See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319751942

Introduction to quantitative engineering design methods via controls

engineering

Article in Artificial Intelligence for Engineering Design Analysis and Manufacturing · May 2017
DOI: 10.1017/S0890060417000415

CITATIONS READS

3 6,813

3 authors, including:

Matthew Adams Cameron J. Turner

Arizona State University Clemson University
5 PUBLICATIONS 5 CITATIONS 109 PUBLICATIONS 517 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Matthew Adams on 05 April 2018.

The user has requested enhancement of the downloaded file.

Artificial Intelligence for Engineering Design, Analysis and Manufacturing (2017), 31, 458–475.
# Cambridge University Press 2017 0890-0604/17
doi:10.1017/S0890060417000415

Introduction to quantitative engineering design methods via

controls engineering

BRIANA M. LUCERO,1 MATTHEW J. ADAMS,2 AND CAMERON J. TURNER3

1
Applied Engineering Technologies, Los Alamos National Laboratory, Los Alamos, New Mexico, USA
2
Ira A. Fulton Schools of Engineering, Arizona State University, Tempe, Arizona, USA
3
Department of Mechanical Engineering, Clemson University, Clemson, South Carolina, USA
(RECEIVED October 15, 2016; ACCEPTED May 27, 2017)

Abstract
Functional modeling is an effective method of depicting products in the design process. Using this approach, product ar-
chitecture, concept generation, and physical modeling all contribute to the design process to generate a result full of quality
and functionality. The functional basis approach provides taxonomy of uniform vocabulary to produce function structures
with consistent functions (verbs) and flows (nouns). Material and energy flows dominate function structures in the mechan-
ical engineering domain with only a small percentage including signal flows. Research suggests that the signal flow gap is
due to the requirement of “carrier” flows of either material or energy to transport the signals between functions. This re-
search suggests that incorporating controls engineering methodologies may increase the number of signal flows in function
structures. We show correlations between the functional modeling and controls engineering in four facets: schematic sim-
ilarities, performance matching through flows, mathematical function creation using bond graphs, and isomorphic matching
of the aforementioned characteristics allows for analogical solutions. Controls systems use block diagrams to represent the
sequential steps of the system. These block diagrams parallel the function structures of engineering design. Performance
metrics between the two domains can be complimentary when decomposed down to nondimensional engineering units.
Mathematical functions of the actions in controls systems can resemble the functional basis functions with bond graphs
by identifying characteristic behavior of the functions on the flows. Isomorphic matching, using the schematic diagrams,
produces analogies based upon similar functionality and target performance metrics. These four similarities bridge the me-
chanical and electrical domains via the controls domain. We provide concepts and contextualization for the methodology
using domain-agnostic examples. We conclude with suggestion of pathways forward for this preliminary research.
Keywords: Bond Graphs; Controls Engineering; Isomorphic Matching; Transfer Function

1. INTRODUCTION and experiences provide alternative solutions, but can fail

to convey the idea of functionality or application due to dif-
Engineering design practices the progression of an idea or
ferences of vocabulary, taxonomies, or methodologies.
conceptualization to a deliverable product. The premise is
Because there is merit to increasing design domains
that a specific function, with defined performance parame-
through analogies, this research provides quantitative options
ters, must be married with a form to produce a product
to incorporate controls engineering designs to the existing
(Pahl & Beitz, 1961). The delivered product expresses the
analogical approaches of mechanical engineering design
bounding requirements of the design by correlating the origi-
(Linsey et al., 2010, 2012; Lopez et al., 2011; Fu, Chan,
nal function to the performance of the final form through a
et al., 2013). The use of performance metrics and critical
prescribed cognitive process. Experience, from which crea-
functionality through a function structure in a design space
tivity stems, provides the limits of the design space (Lopez
may correlate transfer functions with mathematics to produce
et al., 2011). Suboptimal design may occur when engineers
analogies (Lucero, 2014; Lucero et al., 2016). This research
fixate on domain-specific definitions of functions and poten-
seeks to provide the initial frameworks of the matching tech-
tial solutions (Linsey et al., 2010). Interdisciplinary options
niques available to produce such analogies from control sys-
tems to engineering design.
Reprint requests to: Briana Lucero, Applied Engineering Technologies,
Los Alamos National Laboratory, MS-H821, P.O. Box 1663, Los Alamos, More specifically, this research attempts to show the fol-
NM 87545, USA. E-mail: [email protected] lowing:
458
Quantitative engineering design methods 459

1. A gap exists in functional representations of consumer library, encourages analogical design through biological
products for mechanical engineering students with signal sources (Goel & Bhatta, 2004; Goel et al., 2015). The structure,
flows. This deficiency of signal flows may relate to energy behavior, and function modeling language attempts to provide
and material flows acting as “carrier” flows for signals. a programming language capable of automating product teleol-
2. Usability of the schematic and numerical techniques al- ogy (Goel et al., 2009). However this automation seems to re-
ready employed by controls systems engineering with strict the complete abstraction of the designer. The graphical
transfer functions have similarities and applications to modeling language, Systems Modeling Language (SysML),
engineering design. By focusing on key performance has been incorporated into a behavior library at a high-level
parameters (KPPs) of a system, a translation to the engi- representation (Kruse et al., 2012, 2014; Kruse & Shea,
neering design domain is possible and can partially fill 2016). Yet many of these databases or repositories are indepen-
the signals gap. dently operated and do not necessarily support collaboration.
3. There are several similarities and correlations between Nagel (2007) and Nagel et al. (2008) made efforts to move
the two domains outlined herein that provide transla- the FB taxonomy to a formal functional basis modeling lan-
tions of ideas and product decompositions with the cur- guage. Focusing on the signal flows, it is suggested that using
rent domain practices: schematic similarities, perfor- a definitive grammar set improves communication between
mance matching through flows, mathematical function domains. While these rules are beneficial, they have yet to
creation using bond graphs, and isomorphic matching be enacted entirely with the FB taxonomy. The function-
of the aforementioned characteristics allows for analog- based systems engineering methodology implements func-
ical solutions. tional modeling through product planning, conceptual de-
sign, and embodiment design phases (Hutcheson et al.,
2007). Hutcheson et al.’s work defines a behavioral model
2. ENGINEERING DESIGN AND RELATED as a mathematical model of a system representing the ability
WORK to meet the requirements, complete with state variables. A
conceptual behavior model enacts a behavioral model, free
Kleer and Brown (1984) provide a devise-based ontology to of the final form, of the product solution. Both the behavioral
model the functionality of a system with black boxes of inputs models and conceptual behavior models tie back to the re-
and outputs. The function-behavior-state model developed by quirements for full system traceability, yet remains at the sys-
Gero (1990) and the function-environment-behavior-structure tems engineering level.
models (Deng et al., 1999, 2000; Deng, 2002) operate as Popular in systems engineering, a model-based systems en-
environment-centric viewpoints in function representations of gineering (MBSE) follows the standard systems engineering
the system. These models utilize functions as the effect an
design process. SysML supports the MBSE approach via nine
object has on the environment, and relate to the flows as the
diagrams with a standard taxonomy (Friedenthal et al., 2012,
conduits of this change. These approaches rely on the engineer
2015). The activity diagrams show the behavior of the
to provide the abstraction in the development of the system.
product with inputs, outputs, and controls in a sequential
When seeking analogies appropriate for the design pro-
fashion. The internal block diagram provides the interfaces
cess, it is not uncommon to search for previous products.
and connections between the parts of a block. Still, no single
Short of maintaining and establishing independent repositor-
ies, the function-behavior-physical-effect-structure (Qian & diagram quantitatively encompasses the flows of a model, nor
Gero, 1996) searches patents to provide analogies based on are there established relationships between the diagrams spe-
textual descriptions (Gero, 1990). The structure, being the fi- cifically for signals (Hampson, 2015).
nal configuration of the product, uses mathematical models Data flow diagrams are graphical representations of infor-
like physical effect to show the state transition. Patent search mation systems via the inputs and output flows of data (Bruza
tools circumvent the use of official repositories and have been & van der Weide, 1989). Also known as “bubble diagrams,”
researched as potential opportunities for analogies (Montec- data flow diagrams show the path of the data flow, but does
chi & Russo, 2011; Russo, 2012; Russo & Rizzi, 2014). By not show processes, their sequences, nor process timing
utilizing the patent database, matching via graphical mapping (Jayaram, 2002; Jayaram et al., 2003). These are powerful
is achievable and can pull from many domains. While these graphics showing the information only of the system, but
searches provide solutions from various sources, the applic- fail to show full functionality with the collective flows of
ability and functionality to the sourced target is left without the system. In addition, they have yet to be fully adopted in
context. mechanical engineering applications.
Design repositories established by various institutions and/ Several authors (Bracewell et al., 1993, 2001; Wu et al.,
or corporations have sought to capture product functionality 2008; Kypuros, 2013) suggest bond graphs as a means of pro-
and modeling. The National Institute of Standards and Tech- ducing analogies. Many authors have pointed out that bond
nology established the Core Product Model, encompassing graphs only apply to the energy flows and are not appropriate
the standard functional basis (FB) taxonomy (Fenves, 2002). for signal flow as bond graphs work solely for energy. Never-
The Design Study Library, a biologically inspired design theless, it is possible to circumvent this issue by using the
460 B.M. Lucero et al.

functionality of the bond graph components and adding non- verb–object combination correlates to the function-flow of
dimensional analysis to handle the signal flows. the individual functions to aide functional modeling methods
A previous attempt to apply functional decomposition to by utilizing a generic level of specificity and synonyms. This
mechatronics by Yuan and Ljung (2016) suggests that a com- research employs the FB as the cornerstone of functional
bination of bottom-up, physical effect, and top-down decom- modeling and assumes that the vocabulary is comprehensive
position approaches may be more effective than only a top- and adequate for all engineering disciplines.
down decomposition approach. This research focuses on
matching functions, but does not treat flows with the same at-
2.2. Domain specificity
tention. Cao et al. (2011) highlight the issues of SysML to
handle dynamic behavior or continuous dynamics in their Previous functional models maintain a focus on the representa-
models. In the same vein, Wu et al. (2008) show the strength tion of electromechanical consumer products. Many of the
of bond graphs in dynamic systems. available examples in the literature and on the Internet focus
While there have been outstanding efforts to incorporate on these small consumer products. The FB now caters to the
the ideas of functional modeling at a systems level with representation of these types of devices. Based upon a review
methods such as MBSE and SysML, complex domains of functional models by the authors (Lucero et al., 2014), a typ-
such as mechatronics underutilize the potentials of analogies ical model consists of approximately 10 flows and perhaps 20
in developing product solutions. The functional modeling of functions. Of these flows, approximately 90% represent energy
engineering design and the schematic representations of con- or material flows, while only 10% represent signal flows. A
trols systems proffer an opportunity to leverage the dissemi- sample taken by the authors of functional models from design
nation of information (via flows) mathematically to produce textbooks and the Oregon State Design Repository, encom-
analogical solutions. The efforts of this current research passing 40 functional models, indicates that 40% of flows
seek to provide a bridge between these two domains and es- represent material flows, 49% of flows represent energy flows,
tablish a framework for translation of information. and 11% represent signal flows. Similarly, the 20 models ex-
tracted from design textbooks included approximately twice
the number of unique flows as those that were present in the re-
2.1. Functional representation
pository, suggesting more thorough modeling of the flows from
Product representation of individual functionality and the in- the design textbooks than those within the repository. These
termediary processes assists in product decomposition. Pic- models showed a similar dominance of material (40%) and en-
torial and matrix representations of the product behavior are ergy flows (51%) in comparison to signal flows (9%). Neither
common in understanding the sequence of events and the per- textbook models nor those generated from design activities in
formance of the product. Function structures graphically de- the repository appear to include significant signal flows. In a
pict individual, sequential functions of the product in opera- reverse engineering context, these flows are often difficult to
tion. Disassembly of the overall functionality introduces a identify, but for forward design problems, signal flows are a
level of abstraction into the design process. The form of the critical element in the design of the system controller.
final product is then heavily dependent upon the overarching Signal flows are not independent of material or energy
performance required (Pahl & Beitz, 1961). To use the func- flows, as the signal flows require a carrier flow (in the form
tion structure process, the system functionality must be of an energy or material flow) to exist. Therefore, signal flows
broken down into individual purposes or subfunctions. Each can never be more than 50% of all flows. The low percentage
subfunction is a box containing a “verb þ noun” description of signal flows is also indicative of the products typically
representing the purpose with input and output flows. The in- modeled via this approach, and is likely the result of the clear
put and output flows are nouns in the form of an energy, ma- majority of available functional models representing reverse
terial, or signal wherein a flow acts upon as a verb. engineered systems. Most electromechanical consumer pro-
The input and output flows are a manner of accounting for ducts include limited instrumentation and control systems,
the physical system. Thermodynamically, the inputs must and therefore involve significant signal pathways, such as
equate to the outputs regardless of the action occurring in the that which may occur with a rudimentary control system.
function box. The material flows follow the zeroth law of ther- Yet many systems of interest from a modeling standpoint,
modynamics by again equating the input material to the output whether abstracted to a functional level, represented with a
material despite the function acting on them, such as mass. The bond graph model of dynamic behaviors, or represented
energy follows the first law of thermodynamics where all en- with a controls block diagram model, exhibit a much more ex-
ergy entering a system must exit it as well in the law of conser- haustive signal flow network representation than is typically
vation of energy (Sonntag et al., 2002). The signal flow does modeled functionally. A functional model often abstracts the
not flow a separate thermodynamic law, but rather is a combi- inner workings of a control system into a small set of highly
nation of laws zero and one and is not conserved (Pahl & Beitz, abstracted functions with limited signal modeling included.
1961; Sonntag et al., 2002). This is particularly true if the subject of the model is the result
Otto and Wood (2001) propose a taxonomy that links mul- of a reverse engineering exercise whereby the signals
tiple engineering disciplines and applications via the FB. The intended by the design are not necessarily obvious during
Quantitative engineering design methods 461

reverse engineering. Nonetheless, the result is an abstracted utzky, 2010). The edges are connections between two nodes
functional model with limited signal domain contributions. relaying information about the type of power transferred.
The resulting “signal gap” in functional models does pre- Since bond graphs are a method of translating system func-
sent a significant impact upon the design of systems. Without tionality with a defined set of components using power, the
a model of the signals involved in the system, the control sys- physical process of action on the power must occur with
tem for the design becomes a time-intensive post hoc process. two quantities: effort and flow. The effort and flow vari-
Coupling control system design with the physical system de- ables are domain dependent, but when multiplied together,
sign could reduce downstream modifications to the physical result in power in standard units. In addition to effort and
system and may dramatically improve the performance of flow, there are two other categories of use in bond graphs, mo-
the control system and the system. mentum and displacement, which categorize the type of sys-
tem represented. In dynamic systems, the energy changes
over time while using the momentum and displacement vari-
2.3. Design by analogy ables as the energy variables to account for the time fluctua-
Analogies can offer alternative solutions to design problems. tion. The combination of these variables allows for a specific
They provide engineers with analogical solutions based on a definition of the bond graph components as seen in Table 2.
linguistic or visual portrayal of the design problem and de-
scription. Visual analogies increase innovation of the design 3. CONTROLS ENGINEERING
solution for both novice and expert engineers (Casakin et al.,
1999). However, experimental results show that experts em- Controls engineering is a multidisciplinary engineering do-
ploy more analogies than novices employ and in a more crea- main encompassing such factions as mechanical, electrical,
tive fashion (Ball et al., 2004). Where the novices flock to and software engineering. The premise of controls engineering
analogies that closely resemble their current design problem, is to quantify the performance of a system and is common in
the experts draw from numerous examples that can offer par- robotic, computer visualization, mechatronic, and manufactur-
tial analogies and various design domains. ing applications (Mayr, 1970; Karayanakis, 1995; Dorf &
Because there is merit to increasing design domains Bishop, 1998; Doyle et al., 1990; Åström & Kumar, 2014).
through analogies, this research provides quantitative options Controls engineering is the basis for feedback theory and linear
to incorporate controls for engineering designs to the existing system analysis, melding together communications theory and
analogical approaches (Lopez et al., 2011; Linsey et al., 2012; network theory into a singular methodology to track and alter
Fu, Murphy, et al., 2013). The use of performance metrics system response. Control theory deigns performance objec-
and critical functionality to model a function structure in a de- tives are baselined via quantitative analysis by using these con-
sign space may correlate transfer functions with mathematics trols metrics in specifications and design definitions.
to produce analogies (Lucero, 2014; Lucero et al., 2016). This As with the typical engineering design process, the pre-
research seeks to provide the initial frameworks of the match- liminary steps of control theory start with an identification
ing techniques available to produce such analogies. of, and the means to measure, the overarching system’s func-
tionality. In the case of control theory, the “control” of the
identified parameter will be the focus of the remaining study,
2.4. Bond graphs usually with some level of accuracy entrenched in the pa-
rameter definition. With the specified accuracy, sensors
Bond graphs, like block diagrams, allow for pictorial repre-
monitor the parameter to measure the control variable. De-
sentations of systems, but also allow for dynamic scaling
pending on the definition of the problem, the final architec-
from component to full system via the energy in the system
ture of the control system (i.e., open or closed loop, robust-
as seen in Table 1 (Karnopp et al., 1990; Borutzky, 2010; Ky-
ness, signal sensitivities, etc.) can be considered.
puros, 2013). A pictorial representation of a graph with nodes
Open control-loop systems, such as extrusion-based addi-
and edges displays the system as a network. Each of the nodes
tive manufacturing, run through the desired process without
is a port of energy (dE/dt) or power exchange, called a “power
gauging the progress of the designed function. Thus, with ad-
port,” which acts as an intersection for the subsystems (Bor-
ditive manufacturing cases, a part can fail due to thermal or
structural stresses while continuing the build process in a cat-
Table 1. Control system through and across astrophic manner, resulting in the “rats’ nest” conundrum.
variables correlated to bond graph variables. “Open loop control systems employ an actuating device to
control the process directly without using feedback,” whereas
Control System Variable Bond Graph Variable “a closed loop system uses a measure of the output and feed-
Variable through element Flow ( f ) back of the signal to compare with the desired output (refer-
Integrated through variable Displacement (q) ence or command)” (Dorf & Bishop, 1998). Closed-loop sys-
Variable across element Effort (e) tems provide feedback as to the status of the operations and
Integrated across variable Momentum ( p) allow for alterations of some control to guide the desired
outcome. Examples of closed-loop systems are residential
462 B.M. Lucero et al.

Table 2. Bond graph components and their mathematical relationships to bond graph effort and flows

Bond Graph
Component Function Relation Mathematical Relation

Resistors Dissipate energy Directly relates effort to flow q ¼ FR ( f ) or f ¼ F21

R (e)
Capacitor Stores potential energy Directly relates effort to generalized displacement q ¼ FC (e) or e ¼ F21C (q)
Inertia Stores kinetic energy Directly relates momentum to flow p ¼ F1 ( f ) or f ¼ F21
1 (p)
Transformer Effort to effort, flow to flow Directly relates effort to effort, and flow to flow e1 ¼ ne2 or f1 ¼ nf2
Gyrator Effort to flow, flow to effort Directly relates effort to flow, and flow to effort e1 ¼ rf2 or rf1 ¼ e2

thermostats, Crockpots, or soldering irons as simple tools. flow. For example, current (dimensional analysis theory [DAT]
The efforts of this current research focus on the open-loop units; DAT L3T–1 ) is a through variable in an electrical system
system, but intentions to handle closed-loop systems exist. with transfer functions, whereas it is a flow in bond graph ter-
A control system shows the interfaces and interactions of in- minology. The across variable is the voltage potential (DAT
dividual components to produce a desired system response. units ML21T22 ), which is equivalent to a bond graph effort.
Analysis of the system response comes from the linear system When multiplied together, the voltage and current (or effort
theory where a system assumes a cause and effect relationship. and flow) produce power (DAT units ML–2T–3 ). Thus, the fo-
Like the black box models of engineering design, the cause and cal point of the system is the energy and power of the system, in
effect are synonymous with inputs, outputs, and processes. the same manner as bond graphs (Table 3).
There is a relationship established between the input and output
streams based upon the processes performed, providing the
cause and effect on the signal. The graphical representation 4. TRANSLATING TRANSFER FUNCTIONS
of control theory via the block diagrams is akin to the function TO ENGINEERING DESIGN
structures of engineering design by using transfer functions to This section provides the initial framework for connecting the
provide the mathematical representation of the processes. controls systems and engineering design domains as a cohe-
sive methodology. Utilizing the FB and the design-by-anal-
ogy framework, we suggest that four methods exist for en-
3.1. Transfer functions gineers to use in comparing designs:
Transfer functions represent the system variables and dy-
namic relationships of signals. For linear and linear approxi- 1. Schematic similarities: control system block diagrams
mations of systems, the transfer functions are the ratio of the  function structures of functional modeling
Laplace transform of the output variable to the Laplace trans- 2. Quantifiable performance metrics: control variables of
formation of the input variable where all initial conditions are control systems  nondimensional flows of the FB.
zero (Fig. 1). The Laplace transform translates the inputs/out- 3. Mathematical functions: Control system differential
puts from the time domain into the frequency domain through equations of transfer functions  FB flows as defined
a process H(s) shown in Eq. (1): by bond graph characteristics
4. Isomorphic matching: Identifies analogical alternatives
Y(s) by matching the DAT units and bond graph component
H(s) ¼ : (1) similarities
X(s)

The developed transfer function equation compares the in- We work through each step of the approach using an exam-
put parameter to the output parameter via through variables ple problem of a direct current (DC) motor. From the DC mo-
and across variables of individual components operating as tor example problem, we continue to additional examples that
a system (Borutzky, 2010; Kypuros, 2013). Like bond graphs are more scalable in the next section with more comprehen-
with efforts and flows, the variables define the ability to carry a sive discussions.

Fig. 1. Generic block diagram with system inputs and outputs, including mathematical representation.
 Quantitative engineering design methods 463

4.1. Schematic similarities

v × dpt
ML2 T22

ML2 T22

ML2 T22
F × dx,
V × dP
e × dq,

 t × du,
Energy

Energy

Energy

Energy
i × dl
Shared Nomenclature

(E)


This work previously identified gaps in the signal flows of


the FB in functional decomposition. While present in
many systems and products, these flows rarely appear in

F(t) × V(t)

t(t)×v (t)
e(t) × i(t)

ML2 T23

ML2 T23

ML2 T23
Power

Power

Power
Power

the function structures developed by mechanical engineer-

(P)

ing students. Exploiting the field of controls engineering

may increase the appearance of signal flows in function
structures. The block diagrams utilized in control systems
Displacement
proffer similar graphical representation of functional model-
L2 MT21 Q21
Flux linkage
Momentum

difference

Angle
ing with verb þ noun combinations of the FB in function
( p)

X
l

8
structures in the block diagrams. Since controls engineering
does not use a specified taxonomy, there is not a perfect one-
to-one translation between the two domains. However, the
block diagrams utilize mathematical functions, via differen-
L2 MT22 Q21

difference

tial equations, which show the same cause–effect relation-

Velocity

Angular
velocity
Voltage
Bond Graph Nomenclature
Effort

LT21

8T21
(e)

V ships between the control variables and verbs (actions) as

e

seen in the FB.

Controls engineering relies heavily on the reduction of
Table 3. Control system and bond graph nomenclature comparison converging on the shared power and energy terms

block diagrams to simplified diagrams based upon feedback

Displacement

Translational

techniques (Fig. 2). This research does not establish relation-

momentum

momentum
L22 M2 Q2

ML2 T21
Angular
MLT21
Charge

ships to the reduction or simplification techniques, but in-

(q)

P
q

stead proposes that the graphical similarities are still applic-

able regardless of the block diagram complexities. Using
the blocks as the intersections of functions, the interconnect-
ing flows are traced though the system with emphasis placed
ML21 T22

on the signal flows, regardless of the system complexity.

Current

MLT22

Torque
Force
QT21
Flow
(f)

These schematic similarities are akin to the work outlined

i

by Gentner (Gentner & Gentner, 1982; Gentner, 1983),

where the use of conceptual relations or specific operations
can span across various domains to describe functionality.
Displacement
L2 MT21 Q21
Flux linkage

For example, Gentner and Gentner (1982) found electricity

difference
Integrated

Variable
Across

Angle

to be analogous to “flowing” mechanisms such as flow within

X

L
l

pipes. While the flow of electrons is not exactly like the flow
of a water molecule in a pipe, it is an analogy between two
complex systems, describing how to achieve some action.
Thus, the relational structure of how an action is similar yet
Control System Nomenclature

ML2 T22 Q21

difference
Velocity
Variable

Element

Angular
velocity
Voltage

unbound by the object or domain of origin.

Across

LT21

8T21
v
V
e

The block diagrams of controls engineering allow for se-

quential mathematical relationships to be developed based
upon this cause-effect method, which is similar in sequential
product depiction to the functional modeling. The how por-
Translational
momentum

momentum
L22 M21 Q2

tion of an action is depicted in the block diagram mathemati-

Integrated
Through
Variable

ML2 T21
Angular
MLT21
Charge

cally while the function structure shows the systems-level

pt
P
q

analysis of what is done linguistically. Thus, there is no direct

translation between the two domains, but parallels exist from
the graphical representations of the product system. Note that
controls engineering can still provide how the process occurs.
ML21 T22
Through
Variable

Element

Current

MLT22

Torque
Force
QT21

For example, monitoring the amplification signal from an op-

F

t
i

amp or Wheat-stone bridge circuit conveys how much ampli-

fication is occurring in the system.
The block diagrams consist of unidirectional, operational
translational
Mechanical

Mechanical

blocks representing the transfer function on the input and output

Electrical

rotational
System

variables. Block diagrams have the cause–effect relationship al-

lowing for the tracking of specific control variables, such as
those of the signal flows of the FB. Like function structures,
464 B.M. Lucero et al.

Fig. 2. Control engineering system depictions: (a) black box model of process to be controlled, (b) open-loop control system with no
feedback, and (c) closed-loop feedback control system to provide actual output.

the operational blocks can have multiple inputs and output for a entire system, and scalability of models through the set num-
multivariable control system. Likewise, actions occurring on the ber of relevant control variables.
flows result in the cause and effect mentioned previously. Previous research by Coatanéa (2005) and Lucero et al.
(2016) has established the use of the DAT parameters for en-
E XAMPLE . A DC motor is a power actuator device, de-
gineering design using flow parameters. By correlating the
livering energy to a load as resistively shown in Figure 3a
flows to these DAT parameters, it is possible to match the
and as physically sketched in Figure 3b. The DC motor “con-
nondimensional units based on the exponential power of
verts” DC electrical energy to rotational mechanical energy.
the dimensionless units present. This change therefore can
The armature generates the torque of the system (via the rotor
be accounted for via a transfer function that compares the out-
windings and stator windings), which drives an external load.
put to input parameters to identify which action was per-
DC motors have many applications in disciplines such as
formed on the flows by equating the exponential power of
power systems, robotics, and mechatronics. The example
the input and output flows. For example, integrating angular
given is adapted from Dorf and Bishop (1998). B
acceleration (DAT units T–2 ) with respect to time (DAT
units T), results in angular velocity (DAT units T–1 ).
If using the power (DAT units ML2 L–3 ) or energy (DAT
4.2. Quantifiable performance metrics
units ML2 L–2 ) of a system as the performance parameter, it
becomes easy to translate across domains, as these parameters
All engineering design problems revolve around KPPs,
do not alter. The method of obtaining the power might be dif-
which are the overarching target functionality of the system
ferent per domain, but the result of the parameter is the same
or product. Previous work in quantifying the performance
with DAT units ML2 L–3 . For example, multiplying voltage
of these KPPs, and mapping them to flows of the FB, has es-
by current equates to electrical power in the electrical domain
tablished frameworks in technical engineering variables with
while in the thermal heat transfer domain, the temperature
corresponding units (Lucero et al., 2016). The FB flows are
multiplied by the entropy flow rate defines thermal power.
characterizations of their respective material, energy, or sig-
This manner of comparing exponents for the power and en-
nal flow, which can be further decomposed into their second-
ergy parameters allows for the transition between domains
ary levels showing more specificity, allowing for the estab-
via bond graph variable combinations for power and energy
lishment of the individual flows capable of measurement
variables as shown in Table 4 for selective domains.
via sensors. For example, a material flow for a fluid physical
system could be the volumetric flow rate (DAT units L3T–1 ) E XAMPLE . When employing the FB, the statement of work
of some working fluid as measured via flowmeters such as for the DC motor is “convert electrical energy to mechanical
pitot-tubes or ultrasonic Doppler flow meters. (rotational) energy” as shown in Figure 4a. The input into
The DAT provides a relationship between physical quanti- the system is electrical energy in the form of voltage (DAT
ties as defined by their fundamental dimensions of length, units ML–1T–2 ), while the output is mechanical (rotational) en-
mass, time, charge, temperature, and luminosity (White, ergy via torque (t; DAT units ML2T–2 ) and rotational velocity
2003). Through combinations of these base units, it is possi- (v; DAT units of degree T–1 ), waste energy in the form of
ble to produce almost any engineering flow, but with a re- Watts (DAT units ML2T–3 ), and rotational position (degrees
duced number of variables. This dimensional homogeneity u). The function structure of the KPP remains at a systems-level
allows for: the reduction of the number of variables in a sys- depiction where the mechanisms by which the conversion of
tem, fundamental modeling of physical relationships of the the energy is ambiguous. The black box representation merely
Quantitative engineering design methods 465

Fig. 3. (a) Electrical schematic of a motor for voltage applied across the field and (b) physical representation of the electrical schematic.

looks at the inputs to outputs and allows the DAT units to be the can seek mathematical relationships using transfer functions
performance parameters quantified in the system. B as a ratio of the output to input flows via Laplacian transforms
[Eq. (1)]. There can be many mathematical possibilities for
However, the controls block diagram for the DC motor each of the bond graph components, but this method estab-
(Fig. 4b) with the method of conversion defined, established a lishes the mathematical baselines.
transfer function for this specific system wherein the position is From these characteristic baselines of each component
the signal to be monitored. A linear approximation of the actual then, it is possible to quantify the action modeled. For exam-
motor develops from the input voltage to the output position, ple, a regulate function, acts a resistive element that can show
while neglecting second-order effects (i.e., hysteresis and voltage the level of efficiency across the input and output streams.
drop across the brushes). Thus, using sequential differential equa- This behavior provides quantitative bounds in performance
tions, the final positional displacement (degrees u) is calculated. metric analogy searches. In so classifying the behavior of
the resistive element, it is possible to look at any of the sim-
ilarly classified functions as a starting point. Thus, the math-
ematical relationships between the inputs, outputs, and ac-
4.3. Mathematical functions
tions ensuing give additional opportunities to search for
Functions, within the FB, are the performance of an action on analogies based upon similar functionality.
the flows. These actions correlate to the active characteristics Ideally, a repository of the all the mathematical functions as-
of bond graph components and can be thus categorized per sociated with each bond graph component would be estab-
Table 4. With the five bond graph components, engineers lished. For algorithmic analogy programs, such as D-APPS
466 B.M. Lucero et al.

Table 4. Functional basis functions categorized into bond graph components based upon functionality characteristics

Resistors Capacitor Inertial Transformer Gyrator

Branch Position Provision Channel Change Actuate

Change Process Store Export Control magnitude Control magnitude
Channel Regulate Guide Convert Convert
Connect Secure Import Stop
Control magnitude Sense Indicate
Couple Separate Position
Distribute Signal Process
Export Stabilize Provision
Guide Stop Sense
Import Supply Signal
Indicate Support Stabilize
Mix Transfer Supply
Support
Transfer

(Lucero, 2014), it would be possible to have a database of tar- notation of Eq. (3), where the constant armature current (ia )
get analogies that exhibit mathematical relationships similar to is equivalent to Ia .
those of the source (transfer function). More specifically, the
transfer functions provide a quantitative solution which can
bound the range of the performance metrics sought (i.e., Tm (s) ¼ ðK1 Kf Ia ÞIf (s) ¼ Km If (s): (3)
“seek a resistive element that only allows 1808 of rotation).
Equation (4) defines the input voltage of the field as a func-
E XAMPLE . In the DC motor control system, the perfor- tion of the motor inductance and resistance.
mance parameters are the voltage, torque, and rotational ve-
locity. The field terminals have an applied constant current to Vf (s) ¼ (Rf þ Lf s)If (s): (4)
operate. Assuming a linear relationship with the torque and
the air gap flux of the motor as shown in Eq. (2). The motor torque [Tm (s)] of Eq. (5) is the total torque de-
livered to the load as functions of the load torque [TL (s)] and
Tm ¼ K1 Kf if (t)ia (t), (2) the disturbance torque [Td (s)]. It is common to assume Td (s)
is negligible and to define TL (s) as shown in Eq. (6).
where if (t) is the field current and K represents various con-
stants. B
Tm (s) ¼ TL (s)R þ Td (s), (5)
The field current motor provides power amplification,
which allows Eq. (2) to be written in the Laplace transform TL (s) ¼ Js2 u(s) þ bsu(s): (6)

Fig. 4. (a) Direct current motor function structure per the functional basis and (b) controls system block diagram for the energy flows as
outlines in (a).
Quantitative engineering design methods 467

For TL (s) and Tm (s) in Eqs. (7)–(8), rearrange Eqs. (3)–(6) representation provides the foundation for the mathematical
and introduce If (s) as Eq. (9). functions and quantification of the performance parameters.
Isomorphic matching compares the nodes and edges of
TL (s) ¼ Tm (s) þ Td (s), (7) the source against a target. D-APPS employs this iso-
morphic matching to determine how similar a target func-
Tm (s) ¼ Km If (s), (8) tion structure is to various analogous functions structures
(Lucero, 2014). By incorporating the control systems block
Vf (s) diagrams into a graphical design space, tools such as D-
If (s) ¼ : (9) APPS can perform the isomorphic matching to determine
Rf þ Lf s
how close potential solutions can be to the target problem.
Equation (10) relates rotation to voltage via the transfer Consequently, the matching efforts become a graphing
function of the output over the input. problem wherein the various nodal and vertex combinations
are the basis for combinations and permutations capable of
u(s) Km Vf (s) analysis.
¼ If (s) ¼ : (10)
Vf (s) s(Js þ b)(Lf s þ Rf ) Rf þ L f s After review of the function and bond graph characteriza-
tions for many systems, it may be possible to build a
This final transfer functions equation relates back to the FB repository quantifying the actions of the functions through
through the performance metrics and the bond graphs. The the functional categories outlined in Table 4. Similarly,
flow streams match to DAT unit parameters, as modeled by mapping the dimensionless flows of the FB to the transfer
the input/output streams. Once identified, these flows cross function intermediary variables enables comparison of the
the system boundary as the transfer function of Eq. (10) actions in the black box on the exponential units. Opportuni-
and match to the convert function of the FB to provide char- ties are now available to seek analogies within specific signal
acterization of the action. Note that the bond graph compo- flows using both functionality and quantifiable performance
nents characterize the action (Table 4) and thus allow for metrics.
any gyrator component functions to be considered.
E XAMPLE . The DC motor signal flow identified as a KPP
4.4. Isomorphic matching was the rotational position. Building a graph of the convert
function and the other flows, the figure in Figure 5a is possi-
As the schematic similarities (Section 4.1) outlined, the graph- ble with the DAT versions of each performance parameter.
ical representations of the control-system block diagrams Referring to the bond graph component groupings (Table 4),
and those of the function structures offer opportunities to tra- the convert function acts like a gyrator wherein an effort is
verse engineering domains through pictorial representations converted to a flow or vice versa. Per this grouping, the
of system functionality. However, the analogy identification convert and control magnitude functions have similar func-
occurs in this current step where computer matching can iden- tionality and can be possible analogies. B
tify analogical solutions with similar functionality and
comparable performance metrics (via DAT). Using graph A gyroscope is a component commonly used in aerospace
theory as outlined in D-APPS (Lucero, 2014), the functions applications in attitude determination and control efforts. It is
(mathematical or verbs alike) become the nodes of a design a device of a wheel or a disk mounted in gimbals, rapidly
space, while the flows (DAT units) become the connections spinning about an axis leveraging the angular momentum
between the nodes (Gould, 2014). Thus, the functions are of this spin to alter or maintain the direction of the tilt
equivalent to nodes in a design space with interconnections (Mcbride & Cellier, 2001). The control magnitude function
of edges (or vertexes). The edges, with their exponential then is the critical function acting to alter the torque to rota-
DAT units of the performance parameter, can then relay tional velocity and position (Fig. 5b). Since both the DC
upon the mathematical functions (Section 4.3) to find similar- motor and the gyroscope signal flow of rotational position
ities to other functions or bond graph component types. are the output parameter, the analogy can be drawn as
The engineering performance parameters can traverse do- Figure 5c, where the difference between the input flows is
mains by matching the DAT exponents. Treating the flows identified as the delta of the exponential components.
with their dimensionless units as the inputs and outputs of a This correlation of matching is done using critical func-
black box, the function, or action, enacted upon the flows pro- tionality (bond graph component grouping) and performance
vides the mathematical operation by numerically comparing parameters (DAT). With the transfer function established
the input to output unit exponents. These actions become from the DC motor, it would be further possible to quantify
the ratio of the output to input of transfer functions and the extent of the rotational position achieved with this specific
produce a mathematical relationship for each specific flow example. Thus, we are not suggesting that a gyroscope is a re-
conversion. Thus, if laid in the graphical representation of placement for a DC motor, but that the method of sensing the
graph theory, the functions are the nodes, while the flows rotational position has parallels that need further investigation
are the edges or interconnections between the nodes. This from the engineer.
468 B.M. Lucero et al.

Fig. 5. (a) Simplified graph of the direct current motor with performance parameters, (b) graph of a gyroscope, and (c) comparison graph of
the direct current motor and the gyroscope, showing potential analogy identification. Working back from the rotational position, the gyro-
scope functions.

5. EXAMPLE PROBLEMS The function structure displays the signal flows, in orange
arrows, indicating the controlling paths of the temperature
The following two examples demonstrate the ability to traverse
and the glue flow rate (Fig. 6). The KPP of the system, “main-
domains using the methodology discussed previously. Dem-
tain the temperature of the glue gun,” leads to the identification
onstrated with different levels of complexity and process trans-
of two critical pairs: convert electrical energy to heat energy
parency, the two example problems include their function
and regulate melted glue flow. Sensors control the energy
structures, block diagrams, and the transfer functions with sug-
and material flows in the system through control loops and al-
gested analogies. The first problem works through a hot glue
low for the signals flows to be highlighted in this example.
gun, where the key functionality is to extrude melted glue at
The control block diagram of the glue gun lays out the pro-
a controlled flow rate with basic transfer functions. The second
cess by which the energy and material is sensed, and con-
problem shows the applications of a more complex system with
trolled by the glue gun electricity regulator. However, the
a biological sensor application by providing the basis for the
control system does not perform the entire systems analysis
graphical similarities and quantification. Each example is sim-
like the functional modeling but instead focuses on the signal
ply a proof-of-concept validation, and much like engineering
mechanisms. In this manner, the control block diagram shows
design, proffers a single solution among many.
similar graphical characteristics as the function structure, but
is only applicable to the signal mechanism by providing the
5.1. Hot glue gun “how” the heating is controlled. The temperature of the
glue gun can be decomposed into nondimensional tempera-
A hot glue gun operates by melting glue sticks via a heating ture units (u) and traced through the closed-loop system.
element and extruding the melted glue through the nozzle The heat addition to the system is
through the actuation of a trigger. The heating element uses
electricity and a voltage regulator to control the amount of
heat added to the system as required per the material charac- Q(s) 1
¼ , (11)
teristics of the glue sticks, which is dependent upon the q(s) Ct  s þ ðQ  S þ 1=Rt Þ
melting point of the sticks. The glue gun trigger controls
the flow rate of the melted glue as actuated from the hand-
held force applied. Releasing the trigger halts the extrusion, where Q is Qgun – Q0 , which is the temperature difference due
but the thermal regulator maintains the required temperature to the thermal process; Ct is the thermal capacitance; Q is the
to keep the stick melted. This example emphasizes the signal fluid flow rate or constant; S is the specific heat of air; Rt is the
controls process for regulating the temperature through the thermal resistance; and q(s) is the transform of rate of heat
conversion of electrical energy to heat energy. flow of heating element.
Quantitative engineering design methods 469

Fig. 6. (a) Function structure of hot glue gun, and (b) control system block diagram of glue gun operation for a closed-loop system.

Assuming an addition of unit step q(s) ¼ 1/s to alter the Using the KPP of the hot glue gun, the flows can be
temperature, the response of the system becomes nondimensionalized for both the electrical energy and the
temperature, providing the correlation between the function
1=Ct 1 b 1 b=a b=a structure and control system diagrams as seen in Figure 7.
t(s) ¼  ¼  ¼ þ , (12)
QS þ 1=R s s þ a s sþa s The transfer function allows for the quantification of these flows
sþ
Ct as performance metrics for the system while the function struc-
ture provides the linguistic representation of the system. In this
where a ¼ (QS þ 1/R)/Ct and b ¼ 1/Ct .
fashion, the performance metrics and the critical function con-
Taking the inverse Laplace transform yields
vert represented in tandem with bond graphs to provide alterna-
b at b b tive approaches to altering electrical energy to a temperature.
t(t) ¼ e þ ¼ ½1  eat : (13) As an analogy example, consider a hair blow dryer (Fig. 8).
a a a
The glue gun isomorphic graph (Fig. 8a) shows the regulation
Assuming steady state conditions where the temperature of of electrical energy to a temperature through a control mag-
the glue gun remains constant for melting the glue stick once nitude function. By comparison, the blow dryer regulates the
the desired temperature is reached, the time approaches infin- temperature and air flow rate. However, for this example, we
ity (t ! 1) leading to are concerned with the temperature only and how the tem-
perature is sensed to generate feedback within the system. Be-
b 1 cause both the control magnitude and regulate functions are
t(t) ! ¼ : (14)
a Qs þ 1=R considered resistive bond graph elements, it is possible to
470 B.M. Lucero et al.

Fig. 7. Key performance parameters for (a) a function structure of hot glue gun and (b) a control system block diagram of glue gun opera-
tion for a closed-loop system.

consider the output performance parameter of temperature as approximately 50 ms for a 108 saccade, the human eye mus-
a potential functional analogy for the system (Fig. 8c). cles are among the fastest reacting muscles in the human body
Much like the DC motor example presented previous, the (Enderle & Bronzino, 2012). These movements are useful
premise behind analogy mapping is not a one-to-one match when the eyes are locating a target using accurate, but jerky,
providing “how” to utilize the analogy. Instead, this lays motion and without care of the information moving across the
out a method for potential analogies based upon the perfor- retina during the movement. A saccade turns the visual sys-
mance metrics and functionality. How the analogy will be ap- tem off until it is complete. After the saccade, the system op-
plicable to the design engineers is up to them consider the erates in a closed-loop mode to verify that the target is now in
context of their problem and what portions of any analogy view. Use of information from the retinal view of the new
could be useful to them. scene and muscle proprioceptors aides the correction for
any errors between the current and desired eye position.
A typical experimental protocol for observing saccades has
5.2. Saccadic eye movement sensors
the subject seated before a light-emitting diode (LED) target
A saccade is a jerky movement of the human eye to adjust its display. The subject must maintain focus on the lit LED,
focus from one target to another. With reaction times of move their eyes as fast as possible, and avoid false tracking.

Fig. 8. Potential analogy for sensing the temperature of (a) a glue gun and (b) a hair dryer using their key performance parameters. (c) The
analogy can be achieved via isomorphic matching of the bond graph component (resistive) and the exponential values of the nondimen-
sionalized units.
Quantitative engineering design methods 471

The experimenter switches power from one LED to another

LED, which changes the optical signal. This optical signal
converts to electrochemical neural responses in the brain.
The brain must fire the muscles attached to the eyeball by con-
verting the electrochemical neural signal into mechanical ac-
tivation of the appropriate muscles. The mechanical contrac-
tion of the muscle of the eye results in a rotation of the eyeball,
resulting in a saccade, which is the process of focus in this ex-
ample.
From the vantage point of the experimenter, the objective is
to measure the speed and reaction time of the human eye with
respect to a change in the stimulus signal as a position. The
KPP of the system is then “measure the position of the eye after Fig. 10. Example light-emitting diode board to provoke a horizontal saccade
the stimulus triggers.” Therefore, the key flows to track are the eye movement.
eye itself (Fig. 9a), the object providing the stimulus signal
(Fig. 10), and the device(s) used to track the movement of
placement (Westheimer, 1954). The Laplace variable analy-
the eye (Fig. 11a,b). While each of these flows can decompose
sis provides the mathematical expression
into its own independent function structure, it is important for
the experimenter to define the function of the protocol for the
experiment (Fig. 12). For the experimenter, the objective is to J ü þ Bu̇ þ K u ¼ t(t): (15)
collect data for comparison to his or her model (Fig. 9b) of hu-
man eye motion. The experimenter needs to account for the Assuming 08 for initial eye orientation, the standard form
function of the experimental apparatus and whether it will pro- of the transfer function for Eq. (15) is
vide the data for comparison to the experimenter’s equations
for motion [Eq. (15)–(21)] of the eyeball. Y(s) u 1 v2 (s2 þ 2zvn s þ v2n )
H(s) ¼ ¼ ¼ 2 ¼ n ,
In 1954, Westheimer published an eye model (Fig. 9b) for X(s) t Js þ Bs þ ðKÞ K
describing horizontal saccades in response to a 208 target dis- (16)
and for a 208 saccade,
rffiffiffiffi
K
vn ¼ 120, (17)
J

B
z ¼ pffiffiffiffiffiffi ¼ 0:7, (18)
2 KJ

where B is the frictional rotational element, J is the moment of

inertia, and K is the stiffness.
The torque, t(t), generated by the lateral and medial rectus
muscles evokes the saccade and a new eye position, u(t). Per
Westheimer’s data (Westheimer, 1954), the roots of the trans-
fer function are complex and given as
qffiffiffiffiffiffiffiffiffiffiffiffiffi
s1,2 ¼ zvn + jvn 1  z2 ¼ 84 + j85:7: (19)

Assuming a step input for the system, t(s) ¼ g=s, the new
eye orientation, u(t), becomes
  qffiffiffiffiffiffiffiffiffiffiffiffiffi 
g ezvn t
u(t) ¼ 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffi cos vn 1  z2 t þ c , (20)
K 1  z2

where

z
Fig. 9. (a) Muscle anatomy of the ocular motor system and (b) the corre- c ¼ p þ tan1 pffiffiffiffiffiffiffiffiffiffiffiffiffi : (21)
sponding Westheimer second-order mechanical eye model. 1  z2
472 B.M. Lucero et al.

Fig. 11. (a) Depicted for the right eye, a scleral search coil measures eye movement via coils embedded into either a fitted contact lens or a
rubber ring that adheres to the eye. Magnetic fields, from magnets around the eye, generate electric currents in the search coils. By mea-
suring the variations in polarity and amplitude of the current generated from the angular displacement of the eye, the position of the eye can
be determined. (b) Fixed infrared light emitter(s), directed at the eye, will reflect an amount of infrared light to the fixed receiver(s), which
will vary per the eye’s position.

This experiment, when broached from the engineering de- ent lighting level of most facilities does not affect measure-
sign perspective, allows for tracking the saccade in multiple ments because infrared detectors do not detect these light
ways. Two example methods of eye motion tracking are the sources. The input of this system is the position of the eyeball
employment of a scleral (Fig. 11a) search coil and infrared and the output is the change in the infrared radiation detected.
oculography (Fig. 11b). These motion-tracking methods We can use DAT and bond graphs to construct the relation-
share the same function structure for understanding human ship between these two methods. In terms of bond graphs,
ocular mechanics work, but leave the how the tracking is both systems relate the change in the eyeball angle through
done up to the designer. DAT, with bond graph theory, makes a flux linkage. The movement of the eyeball changes the po-
it possible to compare the search coil method to the oculogra- sition and orientation of the scleral coil, resulting in a flux in
phy method as they have equivalent functionality (measure) the magnetic field. Similarly, the change in the position and
and KPP (eye rotational position). orientation of the eyeball produces a flux in the infrared light
A scleral search coil measures eye movement via coils em- detected. Thus, the two solutions presented in Figure 12 are
bedded into either a fitted contact lens or a rubber ring that ad- functionally analogous in their approach to measure the eye
heres to the eye where a wire leaves the eye at the temporal saccade, but vary solely in implementation.
canthus (Fig. 11a). For horizontal saccade experiments, two
inducting coils placed on either side of the head produce mag-
6. CONCLUSIONS
netic fields. During saccade eye movement, the scleral coil
will move, which causes fluctuations in the magnetic fields. Engineering design has sought to produce practices that allow
The new orientation of the eye is determined by measuring for abstraction in the final design. Functional modeling
the variations in the polarity and amplitude of the magnetic methods, such as the FB, establish uniformity of the cognition
field caused by the angular displacement of the scleral coil. process for the mechanical domain with material and energy
The input is the position of the eyeball and the output is the flows. However, there is a deficit in the use and approach to
change in the magnetic field. the signal flows as evidence by the teaching of functional
Infrared oculography uses a fixed infrared light source di- modeling to mechanical engineering students using consu-
rected at the eye (Fig. 11b). The amount of infrared light re- mer products.
flected to a fixed detector varies with the eye’s position. The Because of this gap, the research herein attempted to find
infrared transmitters and receivers mount to frames for eye- approaches upon which methods and processes from other
glasses, and as infrared light is invisible to the eye, it does engineering domains could provide more examples. The
not serve as a distraction to the subject. In addition, the ambi- authors suggest utilizing controls engineering to bridge the
Quantitative engineering design methods 473

Fig. 12. Example function structure for the saccadic eye movement sensor.

gap to provide both similar graphical representations and ioral functionality. The five components of bond graphs
mathematical potentials. Comparing the block diagram used (resistive, capacitive, inductive, transformer, and gyra-
to track and “control” the variables of the system to the func- tor) have specific behavioral characteristics defined by
tion structures of the FB identifies functional similarities. Es- the interactions of the performance parameters. When
tablishing linear, differential equations for products is possi- seeking functional analogies, these characteristic groups
ble when concurrently using the block diagrams and the are the basis for the isomorphic matching.
relationships between the inputs, outputs, and actions of the † Design-by-analogy options are available to fill the sig-
system. nal flow gap by utilizing the bond graphs and block
Some of the outcomes of this research include the following: diagrams of control systems via isomorphic matching.
Treating the functions (via the proposed bond graph
† Current functional modeling pedagogy lacks robust exam- component grouping) as the nodes and the flows (via
ples of signal flows when looking at functional representa- the DAT unit exponentials) as the edges, comparisons
tions of consumer products with mechanical engineering between individual components is possible for analogi-
students. This deficiency in signal flows may relate to en- cal similarities.
ergy and material flows carrying the signal flows. † Applying the proposed methodology to problems of
† Schematic similarities between functional modeling various complexities and scales is feasible due to the
techniques, particularly the function structures, and the use of DAT. Three example products represent the do-
control-system block diagrams allow for translations be- main agnostic application of the developed theories
tween domains. The graphical representation of the sys- and provide evidence of this methodology.
tem models are not identical matches, but functions and
flows correlate through the energy and power variables This research is in the preliminary stages and requires fur-
of bond graphs. ther verification and validation. Additional analysis from stu-
† Quantifying the KPPs of a system is possible for system dent populations will provide the applied basis for these
performance metrics using DAT, such as Buckingham- processes. It is necessary to develop this concept of relating
Pi. The energy and power variables of bond graphs allow controls engineering with engineering design based upon ad-
the translation of the DAT units of the FB flows and the ditional data collection, repository comparisons, and example
control variables of control systems. problems. This framework may allow inclusion of analogies
† The functions of the FB equate to transfer functions of stemming from more engineering domains and not just those
control systems through bond graph component behav- of mechanical or electrical domains.
474 B.M. Lucero et al.

REFERENCES Goel, A.K., Zhang, G., Wiltgen, B., Zhang, Y., Vattam, S., & Yen, J. (2015).
On the benefits of digital libraries of case studies of analogical design:
Åström, K.J., & Kumar, P.R. (2014). Control: a perspective. Automatica documentation, access, analysis, and learning. Artificial Intelligence for
50(1), 3–43. Engineering Design, Analysis and Manufacturing 29(2), 215–227.
Ball, L.J., Ormerod, T.C., & Morley, N.J. (2004). Spontaneous analogising in Hampson, K. (2015). ScienceDirect technical evaluation of the systems mod-
engineering design: a comparative analysis of experts and novices. eling language (SysML). Procedia Computer Science 44, 403–412.
Design Studies 25(5), 495–508. Hutcheson, R.S., Ryan, S., McAdams, D., Stone, R., & Tumer, I. (2007).
Borutzky, W. (2010). Bond Graph Methodology: Development and Analysis Function-based systems engineering (Fuse). Proc. Int. Conf. Engineering
of Multidisciplinary Dynamic System Models. London: Springer. Design, pp. 1–12, Paris, August 28–31.
Bracewell, R.H., Shea, K., Langdon, P., Blessing, L., & Clarkson, P. (2001). Jayaram, M.B. (2002). A Formal Method for Functional Modeling and Con-
A methodology for computational cognitive modelling. Proc. Int. Conf. ceptual Design of Complex Mechantronix Systems. Toronto: University
Engineering Design, pp. 181–188. Glasgow: Design Society. of Toronto Press.
Bracewell, R.H., Bradley, D.A., Chaplin, R.V., Langdon, P., & Sharpe, J.E.E. Jayaram, M., Chen, L., & Xi, F. (2003). Functional modeling of complex me-
(1993). Schemebuilder, a design aid for the conceptual stages of product chatronic systems. Proc. ASME 2003 Int. Design Engineering Technical
design. Proc. Int. Conf. Engineering Design. Zurich: Design Society. Conf. Computers and Information in Engineering Conf., pp. 413–422,
Bruza, P.D., & van der Weide, T.P. (1989). The semantics of data flow Chicago, September 2–6.
diagrams. Proc. Int. Conf. Management of Data (Prakash, I.N., Ed.), Karayanakis, N. (1995). Advanced System Modelling and Simulation With
pp. 66–78, Hyderabad, India. Block Diagram Languages. Boca Raton, FL: CRC Press.
Cao, Y., Liu, Y., & Paredis, C.J.J. (2011). System-level model integration of Karnopp, D., Margolis, D., & Rosenberg, R. (1990). System Dynamics: A
design and simulation for mechatronic systems based on SysML. Mecha- Unified Approach. New York: Wiley.
tronics 21(6), 1063–1075. Kruse, B., Gilz, T., Shea, K., & Eigner, M. (2014). Systematic comparison of
Casakin, H., Goldschmidt, G., & Planning, T. (1999). Expertise and the use functional models in SysML for design library evaluation. Procedia
of visual analogy: implications for design education. Design Studies CIRP 21, 34–39.
20(2), 153–175. Kruse, B., Munzer, C., Wolkl, S., Canedo, A., & Shea, K., (2012). A model-
Coatanéa, E. (2005). Conceptual modelling of life cycle design: a modelling based functional modeling and library approach for mechatronic systems
and evaluation method based on analogies and dimensionless numbers. in SysML. Proc. 32nd Computers and Information in Engineering Conf.,
PhD Thesis. Laboratory of Machine Design, Helsinki University of Parts A and B, Vol. 2., pp. 1217–1227. New York: ASME.
Technology. Kruse, B., & Shea, K. (2016). Design library solution patterns in SysML for
de Kleer, J., & Brown, J.S. (1984). A qualitative physics based on conflu- concept design and simulation. Procedia CIRP 50, 695–700.
ences. Artificial Intelligence 24(1–3), 7–83. Kypuros, J. (2013). System Dynamics and Control with Bond Graph Model-
Deng, Y.M. (2002). Function and behavior representation in conceptual me- ing. Boca Raton, FL: CRC Press.
chanical design. Artificial Intelligence for Engineering Design, Analysis Linsey, J.S., Markman, A.B., & Wood, K.L. (2012). Design by analogy: a
and Manufacturing 16(5), 343–362. study of the WordTree method for problem re-representation. Journal
Deng, Y.M., Britton, G.A., & Tor, S.B. (2000). Constraint-based functional of Mechanical Design 134, 1–12.
design verification for conceptual design. Computer-Aided Design 32, Linsey, J.S., Tseng, I., Fu, K., Cagan, J., Wood, K.L., & Schunn, C. (2010). A
889–899. study of design fixation, its mitigation and perception in engineering de-
Deng, Y.M., Tor, S.B., & Britton, G.A. (1999). A computerized design environ- sign faculty. Journal of Mechanical Design 13, 41.
ment for functional modeling of mechanical products. Proc. 5th ACM Symp. Lopez, R., Linsey, J.S., & Smith, S.M. (2011). Characterizing the effect of
Solid Modeling and Applications, pp. 1–12, Ann Arbor, MI, June 8–11. domain distance in design-by-analogy. Proc. 23rd Int. Conf. Design The-
Dorf, R., & Bishop, R. (1998). Modern Control Systems, 8th ed. Boston: ory and Methodology, pp. 141–151, Washington, DC, August 28–31.
Addison-Wesley. Lucero, B.M. (2014). Design-analogy performance parameter system (D-
Doyle, J., Francis, B., & Tannenbaum, A. (1990). Feedback Control Theory. APPS). PhD Thesis. Colorado School of Mines.
New York: Macmillan. Lucero, B., Linsey, J., & Turner, C.J. (2016). Frameworks for organising design
Enderle, J., & Bronzino, J. (2012). Introduction to Biomedical Engineering performance metrics. Journal of Engineering Design 27(4–6), 175–204.
(Bronzino, J., Ed.), 3rd ed. Waltham, MA: Elsevier. Lucero, B., Viswanathan, V., Linsey, J., & Turner, C. (2014). Identifying crit-
Fenves, S.J. (2002). A Core Product Model for Representing Design Informa- ical functions for use across engineering design domains. Journal of Me-
tion. Washington, DC: US Department of Commerce, Technology Ad- chanical Design 136(12), 121101.
ministration, National Institute of Standards and Technology. Mayr, O. (1970). The origins of feedback control. Scientific American
Friedenthal, S., Moore, A., & Steiner, R. (Eds.) (2012). A Practical Guide to 223(4), 110–118.
SysML, 2nd ed. Waltham, MA: Elsevier. Mcbride, R.T., & Cellier, F.E. (2001). A bond-graph representation of a two-
Friedenthal, S., Moore, A., & Steiner, R. (2015). Modeling event-based be- gimbal gyroscope. Simulation Series 33, 305–312.
havior with state machines. In A Practical Guide to SysML (Friedenthal, Montecchi, T., & Russo, D. (2011). FBOS: function/behaviour-oriented
S., Moore, A., & Steiner, R., Eds.), pp. 273–294. Waltham, MA: search. Proc. 11th ETRIA World TRIZ, Dublin, Ireland.
Elsevier. Nagel, R.L. (2007). Signal flow grammar from the functional basis. Proc. Int.
Fu, K., Chan, J., Cagan, J., Kotovsky, K., Schunn, C., & Wood, K. (2013). Conf. Engineering Design, Paris, August 28–31.
The meaning of “near” and “far”: the impact of structuring design data- Nagel, R.L., Vucovich, J.P., Stone, R. B., & McAdams, D. A. (2008). A sig-
bases and the effect of distance of analogy on design output. Journal of nal grammar to guide functional modeling of electromechanical products.
Mechanical Design 135(2), 21. Journal of Mechanical Design 130(5), 51.
Fu, K., Murphy, J., Yang, M., Kevin, O., Jensen, D., & Wood, K.L. (2013). Otto, K.N., & Wood, K.L. (2001). Product Design: Techniques in Reverse
Investigating the effect of functionality level of analogical stimulation on Engineering and New Product Development. Englewood Cliffs, NJ:
design outcomes. Proc. Korea-Japan Design Engineering Workshops, Prentice Hall.
Kitakyushu, Fukuoka, Japan, November 28–30. Pahl, G., & Beitz, W. (1961). Engineering Design: A Systematic Approach.
Gentner, D. (1983). Structure-mapping: a theoretical framework for analogy. London: Design Council.
Cognitive Science 7, 1–5. Qian, L., & Gero, J.S. (1996). Function-behavior-structure paths and their
Gentner, D., & Gentner, D.R. (1982). Flowing Water or Teeming Crowds: role in analogy-based design. Artificial Intelligence for Engineering De-
Mental Models of Electricity. Washington, DC: Office of Naval Research sign, Analysis and Manufacturing 10, 289–312.
Personnel and Traing Research Programs. Russo, D. (2012). Functional-based search for patent technology transfer.
Gero, J.S. (1990). Design prototypes: a knowledge representation schema for Proc. ASME 2012, pp. 1–11, Chicago, August 12–15.
design. AI Magazine 11(4). Russo, D., & Rizzi, C. (2014). A function oriented method for competitive
Goel, A.K., & Bhatta, S.R. (2004). Use of design patterns in analogy-based technological intelligence and technology forecasting. Proc. 2014 Int.
design. Advanced Engineering Informatics 18(2), 85–94. Conf. Engineering, Technology and Innovation, pp. 1–9, Bergamo, Italy,
Goel, A.K., Rugaber, S., & Vattam, S. (2009). Structure, behavior and func- June 23–25.
tion of complex systems: the SBF modeling language. Language 23(1), Sonntag, R.E., Borgnakke, C., & Van Wylen, G.J. (2002). Fundamentals of
23–35. Thermodynamics. New York: Wiley.
 Quantitative engineering design methods 475
Westheimer, G. (1954). Eye movement responses to a horizontally moving Matthew J. Adams is a PhD student in biological design at
visual stimulus. AMA Archives of Ophthalmology 52(6), 932–941.
White, F. (2003). Fluid Mechanics, 6th ed. New York: McGraw-Hill.
Arizona State University. He is also a Co-Founder of the
Wu, Z., Campbell, M.I., & Fernández, B.R. (2008). Bond graph based auto- health systems engineering company Kintel, LLC. Matt re-
mated modeling for computer-aided design of dynamic systems. Journal ceived his BS degree in physics from the University of Mary-
of Mechanical Design 130(4), 41102.
Yuan, Z., & Ljung, L. (2016). Black-box identification of multivariable
land, College Park, and his MS degree in mechanical engi-
transfer functions—asymptotic properties and optimal input design. In- neering from the Colorado School of Mines. He currently
ternaltional Journal of Control 40(2), 233–256. holds student memberships in both ASME and the Orthope-
dic Research Society.
Briana M. Lucero is an R&D Engineer in the Advanced En- Cameron J. Turner is a Professor in the Department of Me-
gineering Technologies Division at Los Alamos National chanical Engineering at Clemson University. He previously
Laboratory. She holds a bachelor’s degree in engineering worked at the Colorado School of Mines and was also a Tech-
(mechanical and electrical specialties), a master’s degree in nical Staff Member at Los Alamos National Laboratories. Dr.
mechanical engineering, a master’s degree in systems engi- Turner teaches classes in engineering design methods, design
neering from Johns Hopkins University, and a doctoral de- optimization, mechanical systems, computer-aided design/
gree in engineering systems from Colorado School of Mines. engineering/manufacturing, and design of complex systems.
Briana’s current research is on modeling and simulation of Cameron is currently the Program Chair for the ASME CIE
tools in additive manufacturing. Dr. Lucero worked on satel- Division Executive Committee and is a member of the Inter-
lites and scientific imaging instrumentation as an Aerospace national Design Simulation Competition Committee of
Engineer at Ball Aerospace and Technologies Corporation ASME. Prof. Turner won the CSM Design Program Direc-
and Los Alamos National Laboratory. That work informed tor’s Award in 2015 for service to the capstone design
her use of systems engineering. program.

View publication stats