You are on page 1of 10

Building and Environment 45 (2010) 1006–1015

Contents lists available at ScienceDirect

Building and Environment


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

Use of system dynamics as a decision-making tool in building design


and operation
Benjamin P. Thompson*, Lawrence C. Bank
Civil and Environmental Engineering Department, University of Wisconsin–Madison, 1415 Engineering Drive, Madison, WI 53706, USA

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

Article history: This paper presents a decision-making tool based on the system dynamics method for building design
Received 3 June 2009 and operation, and demonstrates the use of the method for the analysis of a building subjected to
Received in revised form a bioterrorist attack. A literature review of prior uses of the system dynamics method, focusing mainly on
29 September 2009
historic uses within civil and environmental engineering and related disciplines is first presented, to
Accepted 13 October 2009
make the case for the applicability of the system dynamics method as a decision-making tool for building
system design, retrofit, and operation. A proof-of-concept system dynamics model is presented for
Keywords:
modeling of an elementary building system subjected to a bioterrorist attack. The proof-of-concept
Decision-making
Buildings model includes evaluation of modifications to the building and its defenses. The results of the model
Terrorism simulation are presented and compared with previously published data. Use of the output data to
System dynamics improve decision-making in building design, retrofit, and operation is discussed. A method to link an
Building information modeling electronic building information model (BIM) of the building to enable the electronic capture of the
relevant building features in the system dynamics model is also discussed and demonstrated. Finally,
future development of the model and other potential areas of applicability for the modeling method-
ology are discussed.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction mechanical design) should be augmented by tools that enable the


interaction of the different discrete sub-system models. This can be
The design of a building requires the participation of many accomplished by using the system dynamics method that allows
different professional disciplines including, for example, structural, the building to be modeled as a feedback system to simulate the
mechanical and electrical engineering, architecture, landscape interactions amongst the various building sub-systems. It is appli-
architecture, security design professions, and law enforcement. In cable to building system simulation because it is well-suited to
addition, the design for unpredictable, anthropogenic hazards such situations where the system to be modeled is extremely complex,
as terrorism includes recognizing that these design-drivers are highly dynamic, or contains a large number of feedbacks. A system
difficult to accurately quantify. During its operation, a building’s dynamics model is focused on fundamental attributes of a system
performance is a function of its components (sub-systems) working and allows for highly uncertain variables to be included.
together in a complex fashion. The actual performance of a building The representation and simulation of a building in a single,
or its sub-systems is not accurately known until long after a design integrated systems model is challenging due to the large number of
decision is made. Other performance measurements of a building sub-systems and components in a building and the complex
system require that trade-offs be made among options that are interactions among these components [1]. Currently, building
difficult to compare in terms of a single metric (for example, models separately represent performance attributes such as
building sustainability). Therefore, traditional sub-system special- energy, light, sound, temperature, fire, safety, and air quality, or
ized analysis, modeling and simulation tools used by building physical attributes such as space, structure, mechanical systems,
designers (e.g., for architectural design, structural design, electrical systems and building envelopes. A major goal of the
building simulation community is the integration of these discrete
models into a single building model, and the optimal use of the
* Corresponding author at: Civil and Environmental Engineering Department,
University of Wisconsin–Madison, 1415 Engineering Drive, Room 1203, Madison,
results of these models in decision-making [1]. A system dynamics
WI 53706, USA. Tel.: þ1 608 262 1262; fax: þ1 608 262 5199. model can provide a framework for integrating these separate
E-mail address: bpthompson@wisc.edu (B.P. Thompson). models into a decision-making process, as illustrated in Fig. 1.

0360-1323/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2009.10.008
B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015 1007

information that must be assembled to manage the life cycle of civil


Architectural infrastructure systems, including the need to ‘‘capture, clarify, and
Model
predict the complex behaviors and interdependencies of infra-
structure systems,’’ to maximize day-to-day efficiency while
Structural Budget
minimizing the vulnerability to catastrophic extreme events, and to
Model Model determine which measures of performance characterize the
performance of the building as a whole [5]. A system dynamics
HVAC Energy Use

room
volume
~
filter
efficiency
rating
airflow
velocity model allows for more effective trade-off analyses among different
airflow drop

design options needed to achieve specific building performance


from filter detection design response
baseline
filter time
airflow exhausted exhaust fraction
efficiency
electrical system air
capacity HVAC Power Demand
HVAC

characteristics, such as security, energy consumption, or produc-


airflow speed response tim
Available Energy
volume per time

Power Provided SD Model outside air


intake typical

tivity. By utilizing user-modifiable weighting parameters it is


Other non HVAC exhaust fraction
Power Demands
Number of
exhaust fraction Incident
lighting efficiency Occupants
Bioagent
factor
Level

possible to compare the effects of different building design vari-


Lighting Power Demand

Mechanical
lighting occ
occupancy
type factor
type air

Energy ables that are traditionally viewed as not comparable.


System
Model
Model The true value of using a system dynamics method to model
building design, operation, and retrofit problems before imple-
mentation is the opportunity to examine ‘what-if’ scenarios before
allocating scarce time, budget, and personnel resources to solve the
Indoor Air problem. Utilizing a system dynamics model to assist in making
Quality Model building design and operations decisions allows the user to
perform trade-off analyses among various courses of action, and to
Fig. 1. Using SD to link building simulation programs. predict the effects of a course of action or modification in one of the
building sub-systems on the overall performance of the building.
In what follows, a review of the use of the system dynamics
Information can flow back and forth between the system dynamics method, with a focus on prior uses in the civil engineering disci-
model and the various building sub-systems models. Information pline, is presented. Following this review, a proof-of-concept
from the sub-system design and simulation models can be used to system dynamics model developed to make decisions regarding
populate a system dynamics model to aid in making decisions building design and operation to minimize the effects of a bio-
related to building design, operation or retrofit. In addition, terrorist attack is presented and future work to develop the model
a system dynamics model can be used in conjunction with various is discussed. Finally, the possibilities for future uses and additional
simulation models to determine preferred values of parameters to developments of such a system dynamics model are presented.
use in a design model to achieve an overall system-wide perfor-
mance objective.
Additional building performance complexity is caused by the 2. Background
interaction of building occupants with the various building sub-
systems during the operation and use phase of a building. The System dynamics (SD) is a modeling method that allows
interaction between buildings and their occupants has been a system (in this case a building) to be represented as a feedback
studied using agent-based simulations of occupants, mostly in the system. It is based on the original work of Forrester, who defined it
area of egress analysis and crowd dynamics [2,3]. Agent-based as ‘‘the investigation of the information-feedback character of
simulations differ from system dynamics in that an agent-based industrial systems and the use of models for the design of improved
model ‘tracks’ every agent (for example, each building occupant organizational form and guiding policy’’ [6]. The model is ‘‘an
exiting the building). System dynamics models, on the other hand, interlocking set of differential algebraic equations developed from
aggregate agents in groups and represent these groups with a broad spectrum of relevant measured and experiential data’’ [7].
specific values (for example, the group of occupants in a room or The equations are represented by a diagram, shown schematically
the group of occupants leaving a room). in Fig. 2, consisting of three element types: (1) stock (or level)
Another complicating factor is that often key design decisions elements (also called state variables); (2) flow elements; and
must be made early in the building design process, when infor- (3) auxiliary variables and constants [8]. SD models depend on the
mation about the final design and operation of the building or structure of the model, time lags, and amplification, which occurs
building system is unavailable, but the ability to affect cost is the through feedback [6]. An SD model allows examination of the long-
greatest. ‘‘The challenge is thus finding a method to use detailed term behavior of complex systems [9]. System dynamics is a direct
simulation tools even during the early stages of design when values descendant of the general systems theory [10], which is a holistic
for many of the variables for the building’s technical sub-systems philosophy of science and engineering based on the principle that
are not yet available’’ [4]. This requires finding ways to provide reducing a system to multiple small, individual pieces for study will
a designer with ‘‘feedback’’ on the building’s performance prior to cause the underlying behavior of the system to escape notice, but
its construction.
However, even combining all the existing building sub-system
models into a single mega-model will not provide an entirely
integrated view of the building as a system, unless the feedbacks
amongst the sub-system models are included in the combined auxiliary
model. Heller asserts that many infrastructure systems follow variable
‘‘patterns that result, yet are not analytically predictable from,
dynamic, non-linear, spatio-temporal interactions among a large
number of components or sub-systems’’ [5]. The same is true of Stock
a building, as a geographically small, but complex aggregation of Inflow Outlfow
sub-systems into an essentially self-contained environment which
has interactions with its surroundings. Heller also discusses the key Fig. 2. Schematic of a system dynamics model.
1008 B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015

studying only the macroscopic behavior of a system while ignoring structures that affect the policy threshold to allow different regu-
the details of the components will cause a lack of understanding of latory environments to be tested (e.g., penalties for non-compli-
the fundamental causes of the model’s behavior [11]. ance, avenues for complaints, community values outcomes, and
predictive quality of the indicator). The study concluded with
2.1. General applications of system dynamics a suggestion that a similar system could be used for natural
disasters, such as when to offer warnings or require evacuations for
The system dynamics method has been used in a wide variety of hurricanes and tsunamis [38]. Sim et al. [39] used SD to model
applications, both in the social sciences and in engineering. It has a network of chemical manufacturing plants and associated infra-
been used to model topics as diverse as political instability [12], structure. The model was then used to examine effects of potential
conflict management [13], organization in the U.S. space program disruptions in the network or its associated infrastructure.
[14], historical trends in world architecture [15], building material A general model template was used for initial modeling of the
resource availability [16], land reclamation in the mining industry plants. The most critical components were identified through this
[17], sustainable cement production [9,18,19], energy and power initial step, and these components were then focused on in greater
systems [20], coastal ecosystem dynamics [21], aviation systems detail in subsequent steps [39].
[22], solid waste forecasting [23], agenda setting and public policy
making [24], and industrial engineering applications, such as 2.2. Civil and environmental engineering applications
supply chain management (e.g., ref. [25]). The system dynamics of system dynamics
method has also been widely used for modeling environmental
systems. Ford [26] includes examples of salmon migration, fisheries System dynamics modeling has also been used in a variety of
management, wildlife population dynamics, air pollution, vehicle applications related to civil and environmental engineering. The
emissions, and environmental policy. Meadows et al. [27], and area of civil engineering that has most commonly used the SD
Randers [28] used SD as a method to model the impact of human method is construction project management, where it has been
population on the earth and the effects of population on resources. used, for example, to study performance enhancement of
Another area of wide application is in health systems. For example, a construction organization [40], effects of project personnel
Koelling and Schwandt [29] reviewed and summarized SD litera- changes [41], delay and disruption claims [42], error and change
ture on health care, addressing topics such as organization of health management [43], quality management [44], rework [45], the
systems, clinical research, delivery, disease prevention, epidemi- design-build process [46,47], conflict management [48,49], and
ology, and dentistry. Modeling of the interactions of health risks, road maintenance budgeting [35]. Sustainable construction has
responses, affected populations and government policy has also also been studied using SD modeling. Shen et al. developed a SD
been undertaken using SD models [7]. Use of the method is also model to assess the sustainable performance of projects using
growing for military and defense modeling. This includes subjects a triple bottom line of: (1) economic; (2) social; and (3) environ-
like preparedness and training [30–32], and capital equipment mental performance [50]. An area of early and continued applica-
management [33]. Smith makes the case that SD methods could tion of SD modeling was in urban planning, development, and land
and should be used for counter-terrorism simulations [34]. Allo- use. One of Forrester’s early works in the SD area modeled urban
cation of finite resources has also been addressed in the SD litera- growth and decay [51]. The earliest reference to SD located in the
ture. For example, Bjornsson et al. [35] modeled allocations in Civil Engineering literature discussed issues related to urban
a road maintenance budget using SD and Linard et al. [31] exam- planning [52]. Later, Drew created a model to illustrate interactions
ined defense budgets. System dynamics has already been used in among four major civil systems: (1) socio-technological; (2) water;
research focused on natural disasters and terrorist attacks. The idea (3) energy; and (4) transportation-land use [53]. Water resources
of modeling a system and then introducing a disruption has also and systems are also a growing area of application. One early study
been studied. Deegan [24] modeled decision-making and agenda applied Forrester’s methodology to urban water supply problems
setting for response to natural disasters, with the intent of using the and developed a case study of the Fort Collins, CO water system
model, ‘‘to find points of leverage in the system, where policy [54]. More recent studies have been undertaken to study water
efforts can be more effective, preventing future damage rather than quality [55], water resource availability and use [56], water
just clearing current damage’’ and to ‘‘develop a tool that admin- resource systems [57,58], flood management [59], and the response
istrators and policy analysts can use to understand the sources of of residents to a flood in their area [60]. Car-following models to
opposition to different policy solutions’’ [24]. This model addressed assess the impact of different intelligent transportation systems
a perceived disconnect between how problems are identified, how technologies or controls on safety and traffic flow have also been
solutions are arrived at, and how policies are implemented based developed using SD methods and software [61].
on the solutions. This work was extended to combine policy anal- Analyses using SD methods have also been used to study
ysis research and policy process concepts [36]. Brown et al. also infrastructure systems, as well as disruptions to those systems.
used a SD method to examine natural disaster assessment and Infrastructure systems, such as transportation and water supply
response during a disaster event [37]. and treatment systems were modeled by Drew [53]. Chasey et al.
Little and Weaver [38] used three social and policy sciences tools examined highway management [62], and de la Garza et al. studied
to develop an indicator for level of safety, a threshold for that budget allocation policies for highway maintenance programs
indicator to be used in setting policy, and a regulatory environment using an SD model [63].
that responds to changes in that threshold: (1) judgment analysis; The National Infrastructure Simulation and Analysis Center
(2) a Taylor Russell diagram; and (3) a SD model. The resultant (NISAC) at Sandia and Los Alamos National Labs1 has used SD
methodology attempted to recognize the difference and impor- methods to improve planning simulation and decision support for
tance of both technical and social elements in analyzing the risk of critical infrastructure analyses. Specifically, NISAC was established
terrorist threats. The first two tools were used to arrive at inputs to to provide a more comprehensive understanding of the interactions
the SD model, which was then used to determine how changes in between the various infrastructure systems, in addition to the
threshold levels for action affect security policies. For instance, as
buildings become safer, the public will demand higher levels of
1
safety. The model included the legal and political regulatory http://www.lanl.gov/news/pdf/TRCounterTerror_NISAC.pdf.
B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015 1009

Complete SD model Simulate disruption to Examine baseline


for overall building overall system (terrorist performance of the
(chosen systems) attack, natural disaster, building in response to
power blackout, etc). disruption

Determine sub-
Economic/ systems/components that Alter pre-selected
Security/ have the greatest impacts building variables to
Social on mitigating effects of the determine their effects
Trade-offs disruption.

Fig. 3. Model use methodology

existing knowledge that has been used to model individual systems decision-makers in prioritizing infrastructure protection measures
in the past. The NISAC has identified key pieces of infrastructure under budget constraints. The SD model interacted with a database
that must be protected or backed up, but is also working to improve and a decision support computer program to determine the effec-
normal functioning of the infrastructure systems. Smith used the tiveness of each alternative, which was measured with respect to
term ‘‘critical service sustainment’’ rather than ‘infrastructure six categories: (i) sector-specific; (ii) human health and safety; (iii)
protection’ [64]. However, buildings are notably lacking from the economic; (iv) environmental; (v) socio-political (perceived risk,
list of critical infrastructure sectors and systems. In this Critical public confidence, trust in government sector-specific effects, and
Infrastructure Protection Decision Support System (CIP/DSS) market confidence); and (vi) national security [65]. The model used
Project being undertaken at Los Alamos and Sandia National a base case, which was then compared pairwise with various
Laboratories, an SD simulation method was used to conduct risk scenarios (policy or resource changes) involving a disruption to the
analyses and consequence assessments with the intent of assisting infrastructure to measure the effectiveness of each scenario. The

Fig. 4. SD building model air system sector.


1010 B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015

Table 1
Simulation of baseline scenario – Kowalski data.

AR = (TA * (1 − OA)) * c * F + (TA * (1 − OA) * c * (1 − F )) * U Time Microbes Microbes in Building Microbes


(h CminD) released (cfu) building (cfu) concentration exhausted
Where: (cfu/m3) (cfu)
AR = Bio-agent removed through recirculation
TA = Total airflow (cfm) (1) (2) (3) (4) (5)
OA = Outside air (%) Agent Remov ed
C = bio-agent concentration (cfu/m3) 0.00 C0D 0 0 0 0
Through Recirculation 0.02 C1D 18,000,000 18,000,000 286.74 353,376
F = Filter removal rate (%)
U = UVGI (ultra-violet germicidal irradiation) removal rate (%) 0.03 C2D 18,000,000 35,646,624 567.85 699,815
0.05 C3D 18,000,000 52,946,808 843.44 1,039,453
0.07 C4D 18,000,000 69,907,356 1113.62 1,372,423
0.08 C5D 18,000,000 86,534,933 1378.49 1,698,855
Fig. 5. Sample variable from air system sector 0.10 C6D 18,000,000 102,836,078 1638.17 2,018,880
0.12 C7D 18,000,000 118,817,198 1892.75 2,332,621
0.13 C8D 18,000,000 134,484,576 2142.33 2,640,204
project has moved into the sensitivity analysis phase. One portion 0.15 C9D 18,000,000 149,844,373 2387.01 2,941,748
0.17 C10D 18,000,000 164,902,625 2626.88 3,237,371
of the CIP/DSS being developed at Sandia National Laboratory uses
economic costs and potential savings to rank alternative preven-
tative measures by running multiple versions of the model. The
goal is to ‘‘investigate and understand the non-equilibrium, non- occupants during a fire [67]. In this model, each room was a stock,
linear dynamics of the system’’ during a disruptive event [66]. with flows being the passages between rooms. SD methods are also
Economic impacts of disruptive events are estimated using the SD finding uses in life cycle assessments of buildings. These assess-
simulation model, which is then used to examine how secondary ments have been performed for embodied energy, CO2 emissions,
economic impacts are generated in interdependent infrastructures. and life cycle costs [68]. The opportunity to combine this decision-
The model is divided into various sectors, and, ‘‘Each sector models making tool with the expanding digital building information that is
the broad capabilities of the infrastructure and is not intended to be embodied in a building information model (BIM) presents a great
a stand-alone, detailed model. Instead its strengths lie in the opportunity for improving the building design process.
representation of first order interactions between sectors and the
ability to model/show how a simple disruption in one sector can 3.1. Use of system dynamics to model a building
propagate to others and disrupt the entire system of critical infra-
structures’’ [66]. In the research study presented in this paper, the SD method
was used to provide a means of simulating a building system and its
3. System dynamics applied to building design and interactions in a broad, ‘‘whole-building’’ fashion. While a large
simulation amount of data can be simulated or collected for the design of
a building, the ability to effectively integrate all these data into
Although SD has been used in a wide variety of disciplines, its a framework for making trade-offs and decisions in the design of
use in the design of buildings has been very limited. However, it has the building is, unfortunately, still lacking. The SD model enables
been applied to some problems related to buildings. Shen used an the utilization of all of the building design information in a holistic
SD program to simulate the evacuation of a building by its decision-making framework. The SD modeling method is especially

Infection

agent
breathing rate concentration

ingress
fraction
Dose

Ingress Rate

Dose Rate

Occupants

fatalities %

Egress Rate

egress
fraction
LD 50
total
fatalities

Fig. 6. SD building model infection sector.


B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015 1011

16000 the basic structure of a system, allowing for highly uncertain vari-
ables to be included. It is useful for identifying key leverage points
Agent Concentration (cfu/m 3)

14000
through use of sensitivity analyses [28]. It allows for simulation
12000 experiments to be conducted on hypothetical buildings or retrofits,
and is well-suited to policy analysis [47]. Another advantage of the
10000 SD method for simulating buildings is that it enables modeling of
8000 the interactions between building systems and building occupants,
and allows for incorporation of ‘soft’ factors that help to capture
6000 human behavior in the model [11]. Incorporation of these soft
factors is important in a building system model to address the
4000
occupants’ perceptions of and reactions to risks and events.
2000 A model constructed using the SD method is transparent to users
[8], and easily manipulated. Users do not need to be system
0
dynamics experts to use a model or to make changes to the
0 100 200 300 400 500 600
parameters in the model.
Time (min)
A SD model allows a building system model to be initially con-
structed with approximate values for parameters, as long as its
Baseline Case 80% Filter Efficiency Case structure is well-defined. This initial model can be used to deter-
mine the overall behavior of the system. Using sensitivity analysis it
Fig. 7. Filter efficiency comparison of two cases from the SD building model. is possible to determine parameters that have the greatest impact
on the model behavior [8]. Thus, even with incomplete data, it is
1.6E+07 possible to determine appropriate policies, or to rank alternatives.

1.4E+07
4. A building system dynamics model
Agent Removed (cfu)

1.2E+07

1.0E+07 In what follows, the development and use of a SD model to


determine the feasibility of using an SD model of a building to set
8.0E+06 priorities for upgrades to defend the building against a bioterrorist
6.0E+06 attack [69] are described. Bioterrorism presents building design
engineers with new challenges. Very little data or experience exists
4.0E+06 to guide building designers and decision-makers on how to miti-
gate this unpredictable hazard [70]. The output of this model is
2.0E+06
a prototype policy decision tool to help decision-makers with these
0.0E+00 challenges, along with a methodology for expanding and
0 100 200 300 400 500 600 enhancing the tool, and for applying it to additional areas within
Time (min) the building systems context.
Baseline Case 80% Filter Efficiency Case Fig. 3 shows a graphic of the methodology used to develop the
SD model of the building system. As indicated in Fig. 3, after con-
Fig. 8. Agent removed through recirculation comparison of two cases from the SD structing the SD model for the overall building system, a model
building model. ‘‘sector’’ is added to simulate a disruption to the system (in this
case, a bio-terrorist attack). The output of the simulation with the
applicable to building system simulation because it is ideal for added disruption provides a baseline building performance.
situations where the system to be modeled is extremely complex, Thereafter, building variables representing possible design
highly dynamic, or highly non-linear (containing large numbers of upgrades or building retrofits are added to the model. With these
feedbacks) [47]. The SD method is also well-suited for focusing on building variables included, additional simulations are carried out.

Table 2
Simple SD building model output data and comparisons with Table 1.

Time Agent in SA Agent in zone Agent concentration Agent in EA Difference from Kowalski
(min) (cfu) (building) (cfu) (cfu/m3) (cfu) (2003) data (%)

(3) vs (8) (4) vs (9) (5) vs (10)

(6) (7) (8) (9) (10) (11) (12) (13)


0 18,000,000 0 0 0 0.00 0.00 0.00
1 18,000,000 18,000,000 286.62 353,236 0.00 0.04 0.04
2 18,000,000 35,646,764 567.62 699,539 0.00 0.04 0.04
3 18,000,000 52,947,225 843.11 1,039,047 0.00 0.04 0.04
4 18,000,000 69,908,178 1,113.19 1,371,892 0.00 0.04 0.04
5 18,000,000 86,536,286 1,377.97 1,698,206 0.00 0.04 0.04
6 18,000,000 102,838,080 1,637.55 2,018,115 0.00 0.04 0.04
7 18,000,000 118,819,964 1,892.04 2,331,747 0.00 0.04 0.04
8 18,000,000 134,488,217 2,141.53 2,639,224 0.00 0.04 0.04
9 18,000,000 149,848,993 2,386.13 2,940,667 0.00 0.04 0.04
10 18,000,000 164,908,326 2,625.93 3,236,195 0.00 0.04 0.04
11 18,000,000 179,672,131 2,861.02 3,525,923 – – –
1012 B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015

airflow

Agent in Zone

Agent in supply Agent in Exhaust

UVGI removal
total rate
Agent Removed
airflow
Through Recirculation

agent filter removal


concentration rate
in minutes outside air %

Fig. 9. Variable changed in 80% filter efficiency case.

By examining the performance of the system across a spectrum of numerically by comparing Table 1 (data from Kowalski [72] for the
these building variables, the user can determine which building first 10 min of simulation) with Table 2 (data output from the SD
variables have the greatest impact on mitigating the effects of the building model). Columns (1)–(5) in Table 1 correspond to columns
disruption. By combining this information with cost information for (6)–(10) in Table 2. Columns (11)–(13) in Table 2 give the differences
various mitigation measures, it is possible to determine appropriate between the SD model output data and the Kowalski [72] data.
methods for allocating resources for cost-effective increases in Column (11) gives the comparison of columns (3) and (8), column
security, both in new buildings during the design process and in (12) gives the comparison of columns (4) and (9), and column (13)
existing buildings by retrofit and operational adjustments. gives a comparison of columns (5) and (10), respectively, as
a percentage of the Kowalski [72] data.
4.1. Model structure Once the SD building model was created, it was possible to
easily adjust building variables to determine their effectiveness at
The pilot SD building model studied is composed of two primary mitigating the effects of a bioterrorist attack. An example of this is
model sectors: (1) the air system sector; and (2) the infection sector. given in Figs. 7 and 8. In this scenario, two cases are examined: (1)
Fig. 4 shows the model of the building’s airflow system, modeled a baseline case; and (2) an 80% filter efficiency case. The baseline
using STELLAÒ software from isee Systems [71]. This model was case includes no air filtration, and the second case includes a filter
developed to simulate a single-zone building, with an airborne that is 80% effective at removing bioagent particles from the re-
release of a biological agent into its ventilation system. Each vari- circulated air. For this simulation, it was assumed that the bioagent
able in the model represents a linear differential equation. An was released downstream of the air handling unit’s filters; thus no
example of a flow variable from the model (agent removed through agent is removed by the filters until the air is re-circulated. These
recirculation) is shown in Fig. 5 with its associated differential scenarios were simulated for 10 h each. As expected, Fig. 7 shows
equation definition. that the agent concentration falls significantly in case (2), as
Fig. 6 shows the infection sector of the SD building model. This compared to the baseline case with no filter. Fig. 8, also as expected,
sector includes an elementary egress model, comprised of an inflow shows that the amount of agent removed increases in the case with
rate, an outflow rate, and a stock of occupants currently in the a filter as compared to the case with no filter. Fig. 9 highlights the
building. It includes a dose rate model to calculate the level of variable in the air system sector that was changed in the second
infection experienced amongst the building occupants. (80% filter efficiency) case.
The SD building model was run through a short, 12-min simu-
lation of a scenario where a biological agent is released down- 4.3. Building variables
stream of any passive defensive measures (filters, UVGI, etc.) over
time into the air handling unit of a single-zone building. This Building variables are chosen to represent the potential
scenario is laid out in Chapter 9 of Kowalski [72]. This scenario was upgrades to a building that would mitigate the effects of a bio-
chosen to yield a model which could be compared with published terrorist attack (e.g., filter removal rate, egress rate). Building
results of a sample building. variables are chosen to span various disciplines that have an impact
on building security. The leverage points for effective protection are
4.2. Preliminary model results/testing found from a subset of these building variables. These variables are
sufficiently diverse such that that there is no other single sub-
The model for the air system of the building, combined with system design model able to incorporate the entire set. This inter-
a sub-model to calculate infection amongst the building’s occu- disciplinarity of the SD model is its strength.
pants has produced results that are consistent with a published In this SD model, three major categories of building variables are
spreadsheet simulation [72]. When run with the published values studied: (1) performance of the air handling system; (2) physical
of building variables, the pilot SD building model provided essen- building security modifications; and (3) occupant behavior modi-
tially an exact replication of Kowalski’s output. This is shown fication. These three types of variables interact mainly with the
B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015 1013

Fig. 10. Linking BIM model with system dynamics building model.

dispersal of the bioagent, the vulnerability of the building to attack, to return data from the SD model back into the BIM model.
and the exposure of the occupants, respectively, although there are Examples of data that are linked into the SD building model from
interactions amongst the three types of variables, as well. The the BIM model include:
simulation of the interactions and effects of these three sets of
diverse variables provide a means to make logical decisions about  room dimensions;
the appropriate use of resources in the design of the building for  doorway maximum flows;
bioterrorism security. A number of these variables, particularly  room maximum occupancy limits; and
those related to the air handling system, are relatively easy to  initial number of occupants.
model because data exist to determine their relationships to the
system. However, other variables related to physical security and Fig. 10 illustrates this procedure graphically.
occupant behavior are more difficult to quantify. These variables
also have interactions with each other. Since many of these types of 4.5. Model potential
variables are difficult to quantify, they need to be varied over
a range of reasonable values to determine which variables have the As the SD building simulation method becomes widely applied
greatest effects on the outcome of the simulations. The outcomes and more data are produced, it may be possible to develop ‘classes’
are measured in terms of a calculated building security index, based of buildings, grouped by their most important characteristics as
on casualties (fatalities), costs (damages), and occupant perceptions determined by the identification of leverage points in the simula-
of security. tion model. Using a set of ‘generic’ building models, these building
classes could then be used as the basis for developing a perfor-
4.4. Integration of BIM model with system dynamics model mance-based design methodology for building security.
This SD methodology could also be applied to other aspects of
In order to populate the SD building model with the data building design, retrofit, operations and maintenance. For example,
necessary to represent the building, a link with a building infor- this model, integrated with the sustainable development ability
mation modeling (BIM) program has been explored. Specifically, model [50] could be modified to simulate a building in terms of
a building model created with Autodesk’s RevitÒ Building2 program environmental sustainability. This simulation would work in an
[73] was linked to the SD decision-making model. The link was analogous manner to the security model presented in this paper, in
accomplished by creating appropriate schedules within the RevitÒ that it would be possible to determine how to cost-effectively
Building model, modifying these schedules as needed to match the create a building that performs in an environmentally sustainable
input requirements of the SD model, and exporting the data from manner. This simulation could also be developed into a type of
these schedules to a spreadsheet. Once the data were in the performance-based analysis, which would allow various possible
spreadsheet, they were imported to the STELLAÒ SD building schemes to be compared and ranked in terms of their environ-
model. Currently the BIM-SD link is manual, but the possibility mental sustainability.
exists of creating an application programming interface (API) to
move data directly from a BIM software to a SD software as well as 5. Conclusion

This paper described the use of the system dynamics method for
2
The latest release of the program is called Revit Architecture. making decisions related to building design, operation and retrofit.
1014 B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015

Previous uses of the method have been reviewed, and the case laid [15] Kunszt G. Formal models in forecasting construction activities and architec-
tural trends on a global level. May, Stockholm: To Build and Take Care of What
out for why this method could be applicable to building design.
We Have Built With Limited Resources; 1983. p. 351-362.
Furthermore, a simple system dynamics building model has been [16] van Vuuren DP, Strengers BJ, De Vries HJM. Long-term perspectives on
developed and described in this paper. It has been tested for proof- world metal use – a system-dynamics model. Resources Policy
of-concept, and the testing has thus far given positive results. The 1999;25(4):239–55.
[17] Elshorbagy A, Jutla A, Barbour L, Kells J. System dynamics approach to assess
pilot SD building model, as described above, has been used to the sustainability of reclamation of disturbed watersheds. Canadian Journal of
simulate a building subjected to a bioterrorist attack [69]. The SD Civil Engineering 2005;32:144–58.
model has been used to simulate a small, relatively simple case [18] Nehdi M, Rehan R, Simonovic SP. System dynamics model for sustainable
cement and concrete: novel tool for policy analysis. ACI Materials
study building, adding building variables that will mitigate the 2004;101(3):216–25.
effects of a biological release in the building, and allowing different [19] Anand S, Vrat P, Dahiya RP. Application of a system dynamics approach for
buildings and different building configurations to be compared in assessment and mitigation of CO2 emissions from the cement industry. Journal
of Environmental Management 2006;79(4):383–98.
terms of their performance in the event of a bioterrorist attack. The [20] Kadoya T, Sasaki T, Ihara S, Larose E, Sanford M, Graham AK, et al. Utilizing
system dynamics building model will continue to be developed and system dynamics modeling to examine impact of deregulation on generation
improved, and will eventually be expanded to include various types capacity growth. Proceedings of the IEEE 2005;93(11):2060–9.
[21] Costanza R, Sklar FH, White ML. Modeling coastal landscape dynamics.
of building disruptions or terrorist attacks, and be useful for finding BioScience 1990;40(2):91–107.
the most effective leverage points across a multi-hazard spectrum [22] Quan C, Trani AA. A system dynamics model for the development of china’s
in the design of a building. The concept of directly linking a BIM aviation system. Airport Modeling and Simulation, Aug. 17–20, Arlington, VA;
1997. p. 226–232.
model with a system dynamics decision-making model has also
[23] Dyson B, Chang N-B. Forecasting municipal solid waste generation in a fast-
been introduced. growing urban region with system dynamics modeling. Waste Management
In addition, the authors anticipate that the system dynamics 2005;25(7):669–79.
method will find a wide range of other uses in the field of building [24] Deegan MA. Extreme event agenda setting and decision making. Proceedings of
the 21st International System Dynamics Society, July 20–24, New York; 2003.
design and operation. Foremost among these uses will be the [25] Vlachos D, Georgiadis P, Iakovou E. A system dynamics model for dynamic
concept of developing a system dynamics building model suitable capacity planning of remanufacturing in closed-loop supply chains.
for analysis of a building in terms of sustainable design. Computers & Operations Research 2007;34(2):367–94.
[26] Ford A. Modeling the environment: an introduction to system dynamics
models of environmental systems. Washington, DC: Island; 1999.
Acknowledgement [27] Meadows DH, Meadows DL, Randers J, Behrens III WW. The limits to growth.
New York: Signet; 1972.
[28] Randers J. From limits to growth to sustainable development or SD (sustain-
This research was supported by the United States Department of able development) in a SD (system dynamics) perspective. System Dynamics
Homeland Security through the National Center for Risk and Review 2000;16(3):213–24.
[29] Koelling P, Schwandt MJ. Health systems: a dynamic system – benefits from
Economic Analysis of Terrorism Events (CREATE) under grant system dynamics. 2005 Winter Simulation Conference, December 4–7,
number 2007-ST-061-000001. However, any opinions, findings, Orlando, FL; 2005. p. 1321–1327.
and conclusions or recommendations in this document are those of [30] Coyle JM, Exelby D, Holt J. System dynamics in defence analysis: some case
studies. Journal of the Operational Research Society 1999;50(4):372–82.
the authors and do not necessarily reflect views of the United States [31] Linard K, Sloper P, Paterson D. Defence preparedness and economic rationalists:
Department of Homeland Security. a system dynamics framework for resource allocation. The 16th International
Conference of the System Dynamics Society, July 20–23, Quebec City, Canada;
1998.
References [32] McLucas AC, Linard KT. System dynamics practice in a non-ideal world:
modelling defence preparedness. Sustainability in the Third Millennium: The
[1] Augenbroe G, Hensen J. Simulation for better building design. Building and 18th International Conference of The System Dynamics Society, Aug. 6–10,
Environment 2004;39(8):875–7. Bergen, Norway; 2000.
[2] Fujii H, Tanimoto J. Integration of building simulation and agent simulation for [33] Coyle RG, Gardiner PA. A system dynamics model of submarine operations and
exploration to environmentally symbiotic architecture. Building and Envi- maintenance schedules. The Journal of the Operational Research Society
ronment 2004;39(8):885–93. 1991;42(6):453–62.
[3] Helbing D, Farkas I, Vicsek T. Simulating dynamical features of escape panic. [34] Smith R. Counter terrorism simulation: a new breed of federation. Orlando, FL:
Nature 2000;407:487–90. Simulation Interoperability Workshop – Fall 2002, March 10–15; 2002.
[4] Brahme R, Mahdavi A, Lam KP, Gupta S. Complex building performance analysis [35] Bjornsson HC, de la Garza JM, Nasir MJ. A decision support system for road
in the early stages of design. Seventh International IBPSA Conference, Aug. 13– maintenance budget allocation. Computing in Civil and Building Engi-
15, Rio de Janeiro, Brazil; 2001. p. 661–668. neering; Proceedings of the Eighth International Conference, Palo Alto, CA;
[5] Heller M. Interdependencies in civil infrastructure systems. The Bridge 2000. p. 702–709.
2001;31(4):9–15. [36] Deegan MA. Extreme event policy design: a conceptual model to analyze policies
[6] Forrester JW. Industrial dynamics. New York: Wiley; 1961. and the policy process for natural hazards. Proceedings of the 23rd International
[7] Homer JB, Hirsch GB. System dynamics modeling for public health: back- System Dynamics Society, Jul. 17–21, Boston, MA; 2005.
ground and opportunities. American Journal of Public Health 2006;96(3): [37] Brown DE, Stile JR, White NM. Using simulation to produce a data fusion deci-
452–8. sion support tool for the assessment of manmade and natural disasters.
[8] Garcia JM. Theory and practical exercises of system dynamics. Barcelona, Proceedings of the 21st International System Dynamics Society, July 20–24,
Spain: Universitat Politecnica De Catalunya; 2006. New York; 2003.
[9] Rehan R, Nehdi M, Simonovic SP. Policy making for greening the concrete [38] Little RG, Weaver EA. Protection from extreme events: using a socio-techno-
industry in Canada: a systems thinking approach. Canadian Journal of Civil logical approach to evaluate policy options. International Journal of Emer-
Engineering 2005;32(1):99–113. gency Management 2005;2(4):263–74.
[10] von Bertalanffy L. General system theory: foundations, development, appli- [39] Sim MS, Lai KW, Seah SB, Khoo WG, Lee CY, Yee KC. Simulating impacts of
cations. New York: Braziller; 1968. disruption in a network of chemical manufacturing plants and supporting
[11] Caulfield CW, Maj SP. A case for systems thinking and system dynamics. infrastructures. The 24th International Conference of the System Dynamics
Proceedings of the IEEE International Conference on Systems. Tucson, AZ: Man Society, Jul. 23–27, Nijmegen, The Netherlands; 2006.
and Cybernetics; 2001. p. 2793-2798. [40] Ogunlana SO, Li H, Sukhera FA. System dynamics approach to exploring
[12] Ellis RE. The impact of instability in Latin and South America. IEEE Engineering performance enhancement in a construction organization. Journal of
in Medicine and Biology Magazine 2004;23(1):187–93. Construction Engineering and Management 2003;129(5):528–36.
[13] Choucri N, Madnick SE, Moulton A, Siegel MD, Zhu H. Linkage between pre- [41] Chapman RJ. The role of system dynamics in understanding the impact of
and post-conflict: exploiting information integration and system dynamics. changes to key project personnel on design production within construction
MIT Sloan Working Paper No. 4521-05. Cambridge, MA: MIT Sloan School of projects. International Journal of Project Management 1998;16(4):235–47.
Management; 2005. [42] Ibbs W, Liu M. System dynamic modeling of delay and disruption claims. Cost
[14] Dulac N, Leveson N, Zipkin D, Friedenthal S, Cutcher-Gershenfeld J, Carroll J, Engineering 2005;47(6):12–5.
et al. Using system dynamics for safety and risk management in complex engi- [43] Lee S, Peña-Mora F. System dynamics approach for error and change manage-
neering systems. Winter Simulation Conference, Dec. 4–7, Orlando, FL; 2005. p. ment in concurrent design and construction. Winter Simulation Conference, Dec
1311–1320. 4–7, 2005, Orlando, FL; 2005. p. 1508–1514.
B.P. Thompson, L.C. Bank / Building and Environment 45 (2010) 1006–1015 1015

[44] Lee S, Peña-Mora F, Park M. Quality and change management model for large [60] Ahmad S, Simonovic SP. Modeling human behavior for evacuation planning:
scale concurrent design and construction projects. Journal of Construction a system dynamics approach. Bridging the Gap: Meeting the World’s Water and
Engineering and Management 2005;131(8):890–902. Environmental Resources Challenges, May 20–24, Orlando, FL; 2001. p. 1–10.
[45] Love PED, Holt GD, Shen LY, Irani Z. Using system dynamics to better under- [61] Mehmood A, Saccomanno F, Hellinga B. Application of system dynamics in
stand change and rework in construction project management systems. car-following models. Journal of Transportation Engineering
International Journal of Project Management 2002;20(6):425–36. 2003;129(6):625–34.
[46] Peña-Mora F, Li M. Dynamic planning and control methodology for design/ [62] Chasey AD, de la Garza JM, Drew DR. Comprehensive level of service: needed
build fast-track construction projects. Journal of Construction Engineering and approach for civil infrastructure systems. Journal of Infrastructure Systems
Management 2001;127(1):1–17. 1997;3(4):143–53.
[47] Chritamara S, Ogunlana SO, Bach NL. System dynamics modeling of design and [63] de la Garza JM, Drew DR, Chasey AD. Simulating highway infrastructure
build construction projects. Construction Innovation 2002;2(4):269–95. management policies. Journal of Management in Engineering 1998;14(5):
[48] Ng HS, Peña-Mora F, Tamaki T. Dynamic conflict management in large-scale 64–72.
design and construction projects. Journal of Management in Engineering [64] Smith R. Complexities of simulating domestic infrastructure protection. Modeling
2007;23(2):52–66. Network Centric Warfare, Information Warfare, and Information Security:
[49] Menassa CC, Peña-Mora FM, Pearson N. Option pricing model to analyze cost- Simulation Interoperability Workshop – Fall 2002, Sep. 8–13, Orlando, FL;
benefit trade-offs of ADR investments in AEC projects. Journal of Construction 2002.
Engineering and Management 2009;135(3):156–68. [65] Bush BB, Dauelsberg LR, LeClaire RJ, Powell DR, Deland SM, Samsa ME. Critical
[50] Shen LY, Wu YZ, Chan EHW, Hao JL. Application of system dynamics for Infrastructure Protection Decision Support System (CIP/DSS) Project overview.
assessment of sustainable performance of construction projects. Journal of Proceedings of the 23rd International Conference of the System Dynamics
Zhejiang University Science 2005;6A(4):339–49. Society, July 17–21, Boston, MA; 2005.
[51] Forrester JW. Urban dynamics. Cambridge, MA: M.I.T. Press; 1969. [66] Dauelsberg LR, Outkin AV. Modeling economic impacts to critical infra-
[52] White KP, Dajani JS, Wright D. System dynamics approach to urban planning. structures in a system dynamics framework. Proceedings of the 23rd
Journal of the Urban Planning and Development Division 1974;100(1):43–56. International Conference of the System Dynamics Society, July 17–21,
[53] Drew DR. Scenariopolis-system dynamics land-use/transport system. Modeling Boston; 2005.
and Simulation, Apr. 19–20, Pittsburgh, PA; 1984. p. 1347–1352. [67] Shen T-S. ESM: a building evacuation simulation model. Building and Envi-
[54] Grigg NS, Bryson MC. Interactive simulation for water system dynamics. ronment 2005;40(5):671–80.
Journal of Urban Planning and Development Division 1975;101(1):77–92. [68] Matsumoto H. System dynamics model for life cycle assessment (LCA) of resi-
[55] Tangirala AK, Teegavarapu RS, Ormsbee L. Modeling adaptive water quality dential buildings. Proceedings of The Third International IBPSA Conference
management strategies using system dynamics simulation. Environmental (Building Simulation 1999), Sep. 13–15, Kyoto, Japan; 1999.
Informatics Archive 2003;1:245–53. [69] Thompson BP. Investigation of system dynamics applied to building simulation
[56] Simonovic S. World water dynamics: global modeling of water resources. for anti-terrorism resource allocation. Ph.D. Thesis. University of Wisconsin–
Journal of Environmental Management 2002;66(3):249–67. Madison, Madison, WI; 2009.
[57] Ahmad S, Simonovic SP. Spatial system dynamics: new approach for simula- [70] Thompson BP, Bank LC. Survey of bioterrorism risk for buildings. Journal of
tion of water resources systems. Journal of Computing in Civil Engineering Architectural Engineering 2008;14(1):7–17.
2004;18(4):331–40. [71] isee Systems. Stella: Systems thinking for education and research. < http://www.
[58] Winz I. Assessing sustainable urban development using system dynamics: the iseesystems.com/softwares/Education/StellaSoftware.aspx.> (Accessed Dec. 4,
case of New Zealand’s urbanwater systems. Systems Thinking and Complexity 2008).
Science: Insights for Action: 11th Annual ANZSYS Conference/Managing the [72] Kowalski WJ. Immune building systems technology. New York: McGraw-Hill;
Complex V, Dec. 5–7, Christchurch, New Zealand; 2005. 2003.
[59] Ahmad S, Simonovic SP. System dynamics modeling of reservoir operations for [73] Autodesk Building Solutions. Revit architecture. < http://usa.autodesk.com/adsk/
flood management. Journal of Computing in Civil Engineering servlet/index?id¼3781831&siteID¼123112;; 2008.> (Accessed Dec. 5, 2008).
2000;14(3):190–8. Autodesk.

You might also like