Untangling the Digital

FEATURE
SPECIAL
Thread: The Challenge and
Promise of Model-Based

AUGUST 2O15
VOLUME 18 / ISSUE 2
Engineering in Defense
Acquisition
Timothy D. West, timothy.west.2@us.af.mil and Art Pyster, art.pyster@stevens.edu

D
Abstract. Although MBSE has made great strides in INTRODUCTION
offering better ways to perform systems engineering, uring a typical U.S. Department of Defense (DoD) weapon
it has not yet incorporated the myriad physics-based development program, literally hundreds of digital models
models generated by the discipline engineers during the simulate the behavior of various components of the
detailed design process. The US Air Force is proposing system at various points along the development timeline.
a related approach, the Digital Thread, as a framework Systems engineers maintain and update very few of these models
to merge these detailed design models with MBSE’s as the system design evolves, and even fewer undergo Verification,
conceptual and top-level architectural models, resulting Validation, and Accreditation (VV&A). These models generally fall
in a single authoritative representation of the system. within one of two groups: the conceptual and top-level architectural
Could this approach be the solution to streamlining models associated with MBSE, and the myriad physics-based and
system development, accelerating production, simplifying mathematical models generated by the discipline engineers during
sustainment, and lowering the total system lifecycle cost? detailed design. As part of DoD’s push for more robust use of digital
This paper introduces the Digital Thread concept, provides system models, the U.S. Air Force is advocating a new Digital Thread
an aspirational scenario describing how Digital Thread concept that offers a potential approach to closing that gap between
may be employed in a future combat aircraft acquisition, MBSE and the physics-based models. The goal of this initiative is
and discusses a number of the challenges that currently to finally achieve the dramatic development cycle time reduction
separate this aspiration from reality. that advocates of model-based weapons development within the Air

Reichwein continued from previous page

expect that the next-generation of PLM and ■■ Eclipse Lyo. 2014a. “OSLC MagicDraw ■■ ——. 2015b. “SysML RDF Vocabulary.”
ALM solutions will most likely be based on Adapter.” https://wiki.eclipse.org/Lyo/ http://www.omg.org/techprocess/
Linked Data, as well as OSLC, to achieve MagicDraw. experimental-rdf/SysML/1.3/.
unprecedented levels of data connectivity ■■ ——. 2014b. “OSLC Simulink Adapter.” ■■ SafeCer. 2013. “Safety Certification of
in engineering.  https://wiki.eclipse.org/Lyo/Simulink. Software-Intensive Systems with Reus-
■■ iFEST. 2013. “Industrial Framework for able Components.” http://www.safecer.eu.
REFERENCES Embedded Systems Tools.” ■■ Siemens PLM. 2015. “What’s New in
■■ Cesar. 2013. “Cost-Efficient Methods http://www.artemis-ifest.eu/. Teamcenter Service Pack 10.1.2.”
and Processes for Safety Relevant Em- ■■ OSLC Software. 2015. “Software – Open https://community.plm.automation.
bedded Systems.” Services for Lifecycle Collaboration.” siemens.com/siemensplm/attachments/
http://www.cesarproject.eu/. http://open-services.net/software/. siemensplm/Teamcenter-Blog/14/1/
■■ Crystal. 2015. “Crystal.” ■■ OSLC4MBSE. 2015a. “OMG OSL- Siemens-PLM-Whats-New-in-Teamcenter-
http://www.crystal-artemis.eu/. C4MBSE Working Group.” http://www. Service-Pack-10-1-2-fs-Y16.pdf.
■■ DCMI. 2015. “DCMI Metadata Terms.” omgwiki.org/OMGSysML/doku.php?id=sys- ■■ SPRINT. 2013. “Software Platform for
http://purl.org/dc/terms/. ml-oslc:oslc4mbse_working_group. Integration of Engineering and Things.”
http://www.sprint-iot.eu/.

45
 Force have been espousing since the 1990s designed solution, and comprehensively and analytical rigor to model-based design.
FEATURE
SPECIAL

(Kraft, et al. 2015). After providing a brief and verifiably answer all the system Currently, mathematical models are
overview of MBSE, this paper describes requirements posed by the problem.” In more the domain of tools such as Math-
Digital Thread – the “single authoritative other words, MBSE provides an integrated works’ ubiquitous MATLAB/Simulink,
digital surrogate” (Kraft, CREATE-AV framework upon which a systems engineer the ANSYS family of simulation tools,
and the Air Force Digital Thread 2014) can assemble a core model of the system; as and 3D CAD-based design packages such
of the system under development – along the design matures, systems engineers can as Dassault Systems’ SolidWorks and
with a similar concept, Digital Twin – “the add layers of detail and complexity to create PTC Creo Parametric’s Pro/ENGINEER.
‘as-built’ piece of the Digital Thread (Kraft, a more robust and representative model MATLAB/Simulink is traditionally used to
AUGUST 2O15
VOLUME 18 / ISSUE 2

et al. 2015).” Subsequent sections describe of the final design. A clear benefit of this model signal processing intensive systems,
the initiatives underway in this area, as well approach is the transition “from controlling whereas the 3D CAD programs model
as the potential challenges this concept is the documentation about the system the geometry associated with the physical
likely to face. The paper concludes with a to controlling the model of the system systems. The richness of both of these CAD
fictional scenario of how Digital Thread (Friedenthal, Moore, & Steiner 2015).” programs comes from the many add-ons
could help develop and sustain a future U.S. A commonly used language for MBSE that automate the discipline engineering
Air Force fighter aircraft. is the Systems Modeling Language analysis required for the product, such
(SysML), developed by INCOSE and the as Pro/ENGINEER’s ability to model the
MBSE 101 Object Management Group (OMG) to non-Newtonian flow into the 3-D shape to
Over the last several years, MBSE has support “the specification, analysis, design, help the engineer virtually identify and fix
become a ubiquitous term within the realm verification, and validation of a broad range areas where a plastic injection mold may
of systems engineering, yet there is no of complex systems,” including “hardware, fail to fill properly. However, few of these
well-accepted definition. The International software, information, processes, personnel, tools provide the “real-time” simulation ca-
Council on Systems Engineering (INCOSE) and facilities” (OMG 2012). Software pability needed to fully leverage the Digital
Systems Engineering Handbook (INCOSE developers such as Vitech Corporation Thread concept.
2011) uses the term several times, but fails and Sparx Systems have leveraged the Systems engineers attempted to add
to precisely define it. (The INCOSE Systems power and simplicity of SysML to create quantitative rigor to MBSE models. Buede
Engineering Handbook does discuss the conceptual modeling products such (2000) describes the software development
Object‐Oriented Systems Engineering as CORE and Enterprise Architect, community’s use of Object-Oriented mod-
Method (OOSEM), whose objectives respectively. These products enable system eling to complement its data, process, and
include the “[i]ntegration of model‐based designers to capture the requirements of behavior models with additional technical
systems engineering (MBSE) methods with the system in a digital database, track the depth. Likewise, Vanderperren and De-
object‐oriented software, hardware, and status of those requirements through the haene (2006) cite examples where Unified
other engineering methods.”) The same is verification and validation process, and Modeling Language (UML, a precursor to
true for four commonly used textbooks on even build architectural models of the SysML used in software development) and
systems engineering (Blanchard & Fabrycky system to simulate its conceptual behavior. Simulink models were loosely coupled in
2006) (Buede 2000) (Kossiakoff and Sweet However, MBSE is still very different what they referred to as “co-simulation,”
2003) (Larson, et al. 2009). The Guide to from “One Model to Rule Them All.” The where separate models run in parallel and
the Systems Engineering Body of Knowledge typical MBSE implementation tends to focus exchange information real-time during
(SEBoK) (BKCASE Editorial Board 2014) on what Buede (2000) described as “three simulation execution. Unfortunately,
describes MBSE as “[t]he formalized categories of qualitative modeling approach- this approach requires the creation and
application of modeling to support system es”: (1) data models (Entity Relationship maintenance of two separate models using
requirements, design, analysis, verification Diagrams and IDEF1 Diagrams), which separate tools. The alternative more fre-
and validation activities beginning in the address relationships between inputs and quently employed is to painstakingly code
conceptual design phase and continuing outputs of the system; (2) process models the model from scratch using Java, C++, or
throughout development and later life (Data Flow Diagrams, IDEF0 Diagrams, a similar programming language.
cycle phases”— a definition it shares with and N2 Charts), which decompose the
INCOSE’s Systems Engineering Vision 2020 system functionally and depict the flow of WHAT IS A DIGITAL THREAD?
(INCOSE 2007). Friedenthal, Moore, & inputs and outputs between these func- Notionally, Digital Thread represents
Steiner (2015) expand upon the INCOSE tional subsystems; and (3) behavior models a framework for merging the conceptual
definition by emphasizing MBSE as a (Function Flow Block Diagrams, Behavior models of the system (the traditional focus
mechanism “to perform the systems Diagrams, and State Transition Diagrams), of MBSE) with the discipline-specific engi-
engineering activities that have traditionally which define “the control, activation, and neering models of various system elements.
been performed using the document-based termination of system functions needed to However, the concept is new enough that
approach,” ideally resulting in a “coherent meet the performance requirements of the a robust, widely accepted definition is still
model of the system” and “enhanced system.” Additionally, MBSE implementa- emerging. In the 2013 Global Horizons:
specification and design quality.” Long tions frequently cross the gap from concep- United States Air Force Global Science and
and Scott (2011) describe MBSE as “a tual design into physical architecture, but fall Technology Vision, the Chief Scientist of
‘layered’ process of analyzing and solving short of incorporating the myriad math- the Air Force described Digital Thread as
systems design problems,” where the ematical models created by the discipline a digital linkage between materials, design,
design problems are handled “in layers engineers during the detailed design process processing, and manufacturing that “will be
of increasing granularity in order to for a complex system. In other words, MBSE, the game-changer that provides the agility
converge strategically on the solution;” although an immensely valuable part of the and tailorability needed for rapid [weapon
they also emphasize MBSE’s “use [of] a systems engineering process, tends to be system] development and deployment,
consistent language to describe the problem conceptual and qualitative, whereas Digital while also reducing risk (U.S. Air Force
and the solution, produce a coherently Thread adds additional layers of quantitative 2013).” Later that year at a National Insti-

46
tute of Standards and Technology (NIST) system at that instant of time. To be the of the aircraft. Early in the design process,

FEATURE
SPECIAL
symposium, Kraft (2013) referred to it as authoritative source of information, it has such models could help developers analyze
a “cross-domain, common digital surro- to be both current and complete. To be alternatives and make initial design trades,
gate of a materiel system” that “provides useful to decision makers, someone must and to identify potential design flaws while
the analytical framework for organizing organize this information in a manner that “sunk” development cost is low. Later in the
output from high-fidelity, physics-based facilitates data mining in a thorough but development process, the model could be
models across the entire lifecycle.” During timely manner. This indeed is a lofty goal. used to guide testing; that is, test where the
a National Defense Industrial Associa- Regarding the subtle differences between model indicates potential problems, where
tion (NDIA) conference in 2014, Chris- DSM and Digital Thread, this paper will the model uncertainty is large, and/or

AUGUST 2O15
VOLUME 18 / ISSUE 2
tian (2014) described Digital Thread as sidestep that issue by taking the viewpoint where the consequences of the model being
“an extensible, configurable and Agency that they are complementary concepts with wrong are too high (Kraft 2014).
enterprise-level analytical framework similar goals and similar challenges; the Kraft’s concept for Digital Thread entails
that seamlessly expedites the controlled fact that multiple organizations within DoD the merger of the probabilistic output from
interplay of authoritative data, information, are pursuing similar concepts may increase various high-fidelity mathematical models
knowledge, and computer software in the the likelihood that the concept will become into a reduced-order response surface,
enterprise data-information-knowledge an accepted practice for complex system which systems engineers frequently update
systems, based on the Digital System development across the Department. throughout the development process,
Model template, to inform decision makers Defense contractor Lockheed Martin resulting in a “single authoritative digital
throughout a system’s life cycle by provid- is credited with coining the term Digital surrogate (Kraft 2014)” of the system under
ing the capability to access, integrate and Thread to describe its approach to using development. Because it is a reduced-order
transform disparate data into actionable the 3D solid models generated during the model, the computational time required to
information.” In a prior presentation at the engineering design process to support execute the model is “fractions of a second
same conference, Zimmerman (2014) of- the manufacturing process for both the instead of [the] hours or days” associated
fered the same definition, less the reference computer numerically controlled (CNC) with executing a large computational fluid
to “computer software.” At the American machining of metal parts and the compos- dynamics (CFD) based solution, yet the re-
Institute of Aeronautics and Astronautics ite programming system (CPS) approach duced-order nature of the model “does not
(AIAA) SciTech 2015 conference, Walker to fiber placement in composite parts: “the necessarily translate into reduced accuracy”
(Kraft, et al. 2015) offered a slightly leaner unbroken data link [between the engineer- (Kraft 2014). The rapid execution of the
definition: “an extensible, configurable, and ing CAD model and the models loaded in model enables designers and acquirers alike
enterprise level framework that seamlessly the CNC or CPS machine] is the digital to quickly run various “what if ” scenarios
expedites the controlled interplay of au- thread” (Newton 2014). Lockheed Martin with the model, enabling more insight into
thoritative data, information, and knowl- has reported a number of successes in system performance and more efficient
edge to inform decisions during a system’s employing this approach to improve part fit design optimization. Because the mod-
life cycle by providing the capability to and reduce rework on mechanical parts for el uses a standard framework, users can
access, integrate, and transform dispa- the F-35 fighter (Kinard 2010). Lockheed incorporate it easily into larger wargaming
rate data into actionable information.” In is also using these same models to gener- engagement models to see how various
addition to deleting “Agency,” “analytical,” ate the graphics for training documents iterations of the system design affect the
and “computer software,” this version omits and maintenance manuals, extending the larger air battle. Replacing the simplistic
“the enterprise data-information-knowl- Digital Thread even further through the models used in wargames today will enable
edge systems, based on the Digital System system lifecycle preventing configuration more accurate trade studies on the effects of
Model template,” where the Digital System mismatches between these documents and various aircraft performance characteristics
Model (DSM) is defined by Christian the actual aircraft. (maximum acceleration, turn rate, climb
(2014) and Zimmerman (2014) as “a digital This direct CAD-to-CNC prototyping rate, radio-frequency signature, and more)
representation of a weapon system, generat- merely scratches the proverbial surface in at the “system of systems” level, and a more
ed by all stakeholders, that integrates the terms of potential uses for Digital Thread. accurate assessment of interoperability
authoritative data, information, algorithms, DoD, and particularly the Air Force, shows between legacy systems. Because the model
and systems engineering processes which considerable interest in the concept, and is updates throughout the system lifecycle, it
define all aspects of the system for the generating new ideas on how to integrate remains the definitive, authoritative repre-
specific activities throughout the system the concept into the entire lifecycle of a sentation of the system that decision mak-
lifecycle.” To clarify the distinction between weapon system. The Air Force Test Center ers can use to better inform billion-dollar
Digital Thread and DSM, Kraft (2013) and (AFTC) is championing an expansion of acquisition decisions.
Christian (2014) describe Digital Thread as Digital Thread into the Test and Evaluation The Air Force Research Laboratory
“the physics-based modeling instantiation” (T&E) arena. Since 1995, Kraft has been (AFRL) has offered an additional spin
of the DSM concept. Zimmerman (2014) proposing the use of physics-based compu- on the concept, proposing the creation
describes it more succinctly as a “frame- tational models to augment and stream- of a “Digital Twin” of each individual
work for DSM use.” line the traditional regimen of ground airframe and engine. Walker (Kraft et
Despite their subtle differences, these and flight testing required to prepare a al. 2015) described the Digital Twin as
definitions share a common theme of “au- new military aircraft for production. For “an integrated multi-physics, multi-scale
thoritative data.” These two words deserve instance, he has proposed the coupling of probabilistic simulation of an as-built
special attention because they communi- airplane aerodynamics, propulsion, heat system that uses the best available physical
cate that the vision for Digital Thread is transfer and structural response models models, sensor information, and input
not just a model, or a collection of models. to give a more complete picture of how data from the Digital Thread and a
Rather, its ultimate goal is to be the defini- an entire aircraft system might respond Digital System Model to mirror the life
tive repository of authoritative information to a given flight condition, and how those of its corresponding physical twin,” and
containing everything we know about the responses would accumulate over the life elsewhere as “the ‘as-built’ piece of the

47
 Table 1. Comparison of systems modeling approaches
FEATURE
SPECIAL

“TRADITIONAL” MBSE DIGITAL THREAD DIGITAL TWIN

Focuses on top half of Systems Addresses complete Systems Engineering Additional emphasis on system O&M
Engineering “Vee” Diagram “Vee” Diagram
Maturing approach with several Fledgling approach with a few successes Notional concept, no demonstrated
demonstrated successes success
AUGUST 2O15
VOLUME 18 / ISSUE 2

Acceptable operation on consumer- Slow to moderate operation on today’s Computational power for full
grade computers fastest supercomputers, depending on the implementation currently unavailable
application
Wide variety of commercial software Prototype software tools in development Tool development awaiting required
available today computer power
Promises accelerated system development, Promises further lifecycle cost reductions
better manufacturing quality & through enhanced failure prediction and
consistency, and lower lifecycle cost true condition-based maintenance

Digital Thread.” According to this concept, Design Tool Compatibility

each time the Air Force takes delivery of • Exploration of System Trade 3D CAD-based design packages such
a new aircraft, it also receives a “[trillion] Space as Dassault Systems’ SolidWorks and
degree-of-freedom” digital model of the • Model-Based Analysis of PTC Creo Parametric’s Pro/ENGINEER
as-built aircraft that is “is ultrarealistic in Alternatives output CNC-friendly data files, essentially
geometric detail, including manufacturing automating the current capabilities of
anomalies, and in material detail, includ­ • System Performance Predictions Digital Thread. As robust as they are,
ing the statistical microstructure level, • Sensitivity Analysis of however, these tools lack the sophistication
specific to this aircraft tail number; ” Performance Results to accurately model the myriad physical
the model would accept “probabilistic aspects associated with an aircraft
• Direct CAD-to-CNC & 3D Printer
input of loads, environmental, and usage design, in addition to the requirements
Prototyping
factors, and it also tightly couples to an management features of Core or Enterprise
outer-mold-line, as-built, CFD model” • System Documentation Architect. As an example, consider the
of that aircraft tail number (Tuegel et al. Development aerothermoelastic behavior of an F-16
2011). After each sortie, data logs from • Accelerated Test & Evaluation aircraft wing. In flight, the aircraft wing
that specific sortie could be downloaded bends because of aerodynamic loading
from the physical aircraft and “flown” on • Digital Interoperability created by increased pressure on the
the Digital Twin to ensure the condition Assessments leading edge and the underside of the wing.
of Digital Twin reflects the same amount • System/Component Life Because air temperature generally decreases
of stress and strain as seen on the real Estimation with altitude, this change in thermal
aircraft. With sufficient computing power, environment also affects the shape of the
AFRL believes they can predict where, Figure 1: Potential applications of Digital wing. As the shape of the wing changes,
when, and how a given component will Thread so do its structural behavior and its
fail, enabling more effective, prediction- aerodynamic performance. In certain flight
based maintenance of the aircraft. Table 1 conditions, the wing and/or the weapons,
compares “traditional” MBSE with Digital sensor pods, and associated suspension
Thread and Digital Twin. • Design Tool Compatibility equipment may encounter flutter or other
• Protection of Intellectual dynamic oscillations. This interaction
CHALLENGES TO THE DIGITAL THREAD Property between aerodynamics, thermodynamics,
CONCEPT • Adequate Computer Processing and wing structural elasticity is a very
The Digital Thread concept described Power complex phenomena, not readily simulated
above potentially offers significantly more by toolsets today. Given their niche
utility than simply leveraging existing 3D • Enabling Software applications, the financial incentive for
CAD drawings to manufacture a single part • Verification, Validation & Dassault or PTC to expand their software
or a simple subassembly. A description of Accreditation to incorporate this behavior is low. ANSYS
potential uses is in Figure 1. When properly and aircraft primes offer rudimentary tools
• Model Maintenance
employed, Digital Thread has the potential for modeling this behavior, but not at the
to dramatically shorten the product devel- • Bureaucracy and Cultural Inertia level of fidelity needed for Digital Thread.
opment timeline, enhance product quality To compensate for this lack of fidelity,
and manufacturing consistency, improve Figure 2: Challenges to the Digital Thread conservative developers tend to over-design
sustainability, and reduce the total lifecycle concept the structure, increasing aircraft weight,
cost of the system. However, to achieve which forces a decrease in payload capacity
these benefits, it must first overcome and/or fuel volume.
several technical, programmatic, cultural, Figure 2. The following section describes Developers should incorporate these
and bureaucratic challenges, as listed in those challenges. phenomena into the Digital Thread to make

48
it more viable. Thus, a common language, contractors to provide CREATE-compat- effects on the wings. Having a common

FEATURE
SPECIAL
framework, and/or interface to enable these ible models, or build its own model. For model framework would also facilitate the
models to run in a simultaneous or iterative example, a robust model of the U.S. Air addition of these models into the larger,
fashion are requirements such that the Force’s F-22 fighter would incorporate the operational-level wargaming simulations,
output of one module becomes the input aircraft’s nose and forward fuselage (built by where the battlespace-wide effects of an
for the next iteration of a different module. Lockheed Martin), wings and aft fuselage incremental increase in a given perfor-
This drives tough questions about what the (built by Boeing), and the F-119 turbine mance parameter could be better assessed.
appropriate framework should be for this engines (built by Pratt and Whitney). Using This larger environment could also provide
process, how loosely or tightly to couple the tools such as CREATE, Air Force engineers a playground for assessing interoper-

AUGUST 2O15
VOLUME 18 / ISSUE 2
individual modules within the framework, could reverse-engineer these models from ability between the proposed systems
and what data exchange must occur in information provided by the contractors and those already in operation. Indeed,
order to make such a framework function. and/or attained through performance test- the DoD Engineered Resilient Systems
Further complicating the development ing. For instance, they could easily generate (ERS) Community of Interest has already
of such a framework is the fact that many a basic rigid-body model of the F-22 using demonstrated the feasibility of this “try
of the mathematical models are incompat- only the aircraft’s Outer Mold Line (OML) before you buy” concept in its 2013 analysis
ible with each other. DoD is attempting to geometry. Expanding that model to include of three candidate design concepts for the
address this issue through its Computa- the aircraft’s aeroelastic behavior, however, next generation cargo transport aircraft
tional Research and Engineering Acquisi- requires substantially more information, (the C-X program) using tools similar to
tion Tools and Environments (CREATE) such as details of the structural design those available in CREATE; the team uti-
program, established in 2008 to enhance underneath the aircraft’s skin. Expanding lized a low-fidelity tool capable of quickly
the development and acquisition of new into the aerothermoelastic regime requires estimating performance trends associated
weapon systems by closing capability gaps an understanding of how various elements with changes in a variety of parameters
in weapon conceptual design, design veri- of that substructure dissipate heat, to where including empty weight, maximum takeoff
fication, and the design environment. The those elements dissipate the heat, how weight, maximum thrust, maximum pay-
CREATE program has already delivered those structures change dimensionally as load range, average fuel burn rate, takeoff
several “multi-disciplinary, physics-based they are heated and cooled, and how the distance, landing distance, lifecycle costs,
simulation software products developed to system’s aerodynamic performance changes and cost per flight hour, and injected these
enable full-vehicle design analysis and test- as a result of this heating and cooling. results into an operational engagement
ing via high-fidelity simulation” – includ- Yet, the model would still be incomplete model to better understand the capabilities
ing one for rotorcraft (Helios) and one for without addressing how the aerothermo- and design tradeoffs inherent in each of the
fixed-wing aircraft (Kestrel) – which share a elastic behavior changes in response to pilot three designs (Kraft 2014).
“common scalable infrastructure” that en- inputs, flap scheduling, and other system Makenna (2012) proposed a similar
ables software developers to share modules stimuli that causes the control surfaces to requirements-definition approach for
across both aircraft-type simulators (US deflect. Building a model with this level of the Navy that utilizes “set-based” design
Air Force 2015). Additionally, both Helios fidelity would require enough proprietary principles, where the user defines and
and Kestrel are compatible with Firebolt, information from the prime contractors gradually narrows down broad sets of
the CREATE tool for modeling propulsion that the resulting model, although built by design parameters as tradeoff information
systems. CREATE has developed a similar Government personnel, would itself likely becomes available until the user identifies
toolset for the Navy (CREATE-Ship) to be proprietary. Clearly, the process would be a best solution. Makenna posits that set-
accelerate ship design (Mackenna 2012). more efficient if the contractors developed based design approach produces a design in
According to Dr. Robert Meakin, the pro- and provided the model. This would also the same amount of time as the traditional
gram manager for the Air Vehicles portion avoid the cost and configuration control point-design methodology, while avoiding
of the CREATE program, tools such as challenges associated with building and the pitfall of later including premature
Helios and Kestrel will enable DoD to pro- maintaining two sets of models. Additional- design decisions, designs near or beyond
duce “physics-based, optimized conceptual ly, it would resolve the dilemma of deter- the edge of feasibility, and costly repairs to
designs,” quantify the performance of those mining which model was authoritative. This correct design deficiencies discovered late
designs, and assess performance sensitivity would not necessarily mean the contractor in the developmental process.
to new technologies or proposed design has to use the CREATE toolset; proprietary
enhancements in “days to weeks” vs. the tools capable of generating a CREATE-com- Protection of Intellectual Property
multi-month process associated with many patible output file would suffice. Another challenge associated with
of DoD’s current Analysis of Alternatives One potential solution would be for building a fully robust, layered model
(AoA) studies, resulting in better-informed the DoD to develop its own models to of the system will be addressing the
acquisition decisions earlier in the acqui- determine the system requirements, then proprietary nature of some of those
sition process, before large portions of the mandate model submission during the sub- layers. Pushing Digital Thread to its full
system lifecycle costs are locked in (Meakin sequent source selection process, followed potential will require striking an optimum
2011). Since CREATE is a DoD-funded de- by regular model updates throughout the balance between complete insight into
velopment, the software is not proprietary; duration of the development effort. This the design approach, manufacturing
it is also designed to “be used by an average would enable the source selection team to processes, physical properties, and more
CFD practitioner with a basic level of train- perform their own system performance of the system with protection of the
ing,” unlike many proprietary tools, which comparisons during the selection process, “secret sauce” that gives one company
require extremely knowledgeable personnel and to conduct sensitivity analyses on per- a competitive edge over another. In the
with extensive training to properly interpret formance results; as an example, the team aerospace industry, for instance, each
the results (Meakin 2011). may assess how the maximum range of the company has intellectual property, such
To realize the full potential of such a tool, aircraft varies with changes in the airflow as unique design tools, development
the DoD must either convince its defense boundary-layer and other drag-inducing processes, manufacturing techniques,

49
 and more, that it carefully guards as companies, and most of these components stakeholders would be impacted by the
FEATURE
SPECIAL

proprietary information or trade secrets. incorporate the trade secrets from the damage and to what degree, and so on.
The Economic Espionage Act of 1996 component developer. This means the com- Each program is unique, and each program
(EEA), which criminalizes the theft or ponent vendor must not only come to trust manager must strike a balance between
unauthorized disclosure of trade secrets, the system integrator and/or model builder, program efficiency and information
defines the term as follows: but also the latter’s mechanisms for protect- protection that is acceptable to everyone
“[T]he term “trade secret” means all ing intellectual property. The F-22 example whose information is at risk.
forms and types of financial, business, cited above would require integration of Kraft’s approach to Digital Thread
scientific, technical, economic, or engineer- Lockheed Martin’s proprietary information sidesteps this issue altogether by enabling
AUGUST 2O15
VOLUME 18 / ISSUE 2

ing information, including patterns, plans, regarding the nose and forward fuselage, contractors to provide reduced-order
compilations, program devices, formulas, with Boeing’s proprietary information on inputs to the single authoritative digital
designs, prototypes, methods, techniques, the wings and the aft fuselage, and Pratt surrogate; this enables contractors to
processes, procedures, programs, or codes, and Whitney’s proprietary F-119 engine in- use their own proprietary design tools
whether tangible or intangible, and whether formation, not to mention the proprietary to generate their respective inputs,
or how stored, compiled, or memorialized data from countless sub-tier vendors who while also protecting their proprietary
physically, electronically, graphically, pho- supplied the three main contractors. In this processes and technologies (Kraft 2014).
tographically, or in writing if (A) the owner example, each company would have to trust The challenge with this approach is that
thereof has taken reasonable measures to the others with their intellectual property, it forces the Government to assume each
keep such information secret; and (B) the in spite of ongoing competition and the contractor’s model is a detailed, accurate,
information derives independent economic risk of theft of trade secrets. Furthermore, and technically-rigorous representation of
value, actual or potential, from not being each would have to trust the others’ ability the system elements — at least until test data
generally known to, and not being readily to protect their digital data from external are available to validate this assumption. A
ascertainable through proper means by, the theft, in spite of growing cyber espionage potential solution to this problem would
public (United States 1996). from potential adversaries. be to have the contractor provide enough
In spite of the EEA and the potential Under such a concept, vulnerability information about the system to enable the
penalties it establishes, economic espionage in one participant’s network puts the Government to generate its own moderate-
does occasionally occur. Even the most intellectual property of all the other players fidelity model with which it could perform
reputable of companies are not immune at risk too. When weapon systems are an order-of-magnitude verification of
from the misdeeds of a few bad actors with- involved, this is not just a financial or the contractor’s model. However, Kraft’s
in its ranks. In 2003, for instance, the US intellectual property risk; it is a national approach also poses a more fundamental
Air Force cancelled over $1 billion worth security risk; directly to the pilots flying quandary: how can a reduced-order model
of space-related contracts with a major those airplanes and then to the people also be authoritative? Should not the model
defense contractor for the latter’s inappro- and property that they protect. Like most with the higher fidelity be authoritative?
priate acquisition of “thousands of pages risks, these have mitigation strategies, but
of trade secrets” from an industry arch mitigation often translates into additional Adequate Computer Processing Power
rival (Pae 2003). That same year, a start-up scope of work and, potentially, cost and A key enabler in the progress to date
company in the space business alleged that schedule growth. Just like the commercial- with Digital Thread has been the steady
a defense contractor “‘blatantly’ stole [its] sector counterpart, a DoD program increase in available computational power.
trade secrets while the two companies were manager must make smart decisions about Computational power has historically
discussing a potential venture” (Pae 2003). how to manage risk without crippling the grown at an exponential rate, doubling ev-
Meanwhile, the Federal Bureau of In- efficiency of the development program. ery 18 months, in accordance with Moore’s
vestigation reports that foreign espionage In the hypothetical case of our F-22 law (Kaku 2011), while the cost per tran-
efforts continue to grow, with U.S. com- model, it could be stored on a computer sistor has decreased by an exponential rate,
panies having suffered financial losses in located in a vault at Lockheed, with no declining by half in approximately the same
excess of $13 billion during fiscal year 2012 external connectivity. This might work 18 month timeframe (Kurzweil 2005). That
(Dilanian 2012). This reportedly includes for Lockheed, but would be challenging exponential growth is not yet sufficient to
“top secret” information on 25 of America’s for Boeing and Pratt & Whitney – and allow real-time modeling of complex aero-
most advanced weapon systems, includ- for the Air Force engineers who need dynamic flow patterns. Sophisticated CFD
ing the F-35 Joint Strike Fighter, the V-22 access to the model. Even at Lockheed, it models often take days to weeks to run,
Osprey, the Aegis Ballistic Missile Defense would likely drive extra overhead costs even on supercomputers today. The compu-
System, and the Black Hawk helicopter, associated with more stringent security tational demands for Digital Twin are even
among others (Millar and Knight 2013). measures, additional personnel screening, greater: for this concept to work, we must
In the case of the F-35, this hacking likely vault maintenance, etc. A less expensive be able to simulate an hour’s worth of flight
slowed down the F-35 fielding, since the but perhaps riskier approach would be to time on the digital model within an hour
developer probably had to rewrite soft- establish encrypted connections between of real time. To put this processing power
ware code and redesign communications the developers and the Air Force. Further into perspective, consider China’s Tianhe-2,
system to overcome vulnerabilities exposed cost reduction occurs by only encrypting currently the world’s fastest computer,
through the security breach (Reed 2012). the sensitive information and sharing which operates at 33.86 petaflops (Knapp
Clearly, potential industrial espionage the non-sensitive information on an 2014). AFRL’s scientists predict that a full
and international hacking into defense con- unclassified server, but is the risk of doing Digital Twin model will require “exaflop”
tractors’ computer systems will complicate this acceptable? In true academic fashion, computer horsepower (100 times faster
any effort to build a consolidated end-to- the answer is “it depends.” It depends than Tianhe-2); as Figure 3 depicts, such
end Digital Thread model for any of these upon how sensitive the information really computational capability is not projected to
weapon systems, since all are composed of is, how much damage will be done if exist until 2022, and without it, “the power
components and subsystems from myriad the information is compromised, which of the Digital Twin for life prediction and

50
1500 Eagle to minimize the flight test matrix in of the increase in raw computing power

FEATURE
SPECIAL
areas where anomalies are unlikely, and (Maccaba 2014).” He also blames the pro-

Processor Speed (Petaflops)

Viable Processor Power to increase test safety where one expects liferation of bloated operating systems for
1200
For Digital Twin anomalies by creating a pre-flight pre- accelerating obsolesce of computer hardware
diction of how severe one anticipates the and exacerbating maintenance requirements
900 anomalies to be. Seek Eagle is even going for legacy software. Clearly, to fully realize
one step beyond basic aerothermoelastic the potential of the Digital Twin requires
modeling by incorporating human-in-the- a healthy dose of software engineering, in
600 loop flight control inputs and the associated addition to hardware enhancements.

AUGUST 2O15
VOLUME 18 / ISSUE 2
digital processing that occurs between the Software development investments
300 pilot’s hand and the F-16’s control surfaces must continue in order to reach the full
to enhance the realism of the simulation potential of Digital Thread. A growth in
output. Simulation and flight test results processor core count without companion
0 compared favorably for the F-16, although modifications in the software to utilize
2015 2017 2019 2021 2023 “millions of CPU hours” were required to those cores results in untapped processor
reach this level of fidelity. To accomplish potential. Because these tools need to be
Figure 3: Predicted growth of computa- this simulation, Seek Eagle used an Army open-source and/or nonproprietary in order
tional power supercomputer with 1,600 processor cores; to reach beyond corporate boundaries, the
Seek Eagle estimates that 4,000 to 8,000 U.S. Government will likely remain the
processor cores will be required within primary source of that funding. The tools
decision making is lost” (Tuegel et al. 2011). two years to handle the growth in mod- generated by the DoD-funded CREATE
This 2022 timeframe assumes that Moore’s el complexity and the demand for faster program fit this profile. The fact that the
Law of exponential computational power computational turn-times: “[t]o be a viable CREATE program is run by the same
increase remains true, which many doubt M&S [Modeling and Simulation] option for office tasked with delivering the massive
– including Gordon Moore himself. When the Test and Evaluation (T&E) community, computational horsepower required by
asked in 2007 about the future of his name- jobs must be submitted and results ana- Seek Eagle and others is a positive sign
sake law, he predicted that it would end in lyzed in days or weeks rather than months” that the DoD understands the importance
“ten to fifteen years” (Kaku 2011). (Clifton, Ratcliff, and Bodkin 2014). of investing in companion software for its
Although the full utility of Digital Twin’s The office tasked with delivering the supercomputing hardware. Whether the
life prediction capability may remain a massive computational horsepower re- DoD continues to fund its HPCMP despite
decade in the future, engineers are already quired by Seek Eagle and others in the near looming budget challenges will be a telltale
exploiting the potential of Digital Thread term and with meeting AFRL’s Digital Twin sign of just how committed it is to achieving
through such efforts as described above for processing requirements in the longer term the Digital Thread/ Digital Twin concept.
the C-X program, and through a related is the DoD High Performance Comput-
study that uses historical wind tunnel ing Modernization Program (HPCMP). Model Verification, Validation, and
test data from the F-22 against which Providing over 2.5 billion processor hours Accreditation
to compare simulation results from the per year, HPCMP consists of five super- Verification, Validation, and Accredita-
CREATE. The intent of this study is to computing centers; its 2014 acquisition of tion (VV&A) are key elements in the pro-
determine the flight regimes where the two supercomputers worth $65 million cess of assuring the model is accurate. DoD
simulation data are accurate, employ a more than doubled its available computing Instruction 500.61, which mandates VV&A
statistically-based Design of Experiments power from 8.1 to 16.5 petaflops (Hint- for DoD models, simulations, and associat-
approach to select the minimum number son 2014). To realize the full potential of ed data, offers the following definitions for
of test points required in these regimes, Digital Thread and Digital Twin, HPCMP these terms (US DoD 2009):
and assess the loss in model fidelity must continue this level of investment for Verification. The process of determining
associated with using the reduced data another decade. that a model or simulation implementation
set; the study team believes it can utilize and its associated data accurately represent
this approach to reduce the four-year Software that Leverages that Computer the developer’s conceptual description and
wind tunnel test campaign traditionally Horsepower specifications.
associated with a fighter aircraft Newer and faster computer hardware Validation. The process of determining the
development program by as much as 25 alone is insufficient to close this gap in degree to which a model or simulation and
percent (Kraft 2014). computational power. Doing so will also its associated data are an accurate represen-
The Air Force Seek Eagle Office, which require new software applications that more tation of the real world from the perspective
certifies weapons, tanks, and external pods optimally exploit these faster speeds. As of the intended uses of the model.
for carriage and release on Air Force fighter Kaku indicates, “brute computational power Accreditation. The official certification
aircraft, is likewise using computational- does not equal intelligence,” or in this case, that a model or simulation and its associat-
ly-intensive, analytical tools to simulate intelligent modeling (Kaku 2011). Maccaba, ed data are acceptable for use for a specific
the aerodynamics associated with weapons who claims that the overall corporate return purpose.
separation and the aircraft flying quali- on investment in information technology To paraphrase Peter Drucker (2010), ver-
ties before and after weapon separation. has declined in spite of exponential growth ification is the process of ensuring we “built
Although Seek Eagle’s approach utilizes in hardware capability, points to size and the model right” by determining whether it
a variety of tools instead of CREATE, the complexity (vs. capability) as the soft- meets the associated design-to or build-to
objective is similar: to verify conditions ware features that follow Moore’s Law as a specifications; validation is the process
where adverse flying qualities or unaccept- primary reason for this decline, noting that of ensuring we “built the right model” by
able flight control behavior are likely to the growth in “the size and complexity of determining whether it satisfies the needs
occur. This information would enable Seek operating systems has largely offset much of the stakeholders; and accreditation is

51
 the process of officially certifying that the F-119 engine is a composite model built exist in each of these tribes, especially at
FEATURE
SPECIAL

model is acceptable for its intended use. In from Digital Twin models of ten major the early stages of implementation. Digital
the case of a Digital Thread/Digital Twin, components, and that the fleet requires 500 Thread may even encounter inertia from
none of these should be a one-time, Big engines. Neglecting the spares associated the systems engineering community,
Bang event, and all rely on mature software with those ten components, this would where MBSE has already proven to be fit
development processes that includes clear require the developer to maintain 5,500 for its intended purpose. Additionally,
requirements definition, rigorous design Digital Twins – one for each of the major the DoD’s acquisition system, contractual
documentation, a formal configuration components for each of the engines, in language, and other bureaucratic processes
control process, and a thorough test and addition to the 500 full-up engine models. will require modification to better enable
AUGUST 2O15
VOLUME 18 / ISSUE 2

evaluation program. A novel approach to Following each flight, the models associated Digital Thread. Strong leadership and
the latter is the Test-Driven Development with the 20 components installed in that demonstrated results will be the keys to
(TDD) process recommended by the particular pair of engines would have to overcoming this inertia, results such as
Extreme Programming (XP) software de- be aged to reflect the stresses and strains those described in the following scenario.
velopment methodology; in this approach, generated by that flight, and the composite
software developers write the test code for Digital Twin for each of the two engines PROSPECTIVE DIGITAL THREAD
each block of software before writing the updated to reflect the current condition of SCENARIO: F-X DEVELOPMENT
software itself, forcing the developer to its constituent components. Never mind Digital Thread and Digital Twin are both
articulate module requirements, associated that the actual F-119 has significantly more intriguing concepts that promise significant
inputs and outputs, and potential ways that than ten components and each of those improvement over the traditional MBSE
a user might break the code before the first has subcomponents requiring modeling to approach in terms of reducing technical
line of software is written (Larman 2005). accurately predict the likelihood of one of and programmatic risk, accelerating the
Formal VV&A of a model can take those parts failing on the next sortie. The development process, and minimizing the
longer to accomplish than building the effort required to maintain accurate Digital overall lifecycle cost. To illustrate how to
model itself, and can account for up to 30 Twins of the aircraft engine inventory, let realize these potential benefits, consider
percent of the total cost of developing the alone the entire aircraft fleet, is a significant
the following hypothetical scenario.
model (Farr 2007). This is especially true effort and would only be feasible if the In October 2017, the U.S. Air Force
when decisions made based on the output approach to maintaining the model and its received authorization from Congress
of that model potentially affect life and accreditation is simple, fast, and affordable to conduct an analysis of alternatives
limb. For instance, when an inspection – adjectives that are not traditionally to finalize requirements for the next-
reveals micro-cracking in a wing spar, but associated with model accreditation. generation fighter aircraft, designated F-X.
the Digital Twin model says we can fly To receive buy-in across the acquisition A team of DoD experts developed five
the jet for another 100 hours, do we trust community, the requirement is real data notional concepts, using the performance
the model? Do we waive the inspections to show that the model maintenance is characteristics of the F-22 as their initial
altogether until the model says they are actually cheaper than the cost associated baseline. Utilizing a set-based design
needed? For this concept to work, we with scrap, rework, unnecessary approach and the DaVinci tool created
must develop an affordable mechanism for maintenance actions, and more. under the CREATE program to perform
assessing model accuracy, understanding parametric analyses of aircraft concepts,
inherent assumptions and limitations, and Bureaucracy and Cultural Inertia they evaluate each of these concepts at
determining where we can and cannot Just as MBSE adoption did not occur various altitudes, airspeeds, g-loads, and
consider its output as definitive. overnight, neither will Digital Thread. angles of attack, within a system of systems
Although the CAD-to-CNC variant of realistic threat environment. A team of
Model Maintenance Digital Thread is receiving mainstream seasoned combat pilots reviewed the results
Assuming we do succeed in attaining an acceptance, the more revolutionary aspects of the simulation to provide face validity,
accurate model and in getting the model of this concept will likely face significant assess the trade space associated with each
accredited, the accuracy of the Digital cultural inertia. The test community will design, and selected a high-performance,
Thread will quickly degrade if not kept up have to be convinced that the digital model twin-engine, tailless aircraft with extended
to date with each design change. In the case is accurate, and that systems engineers can range capability. The DaVinci results for
of the Digital Twin concept, it must remain properly calibrate it with a much smaller this configuration defined the system-level
current with the replacement of each part, test matrix. The model accreditors will specifications for the aircraft.
even when the replacement has the same have to be convinced to minimize, and in Following approval of the specifications
form, fit, and function as the original. some cases eliminate, actions needed to and formal authorization to initiate
Perhaps a different CNC machine made reaccredit the model each time it changes; procurement, the Air Force released two
the new part or it was subject to a different Digital Thread is, after all, intended to be a Requests for Proposals (RFPs). These
head treatment. For myriad reasons, it may living model, implying myriad changes over RFPs were unique in that they were the
possess slightly different electrical, thermal, the lifecycle of the system. Maintainers will first to require offerors to submit Kestrel-
or mechanical properties, taken into have to be convinced that Digital Thread compatible models of aircraft design
consideration by such predictive models. and Digital Twin offer a safe and more concepts and Firebolt-compatible models of
Additionally, if one desires to include the effective approach to system maintenance engine designs. The Air Force received three
effects of wear and usage on the Digital than the traditional time-based and usage- engine and three aircraft proposals; one of
Twin, a separate digital instantiation would based preventative maintenance strategies. the engine proposals was “nonresponsive”
be required for each part or subcomponent The acquisition community will have to be – the eliminated offeror failed to include
in the inventory, even if those parts convinced that the return on investment a Firebolt model with its proposal. The
are initially identical in every way. For in Digital Thread justifies the added up- reviewers performed a six-model virtual
instance, assume the Digital Twin of the front cost to the system. Naysayers will “fly-off,” comparing each aircraft/engine

52
combination. The fly-off revealed that of-the-envelope test points, where Kestrel aircraft’s Digital Twin. Additionally, the Air

FEATURE
SPECIAL
one of the three candidate aircraft designs had proven to be most accurate. A similar Force determined that the time, expense,
was inferior and failed to meet critical approach optimized the test matrix for the and opportunity cost to do so did not
performance parameters; that revelation ground testing of an engine prototype. The justify modeling every system, even if the
eliminated an offeror from the competition. results of both tests updated the Digital computational resources were available. In-
The reviewers compared the simulations Thread representation of F-X. Additionally, stead, engineers modeled only the engine,
resulting from the four remaining the HPCMP funded the Air Force Research airframe, and key structural components at
configurations along with more traditional Laboratory to investigate areas where the IOC; for three squadrons of 18 planes each,
source selection criteria (development Kestrel predictions were incorrect and to even this demanded a large chunk of the

AUGUST 2O15
VOLUME 18 / ISSUE 2
cost, unit procurement cost, estimated update the aerodynamic algorithms to HPCMP resources. Initially, the Air Force
lifecycle cost, proposal risk) to choose the reduce these errors on future simulations. employed a traditional flight-time based
winning aircraft and engine design. The As a result of a solid initial design, a maintenance program. For instance, the
new contracts required the winning vendors stable funding profile, and a fixed require- Air Force initially removed and thoroughly
to maintain model currency throughout ments set, the first F-X prototype was deliv- inspected engines after every 2,000 hours
the developmental program and to deliver ered in 2025 – a full year ahead of schedule. of operation. At each inspection, engi-
updated models to each other and to the The key to early delivery was design stabil- neers compared each engine to its Digital
Government’s F-X Program Office at least ity and accurate performance predictability Twin, and updated the latter to reflect the
quarterly. The prime contractors signed throughout the design process. Test pilots inspection results. Within two years, as the
Nondisclosure Agreements (NDAs) with at the Air Force Test Center conducted a Digital Twin became more accurate, the
each other that assessed stiff penalties performance, stability, and control flight inspection interval was expanded to 5,000
for intellectual property violations and test matrix that was 15 percent smaller than hours or projected failure. It took another
for negligent handling of proprietary that proposed for the F-22 twenty-five years 10 years to convince the Air Force main-
information. The NDAs also authorized the earlier; unlike the explosive growth in test tenance community to trust the Digital
prime contractors to inspect each other’s matrix size seen once the F-22 began flying, Twin projections for the full 10,000-hour
information protection protocols. The the F-X matrix only grew by three percent engine life. By that time, maintainers used
primes established similar NDAs with any over the two-year performance, stability, commercially available tablet computers to
subcontractors with access to proprietary and control test phase. perform post-flight aging.
information. The one major setback to the F-X devel- The above scenario is full of aspirational
By the time the Engineering and Man- opment program occurred during avionics optimism, but it clearly illustrates the
ufacturing Development phase started in systems testing. The CREATE program’s potential technical benefits of Digital
2019, the HPCMP possessed 200-petaflop early emphasis on aerodynamics, thermal Thread and Digital Twin as an instantiation
computational capability, reducing the performance, and structural loading result- of the DSM concept. These benefits
required run time for a given condition ed in minimal investment in simulating the include an earlier understanding of how
by 85 percent. Additionally, the HPCMP in-flight behavior of advanced electron- the proposed requirements translate into
had released a new version of Kestrel with ics. The four-year avionics test program projected system performance, a better
a more refined mesh generation logic that stretched to five because of electromagnetic understanding of the requirements trade
automatically varied the grid density with interference from the new radar, which space and the system impact associated
the complexity of the aircraft geometry caused spurious output on the aircraft with those trades, earlier opportunity
and the complexity of the airflow over datalink system. The F-X program reached for user buy-in based on more realistic
the aircraft surface. For instance, the grid Initial Operational Capability (IOC) in late models, earlier identification and
was smaller near the leading and training 2032, following the fielding of three squad- resolutions of design issues, more accurate
edges of the wing than in the middle of rons at Langley Air Force Base in Virginia. failure prediction, and more effective
the wing. As the angle of attack increases, Throughout the test program, the on- lifecycle sustainment. The scenario also
and with it the propensity for turbulence board aircraft instrumentation recorded highlights some important issues, such
and flow separation on the wing surface, detailed information about every second as the need for stable requirements.
Kestrel automatically increases grid density. of flight. In 2025, shortly after the HPC- Developers cannot stabilize the design if
This dynamic mesh generation capability MP reached 2,000-petaflop computational the requirements are in constant flux; no
eliminated the need to develop a worst-case capability, the engine vendor implemented matter how much computational power
grid, reducing the number of calculations a Digital Twin trial. The Air Force installed they have at their disposal. Likewise,
required for a given condition, yielding two new, heavily pedigreed engines in the when the budget changes dramatically,
another 15 percent reduction in run time. F-X airframe. After each sortie and each the requirements must also change to
The dramatic reduction in compu- engine run, engineers exposed the Digital balance the books; hence, the need for
tational run time generated by the new Twin for each engine to the same condi- predictable funding. The scenario also
supercomputers and the updated Kestrel tions seen by the real engines. At first, each highlighted the need for the customer
software enabled the test team to generate a hour of simulated aging required an hour (in this case, the Air Force) to drive the
high-fidelity prediction of aircraft perfor- of supercomputer runtime per engine; at implementation through such measures as
mance throughout the flight envelope. The IOC, the HPCMP had 20,000 petaflops (or contractual clauses. Additionally, in order
test team leveraged the Kestrel predictions, 20 exaflops) of computational capability, to make the digital fly-off fair, all vendors
in conjunction with a Design of Experi- enabling an hour of simulated aging on were required to submit models that
ments-based approach, to develop a wind both engines in less than five minutes. were compatible with the Government’s
tunnel test matrix that was 20 percent In spite of the massive processing CREATE toolset, which also enabled the
smaller than used on the F-22, despite capability available in 2032, this was still Government analysts to make adjustments
the larger flight envelope of the F-X. This insufficient to model every failure mode in for differing assumptions in the various
was achieved by eliminating most heart- every component in every system on the models. In addition to demonstrating the

53
 importance of having the right software, design solution can meet the stringent cost tiple fronts to be successful. Mechanisms,
FEATURE
SPECIAL

the scenario also demonstrated how an and capability requirements needed to such as new contract language, could
imbalanced emphasis on one design area ensure viability of the program. demand delivery of these models with the
(e.g., air vehicle performance) could weapon system. We need improvements
cause the design team to overlook other TWO AUTHORITATIVE DIGITAL SURROGATES? to better protect digital data from hacking
crucial aspects of the design – in this Assume for a moment that Digital and espionage. The DoD could champion
case, the electromagnetic interference, an Thread is able to overcome all of the the effort to develop the framework and the
area where many recent fighter aircraft above challenges. That still leaves a major interface to interconnect the various mod-
development programs have struggled (The elephant in the room: Digital Thread ules; the CREATE tools described above are
AUGUST 2O15
VOLUME 18 / ISSUE 2

CREATE program is currently developing and MBSE appear to be on two separate a step in this direction. Each of the Services
a radio-frequency module to help avoid and distinct paths, with the Systems could make a requirements definition using
such electromagnetic issues). Finally, it Engineering community supporting the set-based design process and source se-
acknowledges that just because we can MBSE and those such as Kraft and Walker, lection via virtual fly-off essential elements
model every component in the system does whose backgrounds are rich in discipline in its acquisition strategy for all major
not necessarily mean we should. engineering, advocating the Digital programs; Walker’s identification of five
This scenario might not be as far- Thread framework. If not addressed, Air Force pathfinder programs is a major
fetched as it appears. During his AIAA this schism could lead to a two-model step in this direction. The DoD could also
presentation, Walker (Kraft 2015) indicated scenario. Although this may initially look for creative ways to employ the Digital
the Air Force has already identified four seem to be a minor problem, it results Thread throughout the lifecycle, and invest
“pathfinder” acquisition programs for in the development, maintenance, and in those opportunities when they arise. This
application of the Digital Thread/Digital accreditation of two separate models, which is currently happening in Kraft’s study to
Twin concept: the Joint Surveillance and drives additional cost and scope into the compare Kestrel aerodynamic predictions
Target Attack Radar System (JSTARS) program. It also creates the potential for for the F-22 with actual wind-tunnel data in
Recapitalization Program, the Air Force model mismatches caused by configuration order to determine where to reduce tunnel
Next Generation Trainer (T-X) Program, control issues. For instance, an update testing on future aircraft developments, and
the Global Positioning System (GPS) in one model was never added into the to determine where to enhance the Kestrel
Operational Control Segment (OCX) other model, or perhaps it was incorrectly algorithms to better reflect aerodynamic
Upgrade, and the Long Range Standoff incorporated. To avoid this problem, a reality. However, none of these activi-
(LRSO) Cruise Missile Program. These common language and/or interface needs ties will yield the full potential of Digital
four pathfinders represent a diverse sample to be established that integrates (or, as a Thread without continued investment in
set to validate this new approach: T-X minimum, federates) these two systems people, processes, and processor power.
will likely be a commercial-off-the-shelf into a unified whole. Without this bridge, Each of the Services must remain commit-
(COTS) aircraft, LRSO a from-scratch the potential benefit of the Digital Thread ted to restoring the technical expertise they
missile design, JSTARS a COTS airframe concept will significantly diminish. possessed 20 years ago, and to growing a
with specialized equipment, and GPS new generation of in-house scientists and
OCX a ground-based satellite control THE WAY AHEAD engineers capable of fully utilizing the pow-
system. Walker emphasized the need for Clearly, we are very far from the “One er of the CREATE toolset and the Digital
Digital Thread on LRSO, where accurate Authoritative Model” utopia. However, that Thread concept. Through continuing efforts
system models, including manufacturing fact should not dissuade us from contem- such as this, the model-based engineering
and production models, will be required plating and perhaps taking steps in that di- community could make progress towards
up front in order to ensure the proposed rection. Indeed, we require action on mul- the vision of “One Authoritative Model.” 

REFERENCES
■■ BKCASE (Body of Knowledge and Curriculum to Advance ■■ Dilanian, K. 2012. “Foreign Spying Against U.S. Companies on
Systems Engineering) Editorial Board. 2014. The Guide to the Rise, FBI Says.” Los Angeles Times. 29 June.
the Systems Engineering Body of Knowledge. Version 1.3.1. R. http://articles.latimes.com/2012/jun/29/business/la-fi-economic-
D. Adcock, ed. Hoboken, US-NJ: The Trustees of the Stevens espionage-20120629/.
Institute of Technology. ■■ Dilanian, K., and C. Parsons. 2013. “White House Adopts New
■■ Blanchard, B. S., and W. J. Fabrycky. 2006. Systems Engineer- Strategy to Safeguard Intellectual Property.” Los Angeles Times.
ing and Analysis. 4th ed. Upper Saddle River, US-NJ: Pearson 20 February. http://articles.latimes.com/2013/feb/20/business/
Education, Inc. la-fi-obama-cyber-20130221/.
■■ Buede, D. M. 2000. The Engineering Design of Systems: Models ■■ Drucker, P. F. 2010. The Drucker Lectures: Essential Lessons
and Methods. Hoboken, US-NJ: John Wiley & Sons, Inc. on Management, Society and Economy. New York, US-NY:
■■ Christian, T. 2014. “Determining the Contents of the Digital McGraw-Hill.
System Model.” Paper presented at the NDIA 17th Annual ■■ Farr, J. V. 2007. Simulation of Complex Systems and Enterprises.
Conference on Systems Engineering. 30 October. Springfield, Hoboken, US-NJ: Stevens Institute of Technology.
US-VA: National Defense Industrial Association. http://www. ■■ Friedenthal, S., A. Moore, and R.Steiner. 2015. A Practical
dtic.mil/ndia/2014system/16968ThursdayTrack4Christian.pdf/. Guide to SysML - The Systems Modeling Language. 3rd ed.
■■ Clifton, J. D., C. J. Ratcliff, and D. J. Bodkin. 2014. Stability Waltham, US-MA: Elsevier, Inc.
and Control Test and Evaluation Process Improvements through ■■ Haskins, C., ed. 2011. Systems Engineering Handbook: A Guide
Judicious Use of HPC Simulations. Eglin Air Force Base, US-FL: for System Life Cycle Processes and Activities. Version 3.2.2.
US Air Force Seek Eagle Office. http://www.hpc.mil/images/hpc- Revised by K. Forsberg, M. Krueger, D. Walden, and R. D.
mpo_images/featured/clifton_18jun13.pdf. Hamelin. San Diego, US-CA: INCOSE.

54
■■ Hintson, R. 2014. “High Performance Computing Moderniza- Association. http://www.ndia.org/Divisions/Divisions/System-

FEATURE
SPECIAL
tion Program Doubles Capabilities.” 25 January. Lorton, US- sEngineering/Documents/Committees/M_S%20Committee/2011/
VA: US Department of Defense. http://www.hpc.mil/images/ February/NDIA-systems engineering-MS_2011-02-15_Meakin.pdf.
hpcdocs/newsroom/HPCMP_Doubles_Capabilities.pdf/. ■■ Millar, L., and Knight, B. 2013. “Chinese Spies Hacked Secret
■■ INCOSE Technical Operations. 2007. Systems Engineering US Weapons Systems Including F-35 Joint Strike Fighter.” ABC
Vision 2020, version 2.03. Seattle, US-WA: International News Online. 28 May. http://www.abc.net.au/news/2013-05-29/
Council on Systems Engineering. reports-chinese-hackers-targeted-us/4719352/.
■■ Kaku, M. 2011. Physics of the Future: How Science Will Shape ■■ Newton, R. 2011. “Manufacturing With a Digital Thread.”
Human Destiny and Our Daily Lives by the Year 2100. New GraphicSpeak. 21 April. http://gfxspeak.com/2011/04/21/

AUGUST 2O15
VOLUME 18 / ISSUE 2
York, US-NY: Doubleday. manufacturing-with-a-digital-thread/.
■■ Kinard, D. 2010. “The Digital Thread – Key to F-35 Joint Strike ■■ Open Source Specification Project. 2012. OMG Systems Model-
Fighter Afforadability.” Aerospace Manufacturing and Design. ing Language (SysML). Version 1.3. Needham, US-MA: Object
1 September. http://www.onlineamd.com/amd-080910-f-35-joint- Management Group. http://www.sysml.org/docs/specs/OMGSys-
strike-fighter-digital-thread.aspx#.VXwx78vbKcw/. ML-v1.3-12-06-02.pdf/.
■■ Knapp, A. 2014. “China Once Again Boasts The World’s Fastest ■■ Pae, P. 2003. “$1 Billion in Contracts Yanked From Boeing.”
Supercomputer.” Forbes. November 17. http://www.forbes.com/ Los Angeles Times. 25 July. http://articles.latimes.com/2003/
sites/alexknapp/2014/11/17/china-once-again-boasts-the-worlds- jul/25/business/fi-boeing25/.
fastest-supercomputer/. ■■ ––. 2003. “Boeing Satellite Unit Stole Trade Secrets, Suit Says.”
■■ Kossiakoff, A., and W. N. Sweet. 2003. Systems Engineering: Los Angeles Times. 28 May. http://articles.latimes.com/2003/
Principles and Practice. Hoboken, US-NJ: John Wiley & Sons, may/28/business/fi-boeing28/.
Inc. ■■ Reed, J. 2012. “Did Chinese Espionage Lead to F-35 Delays?”
■■ Kraft, E. M. 2014. “CREATE-AV and the Air Force Digital DefenseTech. 6 February. http://defensetech.org/2012/02/06/
Thread.” Paper presented at the 53rd AIAA Aerospace Sciences did-chinese-espionage-lead-to-f-35-delays/.
Meeting, Kissimmee, US-FL. American Institute of Aeronau- ■■ Stackley, S. J. 2012. “Use of In-House Engineering and Techni-
tics and Astronautics. doi: 10.2514/6.2015-0042. cal Resources.” Washington, US-DC: Assistant Secretary of the
■■ Kraft, E. M. 2013. “Expanding the Digital Thread to Impact Navy. http://www.secnav.navy.mil/rda/workforce/Documents/
Total Ownership Cost.” Paper presented at the 2013 NIST ASNRDAUseofInHouseEngineeringTechResourcesFeb20121.pdf.
Model-Based Enterprise Summit, Gaithersburg, US-MD. ■■ Tuegel, E. J., A. R. Ingraffea, T. G. Eason, and S.Spottswood.
December 18. National Institute of Standards and Technology. 2011. “Reengineering Aircraft Structural Life Prediction
http://www.nist.gov/el/msid/upload/1Kraft_DigitalThread.pdf/. Using a Digital Twin.” International Journal of Aerospace
■■ Kraft, E. M., K. Baldwin, J. P. Holland, and D. E. Walker. 2015. Engineering. January. New York, US-NY: Hindawi Publishing
“The Digital System Model: The New Frontier in Aerospace & Corp. http://www.hindawi.com/journals/ijae/2011/154798/.
Defense Acquisition.” Paper presented at the AIAA SCITECH ■■ US Department of Defense. 2009. DoD Instruction 5000.61:
2015. 12 January. Kissimmee, US-FL. American Institute of DoD Modeling and Simulation (M&S) Verification, Validation,
Aeronautics and Astronautics. http://jamesdrewjournalist.com/ and Accreditation (VV&A). http://dtic.mil/whs/directives/corres/
tag/scitech-2015/. pdf/500061p.pdf/.
■■ Kurzweil, R. 2005. The Singularity Is Near: When Humans ■■ United States. 1996. Public Law 104-294: Economic Espionage
Transcend Biology. New York US-NY: Penguin. Act of 1996. http://www.gpo.gov/fdsys/pkg/PLAW-104publ294/
■■ Larman, C. 2005. Applying UML and Patterns: An Introduction pdf/PLAW-104publ294.pdf/.
to Object-Oriented Analysis and Design and Iterative Develop- ■■ US Air Force. 2013. Global Horizons: United States Air Force
ment. 3rd ed. Westford, US-MA: Pearson Education. Global Science and Technology Vision. Washington, US-DC:
■■ Larson, W. J., D. Kirkpatrick, J. J. Sellers, L. D. Thomas, and US Air Force Office of the Chief Scientist.
D. Verma. 2009. Applied Space Systems Engineering. Boston, http://www.defenseinnovationmarketplace.mil/resources/Global-
US-MA: McGraw-Hill Companies, Inc. HorizonsFINALREPORT6-26-13.pdf/.
■■ Long, D., and Z. Scott. 2011. A Primer For Model-Based Sys- ■■ US Air Force. 2015. Statement of Objectives (SOO) for CREATE
tems Engineering. 2nd ed. Blacksburg, US-VA: Vitech Corpo- AV Assessment & Analysis Technology Development Broad Area
ration. Announcement (BAA) Call #3. Wright-Patterson Air Force
■■ Maccaba, B. 2014. “Why Software Doesn’t Follow Moore’s Base, US-OH: US Air Force Lifecycle Management Center.
Law.” Forbes. 19 May. http://www.forbes.com/sites/ciocen- https://www.fbo.gov/index?tab=documents&tabmode=form&subt-
tral/2014/05/19/why-software-doesnt-follow-moores-law/. ab=core&tabid=3d53de10d4bf5be5cedee2b3adb6b8ef/.
■■ Mackenna, A. 2012. “Using CREATE’s Rapid Ship Design ■■ Vanderperren, Y., and W. Dehaene. 2006. “From UML/SysML
Environment to Perform Design Space Exploration for a Ship to Matlab/Simulink: Current State and Future Perspectives.”
Design.” Paper presented at the NDIA Conference on Phys- Paper presented at the Design, Automation and Test in Europe
ics-Based Modeling in Design & Development for US Defense. (DATE) Conference. 6-10 March. Munich, DE.
7 November. Denver, US-CO: National Defense Industrial http://www.omgsysml.org/from_umlsysml_to_simulink.pdf/.
Association. http://www.dtic.mil/ndia/2012physics/Wednes- ■■ Zimmerman, P. 2014.. “Digital System Model Development
day14965_Mackenna.pdf/. and Technical Data.” Paper presented at the 17th Annual
■■ Meakin, R. L. 2011. “Multi-Disciplinary, Physics-Based NDIA Systems Engineering Conference. 30 October.
Simulation Software Products of the CREATE-AV Project.” Springfield, US-VA: National Defense Industrial Association.
Paper presented at the NDIA SE Division M&S Committee. 15 http://www.acq.osd.mil/se/briefs/16969-2014_10_30-NDIA-SEC-
February. Washington, US-DC: National Defense Industrial Zimmerman-DSM-vF.pdf/.

55