Ecological Principles and

Metrics for Improving Material

Astrid Layton1
Mem. ASME
Cycling Structures in
School of Mechanical Engineering,
Manufacturing Networks

Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020

Georgia Tech Lorraine,
57070 Metz, France
e-mail: alayton6@gatech.edu
Bert Bras
Mem. ASME
School of Mechanical Engineering,
Georgia Institute of Technology,
Atlanta, GA 30332-0405
e-mail: bert.bras@me.gatech.edu
Marc Weissburg
School of Biology,
Georgia Institute of Technology,
Atlanta, GA 30332-0230
e-mail: marc.weissburg@biology.gatech.edu
A key element for achieving sustainable manufacturing systems is efficient and effective
resource use. This potentially can be achieved by encouraging symbiotic thinking among
multiple manufacturers and industrial actors and establish resource flow structures that
are analogous to material flows in natural ecosystems. In this paper, ecological princi-
ples used by ecologists for understanding food web (FW) structures are discussed which
can provide new insight for improving closed-loop manufacturing networks. Quantitative
ecological metrics for measuring the performance of natural ecosystems are employed.
Specifically, cyclicity, which is used by ecologists to measure the presence and strength
of the internal cycling of materials and energy in a system, is discussed. To test applic-
ability, groupings of symbiotic eco-industrial parks (EIP) were made in terms of the level
of internal cycling in the network structure (high, medium, basic, and none) based on the
metric cyclicity. None of the industrial systems analyzed matched the average values and
amounts of cycling seen in biological ecosystems. Having detritus actors, i.e., active recy-
clers, is a key element for achieving more complex cycling behavior. Higher cyclicity val-
ues also correspond to higher amounts of indirect cycling and pathway proliferation
rate, i.e., the rate that the number of paths increases as path length increases. In FWs,
when significant cycling is present, indirect flows dominate direct flows. The application
of these principles has the potential for novel insights in the context of closed-loop manu-
facturing systems and sustainable manufacturing. [DOI: 10.1115/1.4033689]

Introduction 240,000 tons, water savings of 3  106 m3 through recycling and

reuse, 30,000 tons of straw being converted to 5.4  106 l of etha-
Substantial progress has been made in the past two decades in
nol, 150,000 tons of yeast replacing 70% of the soy protein in tra-
the field of sustainable manufacturing. Many are looking for tech-
ditional feed mix for more than 800,000 pigs, and the recycling of
nologies that will solve the environmental problems related to
150,000 tons of gypsum from desulfurization of flue gas (SO2),
manufacturing, see, e.g., Refs. [1] and [2]. In 2001, a comprehen-
replacing the import of natural gypsum (CaSO4) [5].
sive study on environmentally benign manufacturing, however,
Unfortunately, symbiotic relationships do not guarantee suc-
found that there was no evidence that the environmental problems
cess, as the world of production and development is constantly
from our production systems are solvable by a “silver bullet” tech-
fluctuating and can be difficult to predict in the long term. Mone-
nology [3,4]. Rather, the need for systems-based solutions was
tary problems halt the implementation of many exciting EIP plans,
noted. Achieving true sustainable manufacturing requires adop-
companies that must fill such plans may remain unconvinced that
tion of a systems view that goes beyond a single factory and com-
moving locations would be financially beneficial, and things may
pany. Close collaboration between multiple companies,
fall apart for a number of reasons between an EIP’s early develop-
stakeholders, and even public–private partnerships around internal
ment and its maturation. Furthermore, there is also no real design
and external value chains and resource networks have to be pur-
guidance for establishing sustainable symbiotic industrial resource
sued. Although all manufacturers strive for efficient and effective
networks beyond typical ad-hoc approaches. The solution to these
resource utilization, we still have a long way to go in order to
difficulties may be found within the biological systems that were
move from the current “take-make-waste” society to a truly cycli-
the inspirations for EIPs. Biological systems, specifically ecosys-
cal production paradigm.
tems, have evolved to maintain their function in nonideal and
Nature is an excellent source of inspiration for sustainable man-
fluctuating conditions. A better understanding of biological eco-
ufacturing in the guise of ecological systems, which are canonical
systems is needed in order to identify and apply key components
examples of sustainability. Industrial symbiosis, named for the
from ecosystems that support the development and formation of
analogous close and often long-term interactions between two or
sustainable industrial manufacturing systems.
more biological species, occurs when multiple firms or facilities
The goal of biologically inspired design of industrial manufac-
achieve higher system efficiency through the exchange of “waste”
turing networks is to reproduce the sustainable cycling and recy-
energy and materials. An EIP is characteristic of a type of indus-
cling that characterize ecosystems, ideally achieving a highly
trial symbiosis that occurs among firms collocated in a bounded
efficient closed-loop flow of materials. This paper investigates
geographic area, typically an industrial park. The exchanges char-
the cycling of material/food through ecosystems and industrial
acteristic of EIPs has proven environmental benefits, most often
systems using ecological metrics. The goals of this article are:
reducing resource consumption and emissions. For example, the
(a) to review relevant principles and findings from ecology
mutualistic relationships within the well-known Kalundborg EIP
related to FWs and their functioning; (b) to compare industrial
have resulted in the EIPs yearly CO2 emission being reduced by
case studies against ecological system using key metrics, specifi-
cally cyclicity; (c) to identify key differentiating components
1
Corresponding author.
between current industrial systems and ecologies; and (d) to pro-
Manuscript received December 15, 2015; final manuscript received May 17, vide suggestions on how to improve industrial symbiosis based
2016; published online June 22, 2016. Editor: Y. Lawrence Yao. on the findings.

Journal of Manufacturing Science and Engineering OCTOBER 2016, Vol. 138 / 101002-1
C 2016 by ASME
Copyright V
 This paper demonstrates how insight and improvements to
cyclical networks can be made based on structural information
only. That is, only the presence and direction of flows between
two actors in the network is required and information on flow
magnitude is not necessary to provide insight into how to improve
material cycling. Although detailed flow knowledge allows for
more precise network representations, the trade-off is the effort
associated with collecting this quantitative flow information. The
cost of information is less with structural metrics that only require
knowing a link exists between two actors than with flow metrics
that require knowing amounts of material or energy transfer.
Review of Relevant Ecological Principles
FW Terminology and Industry Definitions. The importance
of characterizing the anatomy of ecological networks is given by
Strogatz: “structure always affects function” [6]. By reproducing
the structure of these biological networks, the hope is that the
functioning of human systems will replicate the inherently sus-
tainable natural world as well. The appropriate application of eco-
logical principles and analyses depends on building models that
specify how principles from biology are translated to industry,
and back again. One biological model for ecosystems is an FW.
Somewhere during the process of translating this model to indus-
try, the defining characteristic of an ecosystem (the weblike struc-
ture) is discarded and industry is left with a unidirectional,
top–down “food chain” [7]. Companies within an industrial park
or components in an industrial cycle are cast as species, and the
material and energy exchanges between them are analogous to the
transfer of caloric energy that supports the species (metabolism).
At first glance, the comparison may seem a complete one; how-
ever, the transfer of ecological properties and principles to indus-
try is highly complex and much is missing. Definitions have led to
the sustainable design slogan “waste equals food,” a slogan that
although consistent with systems in nature does not fully capture
the important workings of ecological systems. The lack of a well-
translated framework has led to many discrepancies in the imple-
mentation and interpretation of ecological principles and how
they advise the organization of industrial system [8]. A framework
built on real and complete ecological knowledge is of the utmost
importance to accomplish this goal. Extensive literature exists to
aid in the successful translation of many desirable properties
found in nature to industry, e.g., Refs. [9–12].
Industry Desirable FW Properties. Models and structural
metrics have been developed to analyze and explain specific prop-
erties of ecosystems, such as the system’s ability as a whole to
withstand environmental fluctuations and support diverse species,
which could be immensely beneficial to industry [13,14]. Findings
indicate that FWs are composed of multiple pathways through
which material and energy flow, strongly connected groups of
interacting species compartments with weak interactions between
the compartments, and a modular structure that is hypothesized to
increase the overall systems stability by localizing interactions
and disruptions [15–17]. This hypothesis, however, has been diffi-
cult to fully resolve [18,19]. Replicating ecosystem robustness
and stability in human systems could ease the damage caused by
supply chain disruptions, which reduce the share price of the
affected companies so significantly that 80% of companies world-
wide consider better protection of supply chains top priority [20].
In 1969, Odum recognized that ecological systems, particularly
mature ones, are associated with a high degree of internal recy-
cling of energy and materials such that the amount of new inputs
into the system is small compared to what is transformed among
the system components [9]. Human systems in contrast (e.g., agri-
cultural ones) are geared for production rather than efficiency,
resembling young rather than mature natural systems. Odum has
suggested mimicking mature systems would help shift the focus
of human systems from production to efficiency [9]. One desirable
101002-2 / Vol. 138, OCTOBER 2016
property of mature systems is a complex FW structure, a prolifera-
tion of connections between species that exchange material and
energy [21]. The centripetal nature of FW structure also is a sell-
ing point for industry. When a species becomes more efficient in
use or acquisition of a resource, its population increases. Centripe-
tality results in this singularly focused positive change being cas-
caded through the system such that all the populations of species
involved are benefited [17,22]. Translated to industry, this would
mean that a change which benefits one company within an EIP
translates into a park-wide positive net effect.
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
A hypothesis within industrial ecology is that diversity, in the
sense that a wide range of species types are contained within any
system, could contribute to a more stable system: when one firm
departs, the system may adapt or recover by another actor(s) step-
ping in to fulfill the supplying role [23].
An analysis of 40 FWs by Briand indicates that connectance,
which is a measure of the number of direct to the total possible
interactions in a web and an important parameter in the previous
ecosystem analyses, declines as variability of the environment
increases [24]. Following a line of reasoning strongly influenced
by May’s theoretical analysis [25], Briand argues that differences
in connectance values for ecosystems in stable and unstable envi-
ronments are the result of limitations in feeding periods caused by
environmental fluctuations, which can lead organisms to depend
upon intermittent, intense feedings. This suggests a structure
which is dependent on the stability of resources, a property of
interest for industries.
Structural FW Analyses. Ecologists use simple unweighted
digraphs to quantify the characteristics of FW, where every link
indicates direction, in this case from resource (prey) to consumer
(predator) [17]. Species or functional groups are represented in
the digraphs such that any species with identical predators and
prey are grouped as trophic species, which has been found to
reduce methodological bias in the data [13,17,26–28].
Ecologists use structural measures and metrics for the analysis
of ecological FWs [14,24,29–31]. The meaning and calculation of
each ecological measure/metric is best understood within the con-
text of an organizational matrix. Organizational matrices are used
by ecologists to collect and document the exchanges between
species or functional groups within the community at hand.
These matrices can document a variety of interactions from
predator–prey exchanges to all interactions in a community,
including any competitive interactions. Three types of matrices
are commonly used by ecologists: a FW matrix [F], a community
matrix [C], and an adjacency matrix [A].
We use an FW matrix for the representation of the exchanges in
industrial systems and ecological FWs. Analogous to a connectiv-
ity matrix [32], an FW matrix is concerned only with the struc-
tural information (links and nodes) of a network and defines the
pathways that exist by which material and energy flows from one
compartment to another. It is blind to information such as flow
rate, quality, and the type of flow. A link exists as long as some
physical quantity directly joins two nodes. Only flow existence
and direction are captured. An FW matrix [F] captures the
observed consumer–producer interactions that include both plant
and animal species (Ecologists commonly use different terms to
describe these interactions such as plant–herbivore versus
Predator–prey but for simplicity we use predator–prey to describe
any interaction where one species is consumed by another.). The
left half of Fig. 1 depicts a hypothetical FW represented as a
directional digraph; the right half represents the web as an FW
matrix [F]. Since a species (N) can be both predator and prey, the
result is a square matrix. A value of 1 indicates the existence of a
flow while 0 indicates the absence of a flow. In Fig. 1-right, a
value of 1 for L13 indicates that species 1 (S1) contributes resour-
ces to species 3 (S3). Alternatively, one can interpret that element
of the matrix to mean S3 “consumes” S1. Each row in an FW
matrix captures the flow of resources from one species to all
Transactions of the ASME
 Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Fig. 1 An example ecosystem (left), its FW (center), and its FW matrix representation (right).
S1–S3 represent the three species highlighted in the ecosystem and L11–L33 represent the link-
ages between them.

species in a web and each column captures the input of resources [F] and dividing this figure by the number of columns with
to a particular species from all species in the web. In other words, nonzero elements (the number of predators/consumers).
if predator j feeds on prey-i, then fij ¼ 1; the interaction (or link,  Vulnerability (V)—The average number of predators per
L) is accounted for exactly once in the FW matrix. The maximum prey in a web, which corresponds to the average number of
number of links L scales as (N)*(N  1) if a given species does consumers a producer interacts with in an industrial network.
not eat itself, and as (N2) if cannibalism is allowed (noted as a 1 Generated by summing the rows in an FW matrix [F] and
on the diagonal). A structural adjacency matrix [A] is the trans- dividing by the total number of rows with nonzero elements
pose of the FW matrix. (the number of prey/producers).
 Cyclicity (kmax)—A measure of the strength and presence of
cyclic pathways present in the system [32,34]. Obtained by
finding the maximum real eigenvalue of the transpose of the
Common Ecological Metrics for Assessing FW Structure. A FW matrix [F].
wide variety of metrics have been developed to understand the Formulae for these metrics are per Eqs. (1)–(13). fij represents
link between structure and behavior of ecological systems [32,33]. the linkage between actor i to actor j and is the ith row and jth col-
The structural measures and metrics used most frequently by ecol- umn entry in the matrix with a value of either 1 (link exists) or 0
ogists can be calculated using the N  N structural FW matrix (no link). All calculations per Eqs. (1)–(13) are simply based on
(such as given in Fig. 1); that is, all calculations are simply based binary information on whether a link exists between two actors in
on binary information on whether or not a link exists between two the matrix, or not, which makes these metrics very suitable for
actors in the matrix. The following are key metrics typically used application to manufacturing networks because it eliminates the
by ecologists to perform structural assessments of FWs: need for obtaining often proprietary material and energy flow data
 Number of species or actors (N)—The total number of actors between manufacturers
in a network, sometimes termed “species richness.” Repre-
sented by the size (number of rows or, as the two are equal, L
LD ¼ (1)
columns) of the FW matrix [F] [24]. N
 Number of links (L)—The number of direct links or interac- 8
tions between actor in a network. Represented by the total > X
n
>
> 1 for fij > 0
number of nonzero interactions in the FW matrix [F] [24] >
<
 Linkage density (Ld)—The ratio of the total number of links j¼1
frow ðiÞ ¼ X
n (2)
to the total number of actors in a network. >
>
> fij ¼ 0
 Number of prey (nprey)—Actors which are eaten by at least : 0 for
>
j¼1
one other. Represented by the number of nonzero rows in the
FW matrix [F]. “Prey” in an industrial network transfer ma- X
m
terial or energy to be utilized by others; that is, they are pro- nprey ¼ frow ðiÞ (3)
ducers [14]. i¼1
 Number of predator (npredator)—Actors that eat at least one 8
other. Represented by the number of nonzero columns in an > X
m
>
>
FW matrix [F]. “Predators” in an industrial network receive >
< 1 for fij > 0
material or energy from others; that is, they are consumers fcol ðjÞ ¼ i¼1
(4)
[14]. > X
m
>
> 0 for fij ¼ 0
 Prey to predator ratio (Pr)—The ratio of the number of >
:
i¼1
actors eaten by another to the number of actors that eat
another. That is, the ratio of prey:predators or producers:con- X
n
sumers. This is the number of nonzero rows in an FW matrix npredator ¼ fcol ðjÞ (5)
[F] divided by the number of nonzero columns. j¼1
 Specialized predator fraction (Ps)—The number of predators
(consumers) eating only one actor divided by the total num- nprey
ber of consumers in the network. This is the sum of the num- PR ¼ (6)
npredator
ber of columns with only one nonzero element in the FW
matrix [F] divided by the total number of columns with non- 8
>
> X
m
zero elements. >
> fij ¼ 1
< 1 for
 Generalization (G)—The average number of prey eaten per i¼1
fscol ðjÞ ¼ X
m (7)
predator in a web, which corresponds to the average number >
>
of producers a consumer interacts with in an industrial net- >
> 0 for fij 6¼ 1
:
work. Generated by summing the columns in an FW matrix i¼1

Journal of Manufacturing Science and Engineering OCTOBER 2016, Vol. 138 / 101002-3
 X
n node, only the maximum (dominant) eigenvalue is left to repre-
nSpredator ¼ fscol ðjÞ (8) sent the pathway proliferation rate of the system as the limit of the
j¼1 number of indirect links (pathways between two nodes which con-
sist of more than one link) goes to infinity.
nSpredator Cyclicity can be 0, 1, or greater than 1.This is illustrated in
PS ¼ (9)
npredator Fig. 2, which is based on the similar figure by Fath and Halnes
[32,35]. Zero cyclicity indicates that no internal cycles are pres-
L ent, Fig. 2(a). In these networks, energy traveling through the sys-
G¼ (10) tem never passes through a component twice. A value of one is
npredator
representative of a network where only simple closed-loop path-

Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020

L ways exist, Fig. 2(b). Those networks that have cycles made up of
V¼ (11)
nprey
L
C¼ 2 (12)
N in Fig. 2(c). The larger the cyclicity, the more complex and
kmax ¼ max; real eigenvalue solution to : 0 ¼ detðA  kIÞ (13)
Cyclicity is a key metric that addresses resource cycling in eco-
systems and will be discussed in more detail in the following
section.
Cyclicity. Cyclicity (kmax) is a measure of the number and pro-
liferation of cyclic pathways present in the system [32] as the
number of steps in the path approaches infinity. Cyclicity is
obtained by finding the maximum real eigenvalue of a web’s
structural adjacency matrix [A] that is the transpose of the FW
matrix [F]. The maximum real eigenvalue (kmax) is a measure of
the proliferation of pathways that originate and terminate at the
same node. There is a greater potential for flows to remain within
the system as pathways proliferate, and so kmax is indicative of the
resulting internal cycling [17].
The use of eigenvalues to determine cyclicity (also known as
“pathway proliferation rate”) of a system combines results from
graph theory and linear algebra [17]. The proof presented by Bor-
rett et al. uses the Perron–Frobenius theorem, which guarantees
that there is only one real eigenvalue that is greater than or equal
to all other eigenvalues (k1  ki for i ¼ 2,…, n) in adjacency matri-
ces associated with a strongly connected network [17]. In net-
works where it is possible to reach every node from every other
one link (self-loops) or have cycling only if link direction is
ignored may have a maximum eigenvalue of either 1 or 0 [17]. A
network with a maximum eigenvalue greater than 1 indicates that
the network is made up of complex looped pathways, as described
numerous the paths are between components, creating a system
that is more interconnected. Most FWs are composed of networks
where large subsets of “nodes” are strongly connected such that
the maximum eigenvalue is greater than 1, indicating the exis-
tence of multiple cyclic pathways.
With respect to cyclicity, the dynamics and stability of FWs are
significantly influenced by nutrient recycling and decomposition
[36]. In ecosystems, the detritivores (earthworms, fungi, and bac-
teria, for example) are responsible for the decomposition of dead
organic matter and the distribution of nutrients to the system,
often known as the “recyclers of the biosphere.” This decomposi-
tion and redistribution create a fixed cyclic structure in the system
as measured by cyclicity [37].
Comparison of EIPs and Ecosystems Based on Cyclicity
Numerical Comparison. Given the importance of cyclicity in
FWs, we now look at how ecological metrics can be applied to
industrial manufacturing systems, specifically, EIPs. In Ref. [38],
the success of proposed versus existing EIPs was examined and
compared to the performance of ecological networks using struc-
tural ecosystem metrics. The goal of mimicking median values of
the FW metrics stems from the belief that form follows function.
EIPs that match the form of FWs will function more like the FWs.

Fig. 2 Examples of the three types of internal structural cycling as represented by cyclicity
(eigenvalues of [A]): (a) no cycling kmax 5 0, (b) weak cycling kmax 5 1, and (c) and strong
cycling kmax > 1

101002-4 / Vol. 138, OCTOBER 2016 Transactions of the ASME

For this comparison, a comprehensive dataset of 48 EIPs was
compiled and analyzed. The EIPs in this dataset can be classified
based on the current (or as current as possible) status of the EIP:
existing, proposed, or failed. EIPs currently (or as current as pos-
sible) active/in operation/existing are often termed “successful” in
the literature as they have been fully or mostly implemented and
are still running. The locations of the EIPs listed span the globe,
ranging from various locations in the U.S. to China to Denmark
and France. The proposed EIPs are often based on an existing
industrial park where additional linkages between existing and
new companies to increase the symbiotic relationships have been
suggested. Some of these EIPs exist entirely on paper. Three EIPs
that were fully or mostly implemented are no longer in operation
for unknown reasons.
Note that this analysis defined the system boundary of the EIP
as not including customers that may consume products (including
energy) created by the EIP actors. Although including this actor
might change some of the properties examined, it would not
change the analysis of cycling since it is not possible in most cases
to account for returns to the EIP from this type of actor. Moreover,
although defining system boundaries is always a challenge, the
goal of a successful EIP is to reduce raw material inputs and waste
outputs by increasing the cycling of material between actors in the
EIP. Material and energy outputs to customers outside of the EIP
boundary do not fall into this category. Thus, the system boundary
definition used here is the most relevant to the industrial goals that
EIPs are designed to achieve.
Cannibalism is excluded from this analysis both for FWs and
EIPs to permit accurate comparisons. Cannibalism in an EIP
would correspond to a particular actor directly recycling or reus-
ing parts of its own waste stream. Although admittedly important,
these self-loops do not contribute to the cyclicity measured in a
system (see below), and information on this potential link is not
easily obtained.
The 48 EIPs are ranked in four groups based upon the existence
and complexity of the internal cycling measured by the ecological
metric cyclicity. As mentioned, the characteristic cycling seen in
FWs is an especially desirable function for developing sustainable
manufacturing as it embodies the major goal for eco-industrial
networks: closed-loop manufacturing. The four groups are as
follows:
 Class A: The EIPs with a designation of class A are represen-
tative of highly complex internal cycling. This is defined as
those EIPs with a cyclicity value greater than or equal to 3
(kmax  3). These EIPs represent the top tier of collected
systems.
 Class B: The EIPs with a designation of class B are represen-
tative of complex internal cycling. This is defined as those
EIPs with a cyclicity value greater than 1 (3 > kmax > 1).
 Class C: The EIPs with a designation of class C contain sim-
ple internal cycling. This is defined as those EIPs with a
cyclicity value equal to 1 (kmax ¼ 1).
 Class D: The EIPs with a designation of class D have no
internal cycling present. This is defined as those EIPs with a
cyclicity value equal to 0 (kmax ¼ 0). All of the EIPs in this
grouping pass along a by-product to another industry for use
rather than disposal; however, they do not have the more
complex cycling that results from the reintroduction of that
by-product into the system.
Table 1 shows the resulting rank ordering of the 48 EIPs with
respect to cyclicity (kmax) as well as the corresponding values for
linkage density (LD), the prey–predator ratio (PR), generalization
(G), and vulnerability (V). The median values for these metrics
from a post 1993 data set consisting of 50 natural FWs also have
been included.
When compared to median values for FWs, the values for EIPs
highlight that EIPs do not match the values characteristic of eco-
systems. All ecological metric values for EIPs markedly differ
with the average values calculated for ecosystems. In Table 2, a
Journal of Manufacturing Science and Engineering
summary of the differences is given in terms of best and worst
values.
The largest and most consistent differences between EIPs and
biological ecosystems occur for cyclicity and linkage density,
which are both metrics that characterize the type and presence of
connections within the system. None of the EIPs reach a cyclicity
or density of links close to that which is found for biological eco-
systems. As seen in Table 2, the closest EIP has a cyclicity of 3.87
and a linkage density of 3.13 as compared to the respective
median FW values of 4.24 and 5.04. The other structural metrics
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
calculated (ratio of prey to predators, generalization, and
vulnerability) also fall short of FW median values but to a lesser
degree.
For all metrics, those EIPs which have high cyclicity values
(three or greater) show the smallest difference from FW averages.
Those EIPs that have no internal cycling (cyclicity of zero) or
basic internal cycling (cyclicity of one) show the largest differ-
ence from FW averages.
EIPs in all groupings come closest on average to matching the
median values for the FW metric vulnerability (V) and prey to
predator ratio (PR), with both metrics coming in with an average
value for all EIP types 19% lower than FWs. However, the EIPs
with the highest cyclicity (kmax  3) come closest to average FW
values for all metrics. This supports the notion that EIP designers
and decision makers should be aiming for structural designs that
converge on FW metrics in order to obtain the highest cyclicity.
The results of the existing and failed EIPs are due partially to a
response to external stimuli. Industrial networks from which EIPs
are built are “complex adaptive systems” where the system does
not adapt with any coordination but rather, it is the components
that change in their own best interest in response to external con-
ditions [39]. EIPs experience a certain amount of purposeful coor-
dination between the participating companies, but still experience
to different degrees the complex adaptive system response. The
best scenarios would be expected to be those created on paper as
these have been ideally designed with the total purpose of opti-
mizing a coordinated network behavior. The proposed EIPs show
a tendency to come slightly closer to FW averages than the exist-
ing or failed EIPs. This is most likely due to the fact that the pro-
posed EIPs have not had to deal with the realities of operation yet;
on paper, one may make a very beautiful design; however, in
actual operation, the design may not be possible. These
“proposed” EIPs, however, still possess cycling well below that
found in biological systems.
An EIP with no internal cycling seems contrary to what one
expects of a bio-inspired industrial network because one of the
most influential and identifying characteristics of biological net-
works is the prevalence and importance of materials and energy
cycling within the system. What is observed in the EIP results rep-
resents the difference between a simple “waste ¼ food” analogy
and a truly biologically inspired FW, the two concepts are illus-
trated by Fig. 3. The industrial networks of class D, all of which
have a cyclicity of zero, follow the linear structure of the food
chain in Fig. 3-left. Even though many of these networks
exchange and reuse by-products, the system is still made up of a
linear chain of relationships, characterized by the food chain in
Fig. 3-left. The industrial networks having cyclicity greater than
zero, those in classes A, B, and C, begin to show some of the eco-
logical benefits characteristic of strong internal cycling, character-
ized by the FW in Fig. 3-right. The EIPs in these higher classes
exhibit median values for all the metrics used here closer to
medians for the FWs.
The characteristic cycling seen in FWs is an especially desira-
ble function for developing sustainable manufacturing. This
cycling of materials and energy in FWs brings with is a host of
other industry desirable properties and functions. The successful
establishment of cycling in EIPs can also be understood through
the presence and relative strength of indirect effects. Indirect
effects have been shown to be very important to the workings of
ecosystems [40,41] and investigation of their influence in
OCTOBER 2016, Vol. 138 / 101002-5
 Table 1 FW metrics for the 48 EIPs, rank ordered based on cyclicity, alongside median values for the 50 FWs collected after 1993

kmax LD PR G V

FWs post-1993 median values (50): 4.24 5.04 1.09 6.18 5.34

A class EIPs Proposed The Green Triangle 3.92 3.25 1.14 3.71 3.25
Exists Pomacle–Bazancourt 3.70 2.67 1.00 3.00 3.00
Proposed Renova (RRP) 3.39 3.00 1.00 3.00 3.00
Proposed Clark Special Economic Zone 3.34 2.55 0.89 2.68 3.00
Exists Copper Industry Web 3.12 3.20 1.00 4.00 4.00

Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020

Exists Kitakyushu RRP 3.00 1.55 0.80 1.70 2.13
B class EIPs Exists Kwinana 2.59 1.89 0.79 2.13 2.68
Exists Ulsan Industrial Park 2.42 1.75 1.00 2.00 2.00
Exists Humber Industrial Symbiosis Project 2.21 1.47 0.64 1.79 2.78
Exists Uimaharju Forest Industry Park 2.15 2.22 0.89 2.22 2.50
Exists UPM Kymi pulp and paper mill 2.08 2.15 1.20 2.80 2.33
Exists Harjavalta Industrial Area 2.00 2.00 0.83 2.00 2.40
Proposed GERIPA (IBS) 1.93 1.75 0.75 1.75 2.33
Exists Kawasaki 1.88 2.00 1.00 2.00 2.00
Exists Kymi EIP 1.82 1.75 1.00 2.00 2.00
Exists Burnside EIP 1.80 1.82 0.82 1.82 2.22
Exists Suzhou Eco-Industrial Park 1.73 1.56 0.89 1.56 1.75
Exists Devens EIP 1.73 1.43 0.82 1.76 2.14
Exists Guitang Sugarcane EIP Project 1.70 1.78 0.78 1.78 2.29
Exists Tianjin Economic Development Area 1.66 1.38 1.33 1.83 1.38
Exists Scotia Investments 1.62 1.43 1.40 2.00 1.43
Exists Kalundborg EIP 1.62 1.50 0.54 1.62 3.00
Exists Seshasayee Paper and Board Ltd.: Agro Industrial Eco-Complex 1.62 1.57 0.86 1.57 1.83
Exists Guayama 1.62 1.33 0.67 1.33 2.00
Proposed Mongstad EIP 1.55 1.73 0.70 1.90 2.71
Failed AES Thames EIP 1.47 1.63 1.60 2.60 1.63
Failed Brownsville EIP 1.41 1.38 1.20 2.20 1.83
Exists Barceloneta 1.41 1.14 0.57 1.14 2.00
Proposed Red Hills EcoPlex 1.32 2.00 1.00 2.00 2.00
Exists Fushan Farms (IBS) 1.27 1.29 0.71 1.29 1.80
Exists Nanning Sugar Company 1.22 1.38 0.75 1.38 1.83
Exists Monfort Boys Town (IBS) 1.22 1.22 0.88 1.38 1.57
Exists Tunweni Brewery (IBS) 1.17 1.13 0.88 1.13 1.29
C class EIPs Proposed Stoneyfield Londonderry EIP 1.00 2.15 0.83 2.33 2.80
Proposed Lower Mississippi Corridor 1.00 1.74 0.78 2.22 2.86
Proposed PV Symbiosis Prop 1.00 1.56 0.75 1.75 2.33
Exists Wallingford Eco-Industrial Park 1.00 1.50 0.82 1.64 2.00
Exists Styrian Recycling Network 1.00 1.13 0.86 1.57 1.83
Exists Landskrona 1.00 1.07 1.00 1.45 1.45
Exists An Son Village 1.00 1.00 0.67 1.00 1.50
Exists Jyvaskyla 1.00 1.00 0.50 1.00 2.00
Exists NIA-KIADB 1.00 0.71 0.67 1.11 1.67
D class EIPs Exists Lubei Industrial Park 0.00 1.89 1.14 2.43 2.13
Proposed Gladstone (with potential links 2008) 0.00 1.13 1.00 1.86 1.86
Exists Pingdingshan Coal Mining Group 0.00 1.00 1.00 1.33 1.33
Failed Triangle J EIP 0.00 0.95 1.88 2.25 1.20
Exists Gladstone (2005) 0.00 0.88 2.00 2.33 1.17
Proposed Connecticut Newsprint 0.00 0.83 0.40 1.00 2.50

industrial networks has been done as well [42,43]. Key aspects on

how to achieve greater cyclicity will be discussed.
Table 2 Summary of important metrics in the FW and EIP the
datasets showing ranges for the number of actors in the
system (“smallest,” “median,” and “largest”) and cyclicity
(“best” 5 highest and “worst” 5 lowest) Critical Components for High Cyclicity: Detritus Actors. As
mentioned, cyclicity is a measure of internal cycling in networks.
kmax LD PR G V As energy and materials savings in EIPs are highly dependent on
the successful cycling of waste and by-products, cyclicity is an
Post 1993 FWs median 4.24 5.04 1.09 6.18 5.34 important metric. The lower cyclicity values highlight the less
Worst FW 0 1.59 1.00 1.28 1.28 complex internal cycling present in the structure of EIPs as com-
Smallest FW 2.68 2.67 1.2 3.2 2.67
pared to FWs. The median value of cyclicity for FWs is more than
Largest FW 10.30 9.74 1.01 19.90 19.60
Best EIP-The Green Triangle 3.92 3.25 1.14 3.25 3.92 one and a half times larger than EIPs, indicating that FWs gener-
(class A) ally have a much more complex set of pathways. Although many
Worst EIP-Connecticut Newsprint 0 0.833 0.4 1 2.5 EIPs fall into the category of having at least one single cyclic loop
(class D) that all connected components participate in (cyclicity ¼ 1), a
number of the EIPs show a cyclicity of zero, meaning no cyclic

101002-6 / Vol. 138, OCTOBER 2016 Transactions of the ASME


Fig. 3 A visualization illustrating the difference in complex cy-
clical interactions between a food chain and an FW in nature.
Adapted from Ref. [40].
structure is present. This is essentially a failure on the part of the
EIP designers to reproduce the structure and function of FWs.
High cyclicity values (1) relate strongly to the overall propor-
tion of the energy retained or used within the system versus that
which is lost or discarded by the system. This relationship is also
reflected in an analysis done on thermodynamic power systems
[45]. The results of that analysis suggest that designing EIPs with
a high cyclicity structure may lead to more efficient closed-loop
industrial networks. Despite consumer and financial support, recy-
cling in industrial systems still only accounts for a small fraction
of mobilized matter. Most recycling is in the form of metals col-
lected and shipped to an offsite recycling facility [46], which from
an ecosystem perspective corresponds to the export of material to
other systems. The potential for onsite reuse of water and other
by-products is immense and much better reflects the role of the
detritivores (decomposers) in an ecosystem. These organisms con-
sume waste or dead animals (detritus), which recycles matter and
energy back into the living components as opposed to being
exported.
The internal cycling in FWs is very strongly influenced by the
presence of detritivores. Over half of all the material in an FW is
connected to a decomposer-type species such as fungi, which
recycles unused material or dead matter (detritus) and returns it
back to the system. Decomposers ensure the presence of FW path-
ways that include all other species in the system because the con-
nections due to this consumption pattern contribute to many other
existing cycles. Even limited connections to an actor that func-
tions similarly in an EIP would dramatically increase connectiv-
ity, and thereby efficiency.
A detritus-type actor for an EIP is an actor that functions in
waste treatment (i.e., composting), recovery and recycling (i.e.,
repair, remanufacture, reuse, resale), or agriculture (i.e., farm,
zoo, landscaping, green house, golf course). Additionally, to qual-
ify as a detritus-type actor, there must be at least one link entering
and leaving said actor. This last criterion is based on the funda-
mental functional description of a detritus/decomposer in an FW
and ensures that the detritus-type actor is an active participant of
the EIP.
The types of interactions present clearly influence the magni-
tude of the differences between EIP and FW performance (Table
1); EIPs fall closer to FWs without cannibalism and detrital inter-
actions [38], suggesting that the failure to include such roles in
EIPs is at least partially responsible for their lower cyclicity rela-
tive to FWs. The top six EIPs listed in Table 1 have one or more
detritus-type actors, which can comprise from one-third to one-
half of the total actors in the group. Four of the top EIPs have
some form of composting or agriculture-type actor. The EIPs in
this top group tended to have a larger than average linkage density
as well.
Journal of Manufacturing Science and Engineering
The presence of detrital actors in EIP results in complex cycling
even when fewer connections exist relative to FWs (i.e., linkage
density is lower). The lowest EIP in the top group, Kitakyushu
Resource Recovery Park in Japan, has a low linkage density and
prey–predator ratio in comparison to the rest of the group, while
still having a high cyclicity. The explanation is that all the interac-
tions in the system are to and from only 1 of the 11 actors: the
resource recovery facility, which is the acting detritus. Thus, the
Kitakyushu Resource Recovery Park has 100% of the total links
in the system passing through its detritus actor. Clark Special
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Economic Zone also has a lower linkage density as compared to a
majority of the top EIPs. Of the 51 links between the 20 actors in
the Clark Special Economic Zone, those actors that saw the most
connections were the five composting/processing/recovery facili-
ties and 84% of the total links in this system passed through these
detrital-type actors.
There are six EIPs listed in Table 1 that ranked as D class,
exhibiting zero internal cycling. These EIPs are characteristic of
cyclicity equal to zero, and low linkage density. Connecticut
Newsprint ranks the lowest out of all the EIPs in comparison to
FWs. Interesting is it does in fact have a composting and a recy-
cling component, but these actors fail to provide any benefits with
regard to structure; they each only have one connection with the
rest of the system. Triangle J located in NC, another EIP in this bot-
tom group, has a wastewater treatment plant which interacts with
three other actors; however, similar to Connecticut Newsprint, it
too fails to be an “active-enough” participant to have an impact on
the internal cycling. So, we see it is not enough to simply have a
“detrital” component in an EIP; it must be an active participant in
the system in order to create cycles of materials and energy.
This role of the detritus actor is a key differentiator between
EIPs. For example, EIP Kalundborg ranks in the bottom half of
the C class EIPs, those exhibiting only basic internal cycling.
Comparing Kalundborg to Pomacle–Bazancourt, the top ranking
EIP, Fig. 4, highlights the level of participation of the detritus
actors in each system. All except one of the 15þ cycles in
Pomacle–Bazancourt involve the two detritus actors. Kalundborg
also has two detritus actors. The difference is that only one of the
two detritus actors participates in only two of the three existing
cycles. So, Kalundborg has far fewer cycles because the detritus
actors are disengaged from a majority of the system, while those
EIPs in the top-performing group have a majority of their total
links involved in a cycle and highly involved detritus actors.
Therefore, it is not enough to simply have a detrital component
in an EIP, but it must be an active participant in the system in
order to create cycles of materials and energy. An EIP with no
internal cycling seems contrary to what one expects of a bio-
inspired industrial network because one of the most influential and
identifying characteristics of biological networks is the prevalence
and importance of materials and energy cycling within the system.
Most of the detrital/decomposer actors in the systems examined
here constitute agricultural users. However, industrial recyclers,
wastewater treatment plants, or other actors that reuse and rema-
nufacture products from inputs derived from waste streams of
other EIP actors would fulfill this same role. The important role of
these actors for cycling should encourage industrial networks that
do not currently have recycling-type facilities to invest in them,
and motivate the search for ways to reduce barriers that discour-
age this role with respect to other types of material flows.
Effect of Detritus Actors on EIP Cyclicity. Due to the ecolog-
ical importance of the decomposer/detritus actor, we also quanti-
fied the frequency of their occurrence in each EIP and plotted this
frequency against EIP cyclicity (Fig. 5). The plot area between
zero and one on the x-axis is shaded to indicate that the value of
cyclicity cannot fall between these two limits.
The two data points circled in Fig. 5 have no active detrital
actors but still have a greater than zero cyclicity. These EIPs are
Harjavalta with a cyclicity of 2.0 and the Lower Mississippi
OCTOBER 2016, Vol. 138 / 101002-7
 Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Fig. 4 A comparison of the internal cycling of materials and energy within the Kalundborg
and Pomacle–Bazancourt EIPs. Double-lined arrows represent linkages which participate in a
cycle, grayed out linkages do not. Actors highlighted in bold are the acting detritus of the EIP.

Corridor with a cyclicity of 1.0. The Harjavalta industrial area in
Finland is a full industrial park that includes a wastewater treat-
ment plant and an industrial cleaning facility, but these two com-
panies are not included in the material and energy exchange
diagram provided in the literature [47]. The existence of a detrital-
type actor cannot be discounted as being a contributor to the
success of this EIP as the wastewater treatment plant and cleaning
facility may contribute behind the scenes to the overall structure.
The material and energy exchanges between firms in the industrial
network as documented in the literature are shown in Fig. 6.
Although none of the companies within the network fall into the
functional categories defined above for a detritus-type actor, they
do all meet the active participant requirement, with five of the six
actors having at least one connection entering and leaving. This is
why the cyclicity is so high.
The outlier in Fig. 5 with a cyclicity of 1.0 but no active detritus
is the Lower Mississippi Corridor. Figure 7 gives a visual descrip-
tion of the material and energy exchanges between firms showing
that the cyclicity of the Lower Mississippi Corridor results from
three bidirectional links between three different pairs of actors.
Technically, a bidirectional link (or two actors linked in both
directions) does create a cycle; however, it is not the complex
cycling of ecosystems that EIPs strive to reproduce.
Quantifying Indirect Flows
Ecology distinguishes between direct and indirect relationships
and flows. A direct relationship is one that is formed between
actors that are linked by a material or energy flow. An indirect
relationship is formed if two actors interact through at least one

101002-8 / Vol. 138, OCTOBER 2016 Transactions of the ASME


Fig. 5 The number of active detritus in an EIP plotted against
network cyclicity
intermediary so that material or energy flows through at least one
other actor [48]. This can be, but does not require spatial or tem-
poral separation. Per these definitions, indirect relationships are
also prevalent in manufacturing systems. For example, the rela-
tionship between original equipment manufacturer and a tier 2
supplier is indirect because it goes through a tier 1 supplier first.
Although many EIPs focus on direct flows from one actor to
another, ecologists have noted the significance of indirect flows,
i.e., flows between actors that have indirect relationships. The
Fig. 7 The Lower Mississippi Corridor EIP. Double-lined linkages indicate connections which participate in a cycle, while
gray linkages do not. Grayed boxes indicate an actor which exclusively participates in incoming or outgoing interactions (is
only a predator or predator or prey).
Journal of Manufacturing Science and Engineering
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Fig. 6 The Harjavalta industrial area in Finland. Double-lined
linkages indicate connections which participate in a cycle, gray
linkages do not. The single grayed box (city of Harjavalta) indi-
cates an actor which exclusively participates in incoming or
outgoing interactions (is only a predator or prey).
ecologists Salas and Borrett found in a set of 50 FWs that when
significant cycling was present indirect flows nearly always domi-
nated direct flows [49]. This and other literature over the last 20
years has established the dominance of indirect effects in ecosys-
tems [48,50]. The characteristic cycling of materials and energy in
FWs is one of the most desirable properties to sustainably minded
industry networks. The apparent relationship between cycling in
the system and indirect effects should also be a design property of
interest for industrial networks and can be examined to enhance
understanding and increase material cycling in manufacturing
networks.
OCTOBER 2016, Vol. 138 / 101002-9
 Indirect flow effects can be determined by looking at paths of
length greater than one. A path is the route traced by following
some quantity of material or energy and is made up of either
chains or cycles. A path with a length greater than one indicates
that the material or energy being followed interacts with more than
two actors in the system. The two methods for path formation are
chains and cycles. Both methods limit flows through transfer effi-
ciencies relating to dissipation and export and chains apply an
additional limitation by way of their length [17,51].
Graph theory enables the calculation of the number of paths of
different lengths in the system by raising the adjacency matrix [A]
to a power that represents the path length being investigated
[5,52]. Thus, [A]4 provides all the paths in the network repre-
sented by [A] that have a length of four. This can also be done for
what is known as the flow intensity matrix [G]. The flow intensity
matrix highlights the amount of flow (kg, kJ, units, etc.) that is
indirectly circulated through the system (circulated using paths of
length greater than one).
A number of distinct patterns have arisen from the investigation
of indirect effects in ecosystems. The tendency for the number of
paths to increase geometrically without bound as path length
increases, known as pathway proliferation, was first applied to the
study of ecosystems in the early 1980s [52–54] and has been stud-
ied more recently in FWs [17,32]. Pathway proliferation only
occurs if there is more than one cycle in the network. As noted
previously, the rate of increase in the number of paths with path
length is measured by cyclicity.
The relative magnitude of cyclicity can be used as a descriptor
of indirect flows in the network. Pathway proliferation has a
strong influence on the development and significance of indirect
flows [17]. A faster rate of pathway proliferation, or a higher
cyclicity, signifies that indirect pathways are more numerous.
Because indirect pathways that involve multiple actors will pro-
cess larger flows, a higher cyclicity increases the possibility that
indirect flows will dominate direct flows [17].
Paths of specific lengths can be found by raising the adjacency
matrix to a power that represents the path length being investi-
gated [5,52]. Thus, to find paths of length two or greater, the ma-
trix [A] is raised to the powers 2, 3, 4, and so on. Figure 8 shows
path lengths of 1–100 for the 48 EIPs investigated. These were
calculated by raising each of the adjacency matrices of the 48
EIPs to the powers 1–100. Each line in Fig. 8 represents an EIP.
The pathway proliferation rate for FWs has been shown to
increase with a power-law degree distribution [17,32,53,54]. This
relationship can be seen in Fig. 8 inset (a) [52]. The curves for
those EIPs with the highest cyclicity, those curves on top in Fig.
8, most strongly resemble the curve found by Patten for ecosystem
behavior shown Fig. 8 inset (a) [52].
Figures 9 through 11 show the pathway proliferation rate of the
48 EIPs plotted in Fig. 8, broken down by EIP rank using
Fig. 8 Path length versus number of paths of that length
totaled for each of the 48 EIPs, for paths of length 1–100, plotted
on a log–log scale. Inset (a) is the path length to number of
paths relationship for FWs as presented by Ref. [52].
101002-10 / Vol. 138, OCTOBER 2016
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Fig. 9 Path length versus number of paths of that length for
the six highest EIPs ranked by cyclicity (class A; cyclicity > 3)
for paths of length 1–100, plotted on a log–log scale
cyclicity. The goal of this break down is to show that the relation-
ship between indirect path lengths and the presence and strength
of cycles as seen for FWs occurs in industry.
Figure 9 plots the top six EIPs, designated as group A, which
have a cyclicity of three or greater. The figure clearly shows that
the EIPs with a relatively high cyclicity display a rapid rate of
increase in the number of paths with path length. The power-law
degree distribution seen for these EIPs in group A closely matches
the FW behavior. This topological similarity means that the net-
work robustness to random node deletion that has been related to
this structure [55,56] may be translated for those EIPs that closest
match the cyclicity seen in FWs. The metric robustness is a
parameter that requires knowledge of the quantities of materials
and energy flowing between actors, information that is not cur-
rently available for EIPs.
Figure 10 shows the class B EIPs, those with cyclicity greater
than one, but note that the rate of increase is not as rapid as that
for the top-tier EPS. This is another supporting factor for EIP
designers to strive for higher cyclicity values in their networks;
EIPs with higher cyclicity will have a structure that supports a
level of dominance of indirect flows that is on par with what is
seen in FWs.
Figure 11 shows the class C and class D EIPs that have a cyclic-
ity of one or zero. Figure 11-left shows that for those EIPs with
some form of basic cycling present in their structure, there is no
guarantee that the number of paths in the system will be able to
increase with path length. Figure 11-right shows that with no
cycling there is no pathway proliferation. Figures 8 through 11
confirm that pathway proliferation can only occur when there is
more than one cycle in the network, and that this can be confirmed
by measuring a cyclicity that is greater than 1. Thus, cyclicity is a
Fig. 10 Path length versus number of paths of that length for
the medium–high cyclicity EIPs (class B; 1 < cyclicity < 3) for
paths of length 1–100, log–log scale
Transactions of the ASME
 Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
Fig. 11 Path length versus number of paths of that length for the lowest cyclicity EIPs. Left
panel shows class C EIPs (cyclicity 5 1) and right panel shows the class D EIPs (cyclicity 5 0)
for paths of length 1–100, plotted on a log–log scale.

key metric for assessing and improving direct as well as indirect proliferation and indirect links is novel in the context of closed-
material flows in manufacturing networks. loop manufacturing systems.

Acknowledgment
Summary and Conclusion This material is based upon work supported by the National
Ecological principles used for understanding ecological FW Science Foundation under Grant Nos. CMMI-0600243, CBET-
structures are relevant to improving manufacturing networks. 0967536, and CBET-1510531. Any opinions, findings, and con-
Groupings of EIPs were made in terms of both their economic sta- clusions, or recommendations expressed in this material are those
tus (proposed, existing, or failed) and the level of internal cycling of the author(s) and do not necessarily reflect the views of the
in the network structure (high, medium, basic, and none) based on National Science Foundation.
the metric cyclicity. When examined using selected structural FW
metrics commonly used by ecologists for FW analysis, the ana- Nomenclature
lyzed groupings create a more complete perspective between each
other and biological FWs in terms of their success in being “bio- EIP ¼ eco-industrial park
inspired.” None of the systems, despite their status, successfully FW ¼ food web
match the average values found for biological ecosystems. Based G¼ generalization
upon these results, it is clear that the biological ecosystem, in the L¼ number of links
sense of the aforementioned structural metrics, has yet to be fully Ld ¼ linkage density
mimicked by industrial networks. For example, it is unlikely that Pr ¼ predator to prey ratio
high cyclicity values can be achieved in EIPs that lack actors ful- S¼ number of species
filling the role of detritus/decomposers. This suggests that EIP V¼ vulnerability
designers must incorporate analogous interactions in their indus- kmax ¼ cyclicity
trial networks to achieve the strong cycling characteristic of FWs.
Many approaches to sustainability emphasize this as a values- References
driven act that provides for individual and societal well-being. [1] Farrokhi, F., Siltanen, J., and Salminen, A., 2015, “A Semantic Information
This is undoubtedly true, but values and motivation themselves Services Framework for Sustainable WEEE Management Toward Cloud-Based
do not constitute a framework to determine how well a circular Remanufacturing,” ASME J. Manuf. Sci. Eng., 137(6), p. 061012.
[2] Linke, B. S., 2015, “Review on Grinding Tool Wear With Regard to Sus-
economy functions, although they may encourage the search for tainability,” ASME J. Manuf. Sci. Eng., 137(6), p. 060801.
appropriate solutions. This paper demonstrates that ecological [3] Allen, D. T., Bauer, D. J., Bras, B., Gutowski, T. G., Murphy, C. F., Piwonka,
networks composed of predator prey linkages can be used to T. S., Sheng, P. S., Sutherland, J. W., Thurston, D. L., and Wolff, E. E., 2002,
interpret and guide the construction of sustainable industrial “Environmentally Benign Manufacturing: Trends in Europe, Japan and the
USA,” ASME J. Manuf. Sci. Eng., 124(4), pp. 908–920.
organizations through analysis of structure–function relationships. [4] Gutowski, T. G., Murphy, C. F., Allen, D. T., Bauer, D. J., Bras, B., Piwonka,
Currently, none of the EIPs identified come close to matching T. S., Sheng, P. S., Sutherland, J. W., Thurston, D. L., and Wolff, E. E., 2001,
median amounts of cycling seen in FWs. Cyclicity is also a mea- “Environmentally Benign Manufacturing,” WTEC Panel Report, International
sure of indirect cycling and pathway proliferation rate, i.e., the Technology Research Institute, World Technology (WTEC) Division, Balti-
more, MD.
rate that the number of paths increases as path length increases. [5] Roberts, F. S., 1976, Discrete Mathematical Models: With Applications to Social,
Higher cyclicity values correspond to the greater amount of indi- Biological, and Environmental Problems, Prentice-Hall, Englewood Cliffs, NJ.
rect cycling and pathway proliferation rates. This is important [6] Strogatz, S. H., 2001, “Exploring Complex Networks,” Nature, 410(6825),
because pathway proliferation rate is representative of indirect pp. 268–276.
[7] Graedel, T. E., 1996, “On the Concept of Industrial Ecology,” Annu. Rev.
links in the system, and it has been shown that in FWs when sig- Energy Environ., 21(1), pp. 69–98.
nificant cycling was present, indirect flows were nearly always [8] Hess, G., 2010, “The Ecosystem: Model or Metaphor? Epistemological Diffi-
found to dominate direct flows. The analysis of pathway culties in Industrial Ecology,” J. Ind. Ecol., 14(2), pp. 270–285.

Journal of Manufacturing Science and Engineering OCTOBER 2016, Vol. 138 / 101002-11
 [9] Odum, E. P., 1969, “The Strategy of Ecosystem Development,” Science,
164(3877), pp. 262–270.
[10] Pimm, S. L., 1982, Food Webs, Chapman and Hall, London, UK.
[11] Pimm, S. L., 2002, Food Webs, The University of Chicago Press, Chicago, IL.
[12] Cohen, J. E., 1982, Food Webs and Niche Space, Princeton University Press,
Princeton, NJ.
[13] Pimm, S. L., Lawton, J. H., and Cohen, J. E., 1991, “Food Web Patterns and
Their Consequences,” Nature, 350(6320), pp. 669–674.
[14] Schoener, T. H., 1989, “Food Webs From the Small to the Large,” Ecology,
70(6), pp. 1559–1589.
[15] Pimm, S. L., 1979, “The Structure of Food Webs,” Theor. Popul. Biol., 16(2),
pp. 144–158.
[16] May, R. M., 1973, Stability and Complexity in Model Ecosystems, Princeton
University Press, Princeton, NJ.
[17] Borrett, S. R., Fath, B. D., and Patten, B. C., 2007, “Functional Integration of
Ecological Networks Through Pathway Proliferation,” J. Theor. Biol., 245(1),
pp. 98–111.
[18] Cohen, J. E., Beaver, R. A., Cousins, S. H., DeAngelis, D. L., Goldwasser, L.,
Heong, K. L., Holt, R. D., Kohn, A. J., Lawton, J. H., Martinez, N., O’Malley,
R., Page, L. M., Patten, B. C., Pimm, S. L., Polis, G. A., Rejmanek, M., Scho-
ener, T. W., Schoenly, K., Sprules, W. G., Teal, J. M., Ulanowicz, R. E., War-
ren, P. H., Wilbur, H. M., and Yodzis, P., 1993, “Improving Food Webs,”
Ecology, 74(1), pp. 252–258.
[19] Polis, G. A., and Strong, D. R., 1996, “Food Web Complexity and Community
Dynamics,” Am. Nat., 147(5), pp. 813–846.
[20] Bhatia, G., Lane, C., and Wain, A., 2013, “Building Resilience in Supply
Chains: An Initiative of the Risk Response Network in Collaboration With
Accenture,” REF150113, World Economic Forum.
[21] Fath, B. D., 2007, “Structural Food Web Regimes,” Ecol. Modell., 208(2–4),
pp. 391–394.
[22] Ulanowicz, R. E., 1997, Ecology, the Ascendent Perspective, Columbia Univer-
sity Press, New York.
[23] Korhonen, J., and Sn€akin, J.-P., 2005, “Analysing the Evolution of Industrial
Ecosystems: Concepts and Application,” Ecol. Econ., 52(2), pp. 169–186.
[24] Briand, F., 1983, “Environmental Control of Food Web Structure,” Ecology,
64(2), pp. 253–263.
[25] May, R. M., 1972, “Will a Large Complex System be Stable?,” Nature,
238(5364), pp. 413–414.
[26] Cohen, J. E., Briand, F., and Newman, C. M., 1990, Community Food Webs:
Data and Theory, Springer, New York.
[27] Montoya, J. M., S. L. Pimm, and R. V. Sole, 2006 “Ecological Networks and
Their Fragility,” Nature, 442, pp. 259–264.
[28] Yodzis, P., 1982, “The Compartmentation of Real and Assembled Ecosystems,”
Am. Nat., 120(5), pp. 551–570.
[29] Briand, F., and Cohen, J. E., 1987, “Environmental Correlates of Food Chain
Length,” Science, 238(4829), pp. 956–960.
[30] Fath, B. D., 2007, “Structural Food Web Regimes,” Ecol. Modell., 208(2–4),
pp. 391–394.
[31] Warren, P. H., 1990, “Variation in Food Web Structure: The Determinants of
Connectance,” Am. Nat., 136(5), pp. 689–700.
[32] Fath, B. D., and Halnes, G., 2007, “Cyclic Energy Pathways in Ecological Food
Webs,” Ecol. Modell., 208(1), pp. 17–24.
[33] Bascompte, J., and Jordano, P., 2007, “Plant-Animal Mutualistic Networks: The
Architecture of Biodiversity,” Annu. Rev. Ecol., Evol., Syst., 38(1), pp. 567–593.
[34] Allesina, S., Bondavalli, C., and Scharler, U. M., 2005, “The Consequences of
the Aggregation of Detritus Pools in Ecological Networks,” Ecol. Modell.,
189(1–2), pp. 221–232.
101002-12 / Vol. 138, OCTOBER 2016
[35] Fath, B. D., 1998, “Network Analysis: Foundations, Extensions, and Applica-
tions of a Systems Theory of the Environment,” Ph.D. dissertation, University
of Georgia, Athens.
[36] McCann, K., 2012, Food Webs, Princeton University Press, Princeton, NJ.
[37] Husar, R. B., 1994, “Ecosystem and the BiosphereE: Metaphors for Human-
Induced Material Flows,” Industrial Metabolism: Restructuring for Sustainable
Development, R. U. Ayres, and U. E. Simonis, eds., United Nations University
Press, Tokyo, Japan, pp. 21–29.
[38] Layton, A., Bras, B., and Weissburg, M., 2015, “Industrial Ecosystems and
Food Webs: An Expansion and Update of Existing Data for Eco-Industrial
Parks and Understanding the Ecological Food Webs They Wish to Mimic,” J.
Ind. Ecol., 20(1), pp. 85–98.
Downloaded from https://www.asmedigitalcollection.asme.org/manufacturingscience/article-pdf/138/10/101002/6404700/manu_138_10_101002.pdf by Gadjah Mada University user on 21 February 2020
[39] Kambhu, J., Weidman, S., and Krishnan, N., 2007, “Part 3: Systemic Risk in
Ecology and Engineering,” Economic Policy Review, 13(2), pp. 25–40.
[40] deCharon, A., 2013, “Fitting Algae Into the Food Web,” Bigelow Laboratory
for Ocean Sciences, East Boothbay, ME.
[41] Strauss, S. Y., 1991, “Indirect Effects in Community Ecology Their Definition
Study and Importance,” Trends Ecol. Evol., 6(7), pp. 206–210.
[42] Halnes, G., Fath, B. D., and Liljenstr€
om, H., 2007, “The Modified Niche Model:
Including Detritus in Simple Structural Food Web Models,” Ecol. Modell.,
208(1), pp. 9–16.
[43] Chen, S., and Chen, B., 2012, “Network Environ Perspective for Urban Metab-
olism and Carbon Emissions: A Case Study of Vienna, Austria,” Environ. Sci.
Technol., 46(8), pp. 4498–4506.
[44] Chen, S., Chen, B., and Su, M., 2015, “Nonzero-Sum Relationships in Mitigat-
ing Urban Carbon Emissions: A Dynamic Network Simulation,” Environ. Sci.
Technol., 49(19), pp. 11594–11603.
[45] Layton, A., Reap, J. J., Bras, B., and Weissburg, M., 2012, “Correlation
Between Thermodynamic Efficiency and Ecological Cyclicity for Thermody-
namic Power Cycles,” PLoS One, 7(12), pp. 1–7.
[46] Ji, G., 2008, “Closed-Loop Supply Chains Based on by Product Exchange,”
IEEE International Conference on Service Operations and Logistics, and Infor-
matics (SOLI), Beiging, China, pp. 2405–2410.
[47] Heino, J., and Koskenkari, T., 2004, “Industrial Ecology and the Metallurgy
Industry. The Harjavalta Industrial Ecosystem,” Waste Minimization and
Resources Use Optimization Conference, E. Pongracz, ed., Oulu University
Press, University of Oulu, Finland, pp. 143–151.
[48] Higashi, M., and Patten, B. C., 1989, “Dominance of Indirect Causality in
Ecosystems,” Am. Nat., 133(2), pp. 288–302.
[49] Salas, A. K., and Borrett, S. R., 2011, “Evidence for the Dominance of Indirect
Effects in 50 Trophically-Based Ecosystem Networks,” Ecol. Modell., 222(5),
pp. 1192–1204.
[50] Wootton, J. T., 1994, “The Nature and Consequences of Indirect Effects in Eco-
logical Communities,” Annu. Rev. Ecol., Evol., Syst., 25(1), pp. 443–466.
[51] Borrett, S. R., Whipple, S. J., and Patten, B. C., 2010, “Rapid Development of
Indirect Effects in Ecological Networks,” Oikos, 119(7), pp. 1136–1148.
[52] Patten, B. C., 1985, “Energy Cycling in the Ecosystem,” Ecol. Modell., 28(1–2),
pp. 1–71.
[53] Patten, B. C., 1985, “Energy Cycling, Length of Food Chains, and Direct Versus
Indirect Effects in Ecosystems,” Can. Bull. Fish. Aquat. Sci., 213, pp. 119–138.
[54] Patten, B. C., Richardson, T. H., and Barber, M. C., 1982, “Path Analysis of a
Reservoir Ecosystem Model,” Can. Water Resour. J., 7(1), pp. 252–282.
[55] Albert, R., Jeong, H., and Barabasi, A. L., 2000, “Error and Attack Tolerance of
Complex Networks,” Nature, 406(6794), pp. 378–382.
[56] Dunne, J. A., Williams, R. J., and Martinez, N. D., 2002, “Network Structure
and Biodiversity Loss in Food Webs: Robustness Increases With Connectance,”
Ecol. Lett., 5(4), pp. 558–567.
Transactions of the ASME