variables and
and structural weight relations as objective functions (Zhao constraints
et al. 2014). One critical aspect missing from these opti- Design variables
mization formulations is the impact of customer demand Wing geometry Planform area, aspect ratio
and production over time. Optimizing the manufacturing Taper ratio, sweep angle
cost for a single component may not be the best solution Design range, rib angle, rib spacing
when a design with higher production rate agility may lead Factory configuration Workstations, workers
to higher profitability over the lifetime of production Tools, shifts
despite a higher per-unit cost. Constraints
Side constraints Surrogate modeling definitions
4.5.1 Optimization problem description Sizing Max. fuel weight, empty weight
Takeoff gross weight, thrust loading
The simultaneous wing geometry and production planning Performance Takeoff field length, landing field length
optimization problem contains both continuous and dis-
Turn rate, stall speed
crete variables and many constraints. The goal is to mini-
Rate of climb, endurance
mize (or maximize) some metric(s) of value, such as cost,
Production Max. backlog, inventory
performance, or backlog health. The parameters being
Number of factory changes allowed
varied are high-level wing geometry variables and factory
configuration variables for each year of production. Mul-
tiple types of constraints need to be considered: side, air-
craft sizing, performance, and production constraints. The
side constraints are dictated by the surrogate models for where f ðX; YÞ is the objective function(s) of interest
performance and production. The models are created using (e.g., minimize cost, maximize aircraft performance, min-
bounds on the input variables, so allowing variables to have imize backlog, etc.). X are the wing design variables
values outside those bounds would result in extrapolation described in Table 6, while Y represents the factory con-
beyond the domain of verified prediction accuracy. Sizing figuration (defined in Table 6) for each year of production.
constraints dictate the maximum weights allowed for the gperformance represents the performance functions of the
entire aircraft. Performance constraints enforce various aircraft specified in Table 6 and Rperformance are the per-
point performance characteristics for the aircraft, such as formance requirements that would be specified by the
landing field length and rate of climb, set either externally customer or from market analysis. gbacklog is the backlog for
by the customer or derived by the designer. Lastly, the each i year of production to be kept under a maximum
production constraints enforce certain business philoso- allowable backlog. ginventory represents potential excess
phies through the way that the factory is operated and production beyond the aircraft orders to be kept under a
backlog managed. maximum inventory level. If orders out-pace production
A summary of the optimization problem’s inputs and (as is likely the case), then gbacklog is possibly active;
constraints is presented in Table 6, with the formal opti- otherwise, it is ginventory that can potentially become active.
mization problem statement being defined as follows:  
Pnyears 1  Yiþ1;j Yi;j 

Minimize: f ðX; YÞ
The i¼1 maxðjYiþ1;j Yi;j j;1Þ  changesmax ; j constraint
Subject to: counts the number of year-to-year changes to the produc-
gperformance ðXÞ  Rperformance tion system (e.g., workstation additions, new hires, etc.).
This constraint is put in place to ensure that changes to the
gbacklog ðX; Yi Þ  backlogmax ;
factory are not made every year. While this would enable
i ¼ 1;    ; nyears the system to better match the demand, this scenario is
ginventory ðX; Yi Þ  inventorymax ; unrealistic. Furthermore, by setting changesmax;j to a low
i ¼ 1;    ; nyears ð1Þ number, the optimizer will search for a factory that can
  handle the entire range of demand with little to no infras-
nyears
X  1 Yiþ1;j  Yi;j 
     changesmax;j ; tructure changes. Finally, X and Y must remain within the
   
i¼1 max Yiþ1;j  Yi;j ; 1 side constraints defined by the surrogate models to main-
j ¼ 1;    ; nFactoryVariables tain applicability.
XL  X  XU With the optimization problem formulated, the algo-
rithm selection process is described in the following
YL  Yi  YU ; i ¼ 1;    ; nyears
section.

123
Res Eng Design

4.5.2 Optimization algorithm selection minimizing material cost, rather than total cost, could
provide a more robust solution in this scenario. Similarly,
Many factors need to be considered when selecting a other categories that could be minimized include labor,
suitable optimization algorithm for this problem. The equipment, or recurring costs.
design variables determining factory configuration are Backlog health is another metric important to the aero-
discrete, which eliminates most gradient-based algorithms. space industry. Indeed, for many major aerospace compa-
Some of the constraints, such as backlog, are highly non- nies today, one metric closely watched by analysts in the
linear functions. Also, due to the hundreds of independent financial sector is the value of orders tied up in backlog and
variables associated with lifetime factory configuration, the the number of years of production that backlog represents
design space most likely has many local optima. To begin (Captain 2015). The choice on how to manage backlog in
to approach the global optimum, an algorithm that has the terms of what is acceptable or desired is a business pref-
ability to sample across the entire design space despite the erence, not a deterministic rule. As such, two objective
presence of local optima is essential. Finally, the objective functions representing different philosophies were added:
and constraint functions are defined via surrogate models, maintaining a desired level of backlog determined by the
so function calls are computationally inexpensive. user, and attempting to reduce the customer wait time
These problem characteristics make a metaheuristic between order and delivery to as much as possible.
search algorithm that effectively handles discrete variables By allowing the user to investigate multiple objectives,
and a complex design space appropriate for the problem the optimization routine enables the designers to identify a
(Gen and Cheng 2000; Sarker et al. 2002; Zitzler et al. family of alternatives that may then be compared based on
2000). Because of the extremely fast objective function the current market requirements. Following the presenta-
evaluations, a genetic algorithm (GA) is chosen over other tion of the visual tradespace environment in the ensuing
metaheuristic search strategies for this problem. The GA section, selected results from the optimization routine are
parameters used for the optimization studies presented in presented with a focus on illustrating the affordability
the Optimization Results section are summarized in trades such an environment enables.
Table 7.

4.5.3 Objective function selection 5 Tradespace environment

One of the most influential considerations in the selection A successful multi-disciplinary design exercise requires
of an optimum design has to do with the objective function that all parties involved, in this case the vehicle design,
used. However, the metric used as a basis to designate the production, and manufacturing engineers, along with the
‘‘best’’ design is highly subjective. Performance, manu- marketing and management personnel, have access to data
facturer economics, or customer economics could all be at an appropriate level of detail to make informed deci-
legitimate objective functions depending on the specific sions. A visual tradespace environment is thus developed
design philosophy and current circumstances. To accom- within JMP from the SAS Institute to effectively commu-
modate subjectivity, multiple objective functions need to nicate and assess the impact of design decisions. In par-
be defined. ticular, this environment supports the needs discussed in
A very common objective function in design problems, Sect. 2.
especially for commercial applications, is maximizing The following subsections describe the capabilities of
profit on each unit sold. If the selling price of the product is the environment with a focus on the visualizations created
fixed, maximizing profit is equivalent to minimizing total to investigate the impact of demand variability on the
expenses. However, it may be favorable to optimize a aircraft’s design, production plan, and program profitability
design in terms of a subcategory of cost rather than total in support of affordability-based trades.
cost. For example, material costs may be a major contrib-
utor to the total cost of a design that cannot be driven down 5.1 Environment capabilities
by learning curve effects or efficiency gains. Therefore,
5.1.1 User inputs

Table 7 Genetic algorithm The user inputs, as presented in Tables 2 and 5, are cate-
Population size 1200
(GA) parameters
Generations 500 gorized in terms of aircraft wing design and aircraft pro-
Crossover fraction 80% duction. These inputs drive the visuals illustrating the
Mutation rate 1% vehicle’s performance (discussed in Sect. 5.1.2) and single
year production performance (discussed in Sect. 5.1.3).

123
Res Eng Design

5.1.2 Aircraft performance The Gantt chart traditionally illustrates a project sched-
ule, along with its inner-dependencies (Herrmann 2006).
A traditional Conceptual level design tool for aircraft This visualization has been modified in the environment to
design is a constraint diagram (Fig. 4). This diagram dis- show the flow time of the sub-processes involved in the
plays mission specific constraints such as takeoff field creation of a complete wingbox (Fig. 5). This modified
length, climb rate, etc. as the colored lines. Feasible regions Gantt chart, thus, helps the designer quickly identify areas
are shown as white space, while the shaded regions indicate of the production floor that need improvement to increase
space that is not feasible due to one or more constraints. throughput (large waiting times on the critical path) (Józe-
The aircraft’s thrust loading (thrust at sea level static fowska and Weglarz 2006). The built-in dependency feature
conditions divided by takeoff gross weight (TOGW)) is of the Gantt chart is used to identify the critical path (golden
represented on the vertical axis, while the wing loading lines) components of the production process. Finally, the
(takeoff gross weight (TOGW) divided by wing area (S)) is Gantt chart gives, for each task, the breakdown of the flow
represented on the horizontal axis. These axes are used time in terms of processing time (green lines under the main
because they are representative of the three main outputs of black or golden) and waiting time (red lines).
Conceptual design, namely aircraft weight, aerodynamics, The process utilization diagram is developed to provide
and propulsion (Raymer 2006; Mattingly et al. 2002). After insight into the efficiencies of the different processes.
overlaying all mission constraints, a design point (green Hence, each box of this utilization chart (Fig. 5) contains a
marker) can be chosen. A typical strategy is to choose a bar chart showing processing (green), idle (red), shut down
design point that lies near the minimum feasible thrust (orange), and off shift (blue) percentages for critical
loading, as this would provide the smallest engine capable components. Furthermore, the critical path is highlighted in
of completing the mission (Mattingly et al. 2002). By dark gold. This diagram, when used in concert with the
being able to parametrically vary high-level wing design Gantt chart, can help identify portions of the production
inputs, the user is provided with the means to instanta- flow that are constraining production.
neously visualize the impact(s) that such changes have on Additional detailed visuals showing how production rate
the constraint diagram (difference between full lines and information is integrated with the vehicle performance
dashed lines on Fig. 4). design are described in the following section.

5.1.3 Production analysis 5.1.4 Additional analysis capabilities

To support the concurrent evaluation of alternative wing Figure 6 presents a contour profiler, which overlays
geometries and factory settings, this environment brings defined constraints onto 2 design variables. The constraints
forward tools and techniques of the type used at the and design variables can vary parametrically based on user
Operational or Tactical level of production design (Wilson inputs, opening or closing the feasible design space (white
2003; Herrmann 2006). The chosen tools, such as a process area) based on their values. Hence, this visualization allows
utilization diagram and Gantt chart, are discussed below. the designer to observe the active constraints and identify
the ones that prevent him or her the most from obtaining
the largest feasible space possible and, consequently, from
gaining the full benefits of the design concept (Mavris et al.
2010). One of the unique and critical features of the con-
tour profiler developed in the context of this research is its
ability to superimpose both performance and production
related constraints/requirements. Doing so allows the
designer to fully capture the impact of production con-
straints on the choice of a design concept.
Hence, by bringing production considerations forward in
the design process, one can ensure that the design concept
selected not only meets the traditional performance require-
ments but also satisfies any production requirements, such as
monthly production rate (represented by the red lines in
Fig. 6), set forth by the aircraft manufacturer. This essential
capability is enabled by the development and use of (para-
metric) surrogate models of the cost and production models
Fig. 4 Parametric aircraft constraint plot and their integration within the existing environment.

123
Res Eng Design

Fig. 5 Parametric Gantt chart and utilization chart

Fig. 6 Contour profiler

5.1.5 Demand variability developed (Fig. 7). The demand profile allows the user to
define a projected order history (blue shaded region)
One of the key capabilities of interest to designers is an derived from market research. The environment has order
assessment of the impact of production layout and wing histories from many major aircraft (Fig. 7 presents the 747
design on profitability over the life of the vehicle. To order history as an example) that may be selected to pop-
support this need, an interactive demand profile is ulate the demand profile. Additionally, the user may create

123
Res Eng Design

a custom demand scenario by dragging points within the


interactive plot to set the projected year to year demand.
Overlayed onto the demand profile is the current proposed
yearly production plan (red dashed line). The production
plan is determined by first choosing a wing planform
design (wing area, aspect ratio, taper ratio, etc.) using the
inputs displayed in Tables 2 and 5, and then defining the
year to year factory layout. The user does this by selecting
a production year, modifying the factory layout (worksta-
tions, technicians, shifts, and tools for the subcomponent
Fig. 8 Cumulative order backlog for the demand and production
lines), and applying the changes. The environment then
defined in Fig. 7
takes these inputs and calculates the yearly production rate
with the surrogate models developed using data from the production of a designed wingbox and maintain a healthy
Simio models. This yearly production rate is then used for order backlog.
all subsequent years (the flat portions of the red dashed line
in Fig. 7) until a new factory configuration is defined by the 5.1.6 Production use case
user in an ensuing year (seen as the jumps in the production
plan line). It is assumed that any rate changes happen For this use case, the demand history is shown in Fig. 7,
somewhat gradually; hence, the first year after a rate and a requirement that the backlog must remain under 300
change produces OriginalRate þ wingboxes is enforced. It can be seen that the baseline
0:8ðNewRate  OriginalRateÞ wingboxes. Once all orders production rate for the wing geometry selected (one that
have been filled, the factory is shut down and the produc- meets customer defined performance goals (Fig. 4)) is
tion rate goes to zero (as in year 51 in Fig. 7). sufficient (red dashed line in Fig. 7 from years 1 to 13) to
A plot of the cumulative order backlog in terms of remain under the backlog constraint until year 14. At this
number of wingboxes is also provided, as shown in Fig. 8. point, the production rate must be increased to keep the
The user can easily see when the backlog exceeds an backlog below 300 orders. The provided Gantt and uti-
allowable threshold during the production run and may use lization charts (Fig. 5) can be used to increase production
this information to help decide when to make changes to by suggesting meaningful changes to the production plan.
the factory layout. These visuals, along with the production In particular, the Gantt chart shows that there is a long time
and performance plots described in Sects. 5.1.2 and 5.1.3, delay at the rib non-destructive inspection (NDI) step,
enable interactive, concurrent aircraft geometry and pro- while the utilization chart shows that the rib NDI station is
duction plan design. The next section demonstrates how the almost fully utilized. This indicates that the NDI station is
designer would use the environment to investigate the not able to meet the demands of the rest of the factory and
is the constraining process. Therefore, an additional station
is added in year 14 to increase the production rate, as seen
by the step in production at year 14 in Fig. 7. Doing so
helps ensure that the backlog remains under 300 orders.
Further details about how to effectively increase yearly
throughput using the visualization environment can be
found in (Siedlak et al. 2014). A similar procedure is fol-
lowed throughout the rest of the production run to maintain
an acceptable backlog as it approaches the constraint.
Figures 7 and 8 illustrate the overall production plan
defined through this exercise. This use case demonstrates
again the importance of developing surrogate models of
detailed analyses in order to be able to investigate the
producibility of a concept early in the design process.
With the capability to understand the orders and pro-
duction rates of the aircraft over time, economic analyses
Fig. 7 Demand profile populated with the 747 order history and a must be integrated to perform cash-flow and profitability
proposed production plan assessments.

123
Res Eng Design

5.1.7 Economic analysis Table 8 Factory layout change summary


Year Non-recurring D yearly recurring Cost/wing
The previous section focused on adapting the production investment cost produced
rate by making changes to the factory layout to maintain a
specified backlog. However, because of the high number of 1 $198,872,000 N/A $430,000
factory configurations leading to similar production rates, 14 $3,000,000 $10,810,000 $386,000
the cost of each configuration must be jointly considered to 22 $3,000,000 $10,457,000 $358,000
determine the best configuration. Therefore, various eco- 27 $0 -$14,017,000 $365,000
nomic statistics and visuals are also provided to allow the 28 $0 -$7,250,000 $430,000
decision makers to instantaneously view the economic
impact of varying a decision variable (wing area, rib NDI
workstations, etc.) alongside the vehicle performance and
production rates.
Using surrogate models developed in (Heckwolf 2014)
for the same advanced composite wingbox and user defined
economic assumptions (e.g., aircraft selling price, buyer
down payment percentage, loan interest rate, and depreci-
ation information), a cumulative cash flow diagram (Fig. 9)
is generated. Using the demand and production rates from
Fig. 7, the cash flow diagram illustrates that the program is
mostly profitable during its mature phase of higher pro-
duction from year 14 to 28. This is partially because more
volume simply means more cash inflows, but the factory is
also more efficient when producing at higher volumes. As
seen in Table 8, which summarizes the economic impact of
each change to the factory configuration during the pro- Fig. 10 Breakeven year and return on investment versus selling price
per wing
duction run, the cost per wing produced during the periods
of higher production is less than the baseline factory even
profitable and is a good measure of risk and of the time
though the total yearly recurring cost is higher than the
before other investments may be made. ROI is a key per-
baseline factory. This is due to the allocation of fixed and
formance indicator used to evaluate a program or division’s
overhead costs over a larger quantity of products during the
profitability (Braun et al. 2010), and it may be compared
periods of higher production and more efficient utilization
against the company’s Minimum Acceptable Rate of
of some workstations throughout the flow.
Return (MARR) to determine whether or not to continue
The break-even year and return on investment (ROI) for
the program (Blank and Tarquin 2012). The marketing
the chosen wing geometry and production plan against the
team will have a price range at which they believe they can
selling price per wing are provided in Fig. 10. The break-
sell the vehicle or component, allowing for determinations
even year describes the time before the program will be
to be made quickly as to the ability of the proposed design
to meet the company’s MARR. Therefore, by providing
access to economic, performance, and production infor-
mation concurrently, the integrated environment allows for
details across disciplines to be brought forward and sup-
ports collaborative, multi-disciplinary design.
A summary of the performance, production, and eco-
nomic responses available through the visualization envi-
ronment is presented in Table 9. The capability to
immediately propagate changes to the vehicle design,
factory layout, demand profile, and economic assumptions
through each of these responses helps to bridge the gaps
identified in current implementations of affordability- and
resiliency-based design methodologies. The results from
the optimization routine, described in the next section,
further illustrate the interactions between wing geometry,
Fig. 9 Cumulative cash flow diagram

123
Res Eng Design

Table 9 Summary of responses


Performance Production Economics
included in the visualization
environment Fuel weight Production rate Capital cost
Gross weight Workstation utilization rate Tooling and machine cost
Take-off field length Component flow time Material cost
Landing field length Component waiting time Labor cost
Turn rate Component processing time Overhead cost
Stall speed Technician utilization Return on investment
Rate of climb Yearly Backlog Cumulative cash flow
Range/endurance Excess Inventory Breakeven year

production, and demand variability and, ultimately, their


impact on the program’s profitability.

6 Optimization results

The previous section described the manual tradespace


exploration that one can perform to identify and quantify
trends during design reviews; however, because of the
sheer size and high dimensionality of the design space,
manual exploration is not practical. Optimization tech-
niques are implemented to address this issue. This section
presents results from a variety of optimization studies using
the developed environment. Section 6.1 investigates how
the optimum wing geometry changes for different cost
objective functions using a simple trapezoidal demand
profile. This helps designers identify design features that
impact different manufacturability and producibility met-
rics. Section 6.2 presents results from a screening test to
Fig. 11 Optimum composite wing geometry for different cost
determine which performance and backlog constraints objective functions
significantly impact the overall production cost and wing
area. The screening test, therefore, helps the design team As expected, the optimal wing geometry changes based
understand the performance and producibility requirements on the objective function used. Figure 11 shows the wing
that impact the design. Results from a multi-objective geometry from three different cost roll-up objective func-
optimization for an aircraft performance metric (Stall tions overlaid on the original F-86 Sabre geometry. The
Speed) and total material cost are presented in Sect. 6.3 to wing area when minimizing material cost is the smallest of
demonstrate a newly available affordability trade. Finally, the three geometries at 240 ft2, while the labor and total
a comparison between this work’s methodology and a more production costs have wing areas of 259 ft2 and 296 ft2,
traditional, weight-based methods is discussed in Sect. 6.4. respectively. A smaller area means smaller components
and less material. The minimum labor cost design has a
6.1 Optimum wing geometry for various cost smaller aspect ratio than the other costs, 4.42 versus 5.45
objective functions for the material cost design. Decreasing the aspect ratio
reduces the number of ribs; through past experience with
The metric chosen for the objective function can have a the advanced composite model, the rib production line and
large impact on the optimum design, especially when its corresponding NDI stations are often on the critical
metrics are inversely correlated. To investigate this effect, path. Alleviating this bottleneck through a smaller aspect
the wing geometry and factory settings are optimized for ratio wing makes the overall production flow more efficient
the unconstrained case over a trapezoidal demand profile to with less idle workers, which in turn translates into the
represent a simple ramp-up, ramp-down order scenario. lowest cost devoted to labor. This exercise demonstrates
Some representative results are shown in Fig. 11. how wing design impacts different cost categories, which is
an important capability to support decision making.

123
Res Eng Design

6.2 Optimization constraint screening tests Table 11 Influential constraints for aluminum wing area optimized
for minimum total production cost
With the optimization algorithm and objective functions in Constraint Relative t-ratio plot p value
place, the design space can be studied by determining the
performance and production constraints that impact the Backlog < 0.0001
choice of wing geometry. One way to quantify sensitivities LFL < 0.0001
is through the use of a factor screening experiment. This TOGW < 0.0001
type of experiment is typically performed to identify fac- TOFL 0.0002
tors that have large effects on the response (Myers et al. Stall speed 0.0005
2009). In this application, the factors are the performance
and production constraints imposed on the design, whereas
the responses are the optimum design variable values for
the given constraint scenario. sensitivities on high-level cost parameters early in the
A major observation from the factor screening study is design process allows the designer to understand the cost
that the backlog requirement has major cost implications implications of certain requirement changes and react
and can even drive the values of certain wing design appropriately.
variables. For both aluminum and composite designs, the The sensitivity of optimal design parameters to the
optimal design for objective functions related to the various requirements can also be studied. This again helps to
breakdowns of cost (e.g., total production, material, labor, identify design characteristics that impact the performance
equipment, recurring, and break-even year) are most and producibility of a design. The sensitivities for the wing
impacted by the backlog threshold. As an example, the area of the aluminum concept optimized for total produc-
screening results for the total production cost associated tion cost are shown in Table 11. As expected, the results
with the composite designs are shown in Table 10. The show that performance constraints such as landing field
backlog-cost relationship is negatively correlated; length, takeoff field length, and stall speed are major dri-
increasing the backlog requirement to a higher, less strin- vers of the resulting wing area. More importantly, the
gent value tends to lower the total production cost. A lower results also show that backlog is the main driver of wing
tolerance to large backlogs means production rate changes area. Wing area is a major driver for the dimensions of
will be more common, which implies higher costs due to each component within the wingbox structure; scaling up
more equipment acquisitions and personnel changes. This the size of the wing increases the necessary thickness of
demonstrates the importance of considering production each component, the lengths of the spars, ribs, skins, and
rates for environments subject to demand variability when stringers, and potentially increases the number of stringers.
one is concerned with the design of affordable concepts. Having a very low backlog tolerance requires the factory to
The original performance requirements set for the have the ability to shift to higher production rates very
design can impact production costs, as illustrated in quickly; a smaller wing would ease the transition to those
Table 10. After backlog, the next most influential con- higher production rates because the components will be
straints on total production cost are landing field length, smaller (and, therefore, have a shorter manufacturing time)
stall speed, and takeoff field length, respectively. These than for larger wings. While trends similar to those iden-
constraints are direct drivers of wing area: increasing the tified in this section are qualitatively understood, the pre-
value of these constraints allows for a smaller wing. A sent methodology provides the unique ability to quantify
smaller wing means fewer raw materials and less manu- their impact, which in turn helps trace and facilitate dis-
facturing time necessary to build the wing, decreasing the cussions and compromises among the various stakeholders
total production cost. Knowing the performance involved in the design.

Table 10 Influential constraints for composite optimum total pro- 6.3 Multi-objective optimization results
duction cost
Constraint Relative t-ratio plot p value
When multiple criteria are critical to the performance and
desirability of a design, especially for competing objec-
Backlog < 0.0001 tives, multi-objective optimization provides insight into the
LFL < 0.0001 best ‘‘compromise’’ designs. Multi-objective capabilities
Stall speed < 0.0001 are included in this environment using a multi-objective
TOFL 0.0128
genetic algorithm, the Non-dominated Sorting Genetic
Algorithm-II (NSGA-II) (Deb et al. 2002). The algorithm

123
Res Eng Design

was modified to accept any two objective functions relating collaborate during the design process and better understand
to production costs, backlog management, and aircraft how their choices, assumptions, and constraints impact the
performance. An example of this capability is shown in performance and cost of the design when compared to
Fig. 12. This specific example shows how stall speed, a traditional design techniques. Such capabilities are identi-
performance requirement that directly influences takeoff fied throughout Sect. 2 as essential to reducing the design’s
and landing capabilities, can be traded against total mate- cost and improving profitability. This section aims to
rial cost for the aluminum production concept. Decreasing qualitatively and quantitatively compare how well the
the stall speed requires a larger wing area and increases the presented methodology achieves these goals when com-
raw materials required for manufacturing. Pareto frontiers, pared to a baseline, traditional design process.
such as the one illustrated in Fig. 12, display the frontier of For the purposes of this comparison, ‘‘traditional design
non-dominated solutions. As such, these are powerful techniques’’ entail the use of aircraft empty weight as a
design tools when conducting compromised tradeoffs surrogate for cost (e.g., the design with the minimum
(Sarker et al. 2002). Additionally, the Pareto frontiers weight that can meet the performance requirements will
display the best value a metric can achieve when the other have the lowest production cost). The capabilities and
metric is held constant. For example, if the aluminum methods presented in this paper can be compared to the
design has a stall speed of 70 ft/s, the absolute best mate- traditional method through a set of trades and experiments.
rial cost that can be achieved for the given stall speed is Many trades and capabilities demonstrated throughout
$44 million. If the stall speed requirement is relaxed to Sects. 5 and 6 support the utility of the presented
100 ft/s, the best material cost is reduced to $41.2 million. methodology over traditional design. First, the contour
Multi-objective optimization can provide insight on the profiler presented in Fig. 6 demonstrates that the produc-
impact of changing requirements on the best design pos- tion rate required to maintain customer satisfaction and
sible with respect to cost, production, or performance make a profit can have a significant impact on the feasible
metrics, which is the primary goal of affordability-based wing design. Hence, without this information, the designer
design and acquisition. would have a limited ability to design a vehicle that could
The following section ‘‘closes the loop’’ by demon- be produced at the required rate without significant
strating how the described analyses come together to ulti- increases in factory cost.
mately generate a more affordable design than when using The collaborative analysis among the performance,
traditional methods. production planning, and the economic analysis teams
discussed throughout Sect. 5 again supports the develop-
6.4 Closing the loop—comparison to traditional ment of an affordable design. Discussions that link a
techniques requirements change (e.g., 5 ft/s decrease in stall speed) to
a design change (e.g., increase in wing area) ultimately to
The primary argument of this paper is that the capabilities the change in production rate, cost, backlog, and cash flow
presented can help designers and decision-makers better can now be supported by instantaneous, quantitative anal-
ysis. Further, as shown in Sect. 6.2, requirements from both
performance and production that drive the cost can be
110 concurrently analyzed. In a real-world implementation, this
insight could prove invaluable when setting requirements
on the design. Finally, the multi-objective optimization
100
capabilities described in Sect. 6.3 can directly inform the
Stall Speed (ft/s)

design requirements by demonstrating the trades between


90 cost and performance. In a full implementation, this could
be expanded to many more dimensions across disciplines
to help better understand the design space across previ-
80 ously hidden or unquantified dimensions.
The following sub-section builds on these observations
70 to quantify the changes to the design that are exposed by
implementing the described methodology.
$40 $42 $44 $46 $48 $50 $52
Total Material Cost ($M, 2014 USD)

Fig. 12 Requirement versus cost Pareto frontier for an aluminum


concept

123
Res Eng Design

6.4.1 Quantitative comparison of the traditional assumptions, the material cost is halved for the second
and described design techniques scenario presented.
While these results do not show a significant variation
This section aims to briefly compare designs produced between the designs achieved, there are some noticeable
using the proposed methodology against designs produced trends between the scenarios. For this particular set of
using a traditional, weight-based technique. To enable this constraints and assumptions, the labor cost per wing and
comparison, a baseline wing design is generated by mini- capital cost per wing can be decreased by increasing the
mizing the wing weight subject to the set of performance wing area and aspect ratio while decreasing the taper ratio.
constraints described in Table 6. This represents the ‘‘tra- As the material cost decreases, the design is allowed to
ditional,’’ empty weight-based design process. The tradi- change further to produce improvements to the labor and
tional optimization produces the wingbox described in capital costs per wing. This analysis, thus, demonstrates the
Table 12. potential to reduce costs by leveraging the presented
The first quantitative comparison is to assess whether analysis methodology while further exhibiting the benefits
the wing designed by minimizing the material cost differs to the design process achieved through cross-discipline
from that obtained by minimizing the wing weight. Upon collaboration. The following section discusses the current
completing this optimization case, the two designs are limitations of the method.
equivalent, which indicates that even though the process-
based costing module incorporates additional material 6.5 Limitation of results
costs related to consumable and trimmed materials, this
does not significantly impact the wing’s design given the While the development of the analyses and assumptions
available information and assumptions. made have been discussed with subject matter experts, the
One of the main goals of the research discussed in this manufacturing cost model and production simulation were
paper is to design a wing that costs less to produce than a not validated against historical data (as it is not available).
traditionally designed wing. This can be examined by Furthermore, this work does not claim to provide improved
minimizing the cost per wing for both scenarios. For the cost estimates when compared to a validated, relevant cost
traditional case, the wing is fixed to the design produced by estimating relationship. As such, the trends and design
minimizing wing weight, and the factory settings are then recommendations made throughout the use case should not
optimized to minimize the cost per wing. This represents a be utilized to guide the design of an aircraft wing. Rather,
best case scenario for the minimum weight design in that the visualizations and results presented throughout Sects. 5
the factory is optimized to minimize cost. This is intended and 6 are intended to demonstrate how the applied
to isolate the impact of freeing the wing design alongside methodology can be utilized to conduct a variety of
the factory variables. For the presented methodology, the affordability-based trade studies. Further, the use case is
wing design is allowed to vary alongside the factory designed to illustrate that inclusion of detailed information
settings. through the integration of surrogate models provides new
For this example, the cost per wing is calculated by insight to the early design of an unconventional concept.
dividing the total yearly material, labor, and capital costs Finally, the optimization formulation as presented rep-
by the number of wings produced over that year. The resents an initial attempt at the full, multi-disciplinary
results are highly dependent on the input assumptions. For optimization ultimately desired. The current formulation,
example, if the labor rate is low but material costs are high, while producing results useful for the sensitivity studies
the minimum wing weight design is likely similar to the presented, is not mature enough to definitively demonstrate
optimal design. This highlights the need for collaboration that the total production costs can be consistently reduced
between stakeholders as enabled by the MInD over the production life-cycle. The primary limitations are:
methodology. (1) the present study encompassed only the wing design as
The results from this experiment are presented in opposed to the entire aerostructure, (2) the performance
Table 13. To demonstrate the impact of the cost constraints considered could be expanded to further restrict
the allowed design, and (3) by optimizing the entire pro-
Table 12 Wing design duction life-cycle, the problem is highly constrained and
Design variable Value limits the ability of MATLAB’s off-the-shelf genetic
produced by minimizing wing
weight Wing area 308.5 ft2 algorithm to maintain a large feasible population.
Aspect ratio 5.88 The following section summarizes the contributions of
Sweep angle 28.0 this work to the affordability- and resiliency-based acqui-
Taper ratio 0.55 sition paradigm and discusses opportunities for future
research.

123
Res Eng Design

Table 13 Comparison of
Minimum weight wing Freely optimized wing
production costs and design for
minimum weight and co- Baseline material cost
optimized wingbox
Cost/wing $470,800 $469,400
Labor cost/wing $164,300 $162,300
Material cost/wing $229,300 $230,900
Capital cost/wing $77,200 $76,200
Wing area 308.5 ft2 312.7 ft2
Aspect ratio 5.88 6.20
Sweep angle 28.0 28.0
Taper ratio 0.55 0.40
Halved material cost
Cost/wing $356,100 $353,700
Labor cost/wing $164,300 $162,000
Material cost/wing $114,600 $115,700
Capital cost/wng $77,200 $76,100
Wing area 308.5 ft2 313.9 ft2
Aspect ratio 5.88 6.25
Sweep angle 28.0 28.0
Taper ratio 0.55 0.36

7 Conclusion and future research 3. Developing and integrating surrogate models to enable
the rapid parametric propagation of design changes
While implementations of PLM, IPPD, and the LAI during throughout multiple disciplines.
the 1990s and early 2000s provided a major step towards 4. Demonstrating a multi-objective optimization routine
affordability-based design, the struggles of more recent that is capable of exploring the design space. The
programs illustrate that there is room for improvement. optimization routine is especially helpful in under-
Furthermore, recent changes to the aircraft manufacturing standing how requirements from multiple disciplines
industry, including the transition to the usage of more impact the design and identifying geometry choices
advanced composites along with the increase in aircraft that impact different cost categories.
demand variability, have exacerbated the challenges faced 5. The developed decision-support system, by leveraging
by designers and manufacturers. More modern design surrogate models of detailed analyses early in the
concepts, including model-based systems engineering and design process, demonstrates the potential for collab-
engineered resilient systems, seek to reverse this trend by oration among stakeholders. In particular, by instanta-
utilizing more detailed modeling techniques (e.g., physics- neously and interactively projecting the impact of
based performance models, process-based costing, dis- various decisions on performance, cost, production,
crete-event simulation, etc.) across disciplines during early and cash flow/profitability, trades that are usually
design. hidden during early design are now apparent. Addi-
The current work supports these methodologies by tionally, through the incorporation of an optimization
addressing the needs identified in Sect. 2. Specifically, the routine, exploration of the design space and identifi-
following capabilities are demonstrated: cation of influential requirements and constraints is
possible.
1. Presenting a digital thread approach to link aircraft
performance, wingbox structural design, process-based While affordability is primarily related to defense
costing, and production simulation tools. acquisition, the ability to balance performance and cost
2. Demonstrating an affordability-based analysis for an during design is imperative for commercial applications as
unconventional concept (e.g. a composite wingbox well. The general method presented in this work can easily
lacking relevant historical cost estimating relation- be adapted to support the design of a commercial aircraft.
ships) by trading the system’s performance with the This would require the development of performance,
manufacturability and producibility of the aircraft’s structural, process-based cost, and production models
wingbox. suitable for the commercial vehicle. The method could also
be utilized during the design of other subsystems and

123
Res Eng Design

components. For example, the affordability of an avionics Becz S, Pinto A, Zeidner L, Khire R, Reeve H, Banaszuk A (2010)
system is related to sensor capability and software com- Design system for managing complexity in aerospace systems.
In: Aviation Technology, Integration, and Operations (ATIO)
plexity (Becz et al. 2010; Szkody 1999). In this case, the Conferences. American Institute of Aeronautics and Astronautics
impact of interactions between the avionics and wing Blair M, Hartong A (2000) Multidisciplinary design tools for
design on the system’s performance and cost could be affordability. In: 41st AIAA/ASME/ASCE/AHS/ASC structures,
investigated. Ultimately, it is expected that the links and structural dynamics, and materials conference and exhibit.
American Institute of Aeronautics and Astronautics
flow of data between the models could be similar and the Blank L, Tarquin A (2012) Engineering economy. Mcgraw-Hill, New
surrogate modeling techniques involved would remain the York
same. Bobinis J, Garrison C, Haimowitz J, Klingberg J, Mitchell T, Tuttle P
Future work will focus on improving and expanding the (2013) 4.6.1 affordability considerations: cost effective capabil-
ity, vol 23, No. 1. In: INCOSE International symposium,
capabilities of the optimization routine to enable further pp 287–303
multi-objective trades, especially between the various Braun KW, Tietz WM, Harrison WT (2010) Managerial sccounting,
manufacturing concepts and additional performance 2nd edn. Pearson Hall, Upper Saddle River
requirements. Depending on the application, the inclusion Briceno S, Pinon O, Laughlin B, Mavris D (2013) Addressing
integration challenges in the design of complex aerospace
of additional ‘‘-ilities’’, such as maintainability and sup- systems. In: NAFEMS World Congress, Salzburg, Austria
portability, could have a large impact on the affordability Campbell F (2003) Manufacturing Processes for Advanced Compos-
of the system. In addition, the proposed framework could ites. Oxford
be extended to study and assess the impact on cost and Captain T (2009) Can we afford our own future? why a&d programs
are late and over-budget = and what can be done to fix the
performance of adding subsystems to the vehicle or problem. Online
designing families of vehicles to be produced in the same Captain T (2015) 2015 global aerospace and defense industry outlook.
production system (Libby et al. 2017). Finally, as aircraft Online
manufacturers increase their reliance on the supply chain, Ceisel J, Witte P, Carr T, Pogaru S, Mavris D (2012) A non-weight
based manufacturing influenced design methodology for prelim-
an analysis on the design’s impact on the performance and inary design. In: 28th International congress of the aeronautical
agility of the supply chain becomes necessary. Tang et al. sciences, Brisbane, Australia
has begun this investigation within ASDL’s manufactur- Cole C (2001) Guide for low cost design and manufacturing of
ing-influenced design research (Tang et al. 2014). composite general aviation aircraft. National Institute for
Aviation Research, Wichita State University. http://www.niar.
wichita.edu/agate/Documents/Advanced%20Manufacturing/WP3.
Acknowledgements This research is funded by Boeing Research and 1-031200-130.pdf. Accessed 20 Aug 2017
Technology (BR&T). In particular, the authors would like to thank Curran R, Kundu A, Raghunathan S, Eakin D (2001) Costing tools for
Howard Appelman, Steven Wanthal, Keith Rupel, Adam Graunke, decision making within integrated aerospace design. Concurr
and Gabriel Burnett from The Boeing Company for their expertise Eng 9(4):327–338
and support. Curran R, Gomis G, Castagne S, Butterfield J, Edgar T, Higgins C,
McKeever C (2007) Integrated digital design for manufacture for
reduced life cycle cost. Int J Prod Econ 109:27–40
References Das S, Kanchanapiboon A (2011) A multi-criteria model for
evaluating design for manufacturability. Int J Prod Res
(1960) Flight Manual: USAF Series F-86F Aircraft. United States Air 49(4):1197–1217
Force De Baets P, Mavris D (2000) Methodology for the parametric
(2015) F-35 joint strike fighter: assessment needed to address structural conceptual design of hypersonic vehicles. In: Pro-
affordability challenges. http://www.gao.gov/assets/670/ ceedings of the 2000 World Aviation Conference, San Diego,
669670.pdf. Accessed 2 June 2016 CA
(2015) Low cost attritable uas demonstration statement of objectives. Deb K, Pratap A, Agarwal S, Meyarivan T (2002) A fast and elitist
https://www.fbo.gov/index?s=opportunity&mode=form&id=14b multiobjective genetic algorithm: Nsga-ii. IEEE Trans Evol
63e83b3345e3bc8fdc5d02d288f62&tab=core&_cview=1. Acces- Comput 6(2):182–197
sed 2 June 2016 Defense Acquisition University (2012) Defense Acquisition
(2016) Air superiority 2030 flight plan. http://www.af.mil/Portals/1/ Guidebook
documents/airpower/Air%20Superiority%202030%20Flight/%20 Eaglesham MA (1998) A decision support system for advanced
Plan.pdf. Accessed 2 June 2016 composites manufacturing cost estimation. Ph.D. thesis, Citeseer
Anderson A, Longo C, Teufel P (2004) New composite design and Estefan JA et al (2007) Survey of model-based systems engineering
manufacturing methods for general aviation aircraft structures. (mbse) methodologies. Incose MBSE Focus Group 25(8):
In: 24th International Congress of the Aeronautical Sciences Gen M, Cheng R (2000) Genetic algorithms & engineering
Asiedu Y, Gu P (1998) Product life cycle cost analysis: state of the art optimization, Wiley series in engineering design and automation,
review. Int J Prod Res 36(4):883–908 vol 7. Wiley, New York
Bao H, Samareh J (2000) Affordable design: a methodology to Goerger SR, Madni AM, Eslinger OJ (2014) Engineered resilient
implement process-based manufacturing cost models into the systems: a dod perspective. Proc Comput Sci 28:865–872
traditional performance-focused multidisciplinary design opti- Haggerty AC (2004) Aircraft systems engineering: lifecycle consid-
mization. In: 8th symposium on multidisciplinary analysis and erations. Online
optimization, AIAA-2000-4839

123
Res Eng Design

Heckwolf A (2014) Integration of demand variability into aircraft SB10001424052702303879604577407861198472788. Accessed


factory cost predictions using manufacturing influenced design 15 Jan 2017
(MInD). Georgia Institute of Technology Myers RH, Montgomery DC, Anderson-Cook CM (2009) Response
Herrmann JW (2006) Handbook of production scheduling, vol 89. surface methodology. Wiley, Hoboken
Springer, New York Neches R (2011) Engineered resilient systems (ers) s&t priority
Holland JP (2015) Engineered resilient systems, Power of advanced description and roadmap. Office of the Deputy Assistant Secretary
modeling and analytics in support of acquisition. In: NDIA 16th of Defense for Systems Engineering. Washington, DC. http://www.
science and engineering technology conference. https://ndiasto dtic.mil/get-tr-doc/pdf?AD=ADA554975. Accessed 8 Jan 2017
rage.blob.core.usgovcloudapi.net/ndia/2015/system/17945_J_ Oppenhein BW (2011) Lean for systems engineering with lean
Holland.pdf. Accessed 20 Aug 2017 enablers for system engineering. Wiley, New York
Józefowska J, Weglarz J (2006) Perspectives in modern project Peters TJ, Rosen DW, Shapiro V (1994) A topological model of
scheduling, vol 92. Springer, New York limitations in design for manufacturing. Res Eng Design
Kamdar N, Smith M, Thomas R, Wikler J, Mavris D (2003) Response 6(4):223–233
surface utilization in the exploration of a supersonic business jet Rahman S (1998) Theory of constraints: a review of the philosophy
concept with application of emerging technologies. In: Proceed- and its applications. Int J Oper Prod Manag 18(4):336–355
ings of the World Aviation Congress & Exposition, Montreal, Rais-Rohani M, Dean EB (1996) Toward manufacturing and cost
QC, Canada considerations in multidisciplinary aircraft design. In: 37th
Kaufmann M, Zenkert D, Wennhage P (2010) Integrated cost/weight AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics
optimization of aircraft structures. Struct Multidiscipl Optim and materials conference and exhibit. American Institute of
41(2):325–334 Aeronautics and Astronautics
Keane A, Nair P (2005) Computational approaches for aerospace Raj P (1998) Aircraft design in the 21st century: implications for
design: the pursuit of excellence. Wiley, New York design methods. In: 29th AIAA fluid dynamics conference
La Trobe-Bateman J, Wild D (2003) Design for manufacturing: use of Ramos AL, Ferreira JV, Barcelo J (2012) Model-based systems
a spreadsheet model of manufacturability to optimize product engineering: an emerging approach for modern systems. IEEE
design and development. Res Eng Design 14(2):107–117 Trans Syst Man Cybern Part C (Appl Rev) 42(1):101–111
Lee Q (2000) How to balance a manufacturing work cell. In: Institute Raymer DP (2006) Aircraft design: a conceptual approach, 4th edn.
of Industrial Engineers—IE Solutions Conference, Cleveland American Institue of Aeronautics and Astronautics, Reston
Lee M, Ceisel J, Liu Z, Mavris D (2012) A parametric, preliminary Rush C, Roy R (2000) Analysis of cost estimating processes used
structural analysis and optimization tool with manufacturing cost within a concurrent engineering environment throughout a
considerations. In: 53th AIAA/ASME/ASCE/AHS/ASC struc- product life cycle. In: 7th ISPE International Conference on
tures, structural dynamics and materials conference, Honolulu, Concurrent Engineering: research and applications, Lyon,
HI France, July 17th–20th. Technomic Inc., Pennsylvania USA,
Leiva C (2016) Demystifying the digital thread and digital twin pp 58–67
concepts. IndustryWeek. http://www.industryweek.com/sys Russell JD (2006) Composites affordability initiative: transitioning
tems-integration/demystifying-digital-thread-and-digital-twin- advanced aerospace technologies through cost and risk reduc-
concepts. Accessed 20 Aug 2017 tion. AMMTIAC Q 1(3):3–6
Libby SL, Siedlak DJ, Solano HD, Pinon OJ, Mavris DN (2017) Cost- Sarker R, Mohammadian M, Yao X (2002) Evolutionary optimiza-
capability analysis of uav family and flexible factory design. In: tion, vol 48. Springer, New York
17th AIAA aviation technology, integration, and operations Schaffner MA, Shihong MW, Ross AM, Rhodes DH (2013) Enabling
conference, p 4250 design for affordability: an epoch-era analysis approach. Tech.
Ligetti C, Simpson TW, Frecker M, Barton RR, Stump G (2003) rep, DTIC Document
Assessing the impact of graphical design interfaces on design Siedlak DJ, Schlais PR, Pinon OJ, Mavris DN (2015) Supporting
efficiency and effectiveness. J Comput Inf Sci Eng 3(2):144–154 affordability-based design decisions in the presence of demand
Lucero S (2011) Engineered resilient systems-dod science and variability. In: Proceedings of ASME 2015 International Man-
technology priority. Tech. rep, DTIC Document ufacturing Science and Engineering Conference. American
Magrab EB, Gupta SK, McCluskey FP, Sandborn P (2009) Integrated Society of Mechanical Engineers, MSEC2015-9422
product and process design and development. CRC Press, Boca Siedlak DJ, Schmidt TM, Pinon OJ, Mavris DN (2014) A method-
Raton ology for the parametric exploration of the impact of production
Marx C, Thut M (2012) Globalization: aerospace suppliers need a planning on the early stages of design. In: Proceedings of ASME
flight plan to sustain growth. PricewaterhouseCoopers LLP. 2014 International Manufacturing Science and Engineering
http://www.pwc.com/en_US/us/industrial-products/assets/pwc- Conference. American Society of Mechanical Engineers,
gaining-attitude-issue-4-sustain-growth.pdf. Accessed 15 Jan MSEC2014-3974
2017 Simpson T, Peplinski J, Koch P, Allen J (2001) Metamodels for
Mattingly JD, Heiser WH, Pratt DT (2002) Aircraft engine design, computer-based engineering design: survey and recomendations.
2nd edn. American Institute of Aeronautics and Astronautics, Eng Comput 17:129–150
Reston Sloan J (2012) A400m wing assembly: challenge of integrating
Mavris DN, DeLaurentis DA, Bandte O, Hale MA (1998) A composites. Composites World
stochastic approach to multi-disciplinary aircraft analysis and Spero E, Avera MP, Valdez PE, Goerger SR (2014a) Tradespace
design. In: Aerospace Sciences Meeting & Exhibit, AIAA exploration for the engineering of resilient systems. Proc
98-0912 Comput Sci 28:591–600
Mavris DN, Pinon OJ, Fullmer Jr D (2010) Systems design and Spero E, Bloebaum CL, German BJ, Pyster A, Ross AM (2014b) A
modeling: a visual analytics approach. In: 27th Congress of research agenda for tradespace exploration and analysis of
International Council of the Aeronautical Sciences (ICAS) engineered resilient systems. Proc Comput Sci 28:763–772
Michaels D (2012) Hit by delays, airbus tries new way of building Stark J (2005) Product lifecycle management: 21st century paradigm
planes. The Wall Street Journal. http://online.wsj.com/articles/ for product realisation. Springer, New York

123
Res Eng Design

Stebbins M (2015) Systems design using the digital thread. Project Wagner R (1998) North American Sabre Jet F-86d/K/L Part One:
call, Digital Manufacturing and Design Innovation Institute. Design/Structure/Testing. GINTER BOOKS
http://dmdii.uilabs.org/projects/calls/dmdii-15-05-systems-design- Werner-Westphal C, Heinze W, Horst P (2008) Multidisciplinary
using-the-digital-thread-ame. Accessed 8 Jan 2017 integrated preliminary design applied to unconventional aircraft
Sundaresan C, Tang Z, Ceisel J, Mavris D (2012) A methodology for configurations. J Aircraft 45(2):581–590
parametric production planning in preliminary aircraft design. Wilson JM (2003) Gantt charts: a centenary appreciation. Eur J Oper
In: The 28th Congress of the International Council of the Res 149(2):430–437
Aeronautical Sciences (ICAS 2012), Brisbane, Australia Wu MS (2014) Design for affordability in defense and aerospace
Szkody R (1999) Integrated sensor system-demonstration results. In: systems using tradespace-based methods. Ph.D. thesis, Mas-
Gateway to the New Millennium. In: Proceedings of 18th Digital sachusetts Institute of Technology
Avionics Systems Conference (Cat. No.99CH37033), Institute of Wu MS, Ross AM, Rhodes DH (2014) Design for affordability in
Electrical and Electronics Engineers (IEEE) complex systems and programs using tradespace-based afford-
Tang Z, Pinon-Fischer OJ, Mavris DN (2014) Identification of key ability analysis. Proc Comput Sci 28:828–837
factors in integrating aircraft and the associated supply chains Zavadlav E, McClain JO, Thomas LJ (1996) Self-buffering, self-
during early design phases. In: 14th AIAA Aviation Technology, balancing, self-flushing production lines. Manage Sci
Integration, and Operations Conference, p 3271 42(8):1151–1164
Tuttle P, Bobinis J (2013) 4.6.2 specifying affordability, vol 23, No. Zhao X, Curran R, Verhagen WJC (2014) Aircraft component
1. In: INCOSE International Symposium, pp 662–676 multidisciplinary design optimization considering cost perfor-
Usher JM, Roy U, Parsaei HR (1998) Integrated product and process mance. In: 14th AIAA Aviation Technology, Integration, and
development: methods, tools, and techologies. Wiley, New York Operations Conference. American Institute of Aeronautics and
Vascik PD, Ross AM, Rhodes DH (2016) Program and portfolio Astronautics
affordability tradeoffs under uncertainty using epoch-era anal- Zitzler E, Deb K, Thiele L (2000) Comparison of multiobjective
ysis, vol 26. In: INCOSE International Symposium, Wiley evolutionary algorithms: Empirical results. Evol Comput
Online Library, pp 2391–2406 8(2):173–195

123

You might also like