Simulation of Dynamic Energy Consumption in Modular

Construction Manufacturing Processes

Haiyan Xie1; Md Mukul Chowdhury2; Raja R.A. Issa, F.ASCE3; and Wei Shi4

Abstract: Although industrialized construction methods, such as modular and manufactured buildings, have evolved over decades, core tech-
niques used in prefabrication plants vary only slightly from those used in traditional site-built construction. With the focus of energy efficiency
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

in modular construction, this research presents the development of a simulation, measurement, and optimization system for determining
energy consumption (EC) in the manufacturing process of modular construction. The system uses the principles of a loosely coupled system
operation to identify the nonvalue-added tasks and possible causes of low energy efficiency. The system includes value stream mapping
(VSM), discrete optimization, and system dynamics (SD) for the demonstration and calculation of EC in modular construction. The benefits of
implementing this system include the quantity analysis of possible waste factors, reductions in the EC, decrease of energy cost, and increase of
profit. In addition, the research reveals the relationships between relocatable workers, energy efficiency, and operation flexibility. This system
should provide superior performance over existing operation evaluation systems, enabling decision-making based on reliable quantities and
delivery of true EC models to facility construction practitioners. DOI: 10.1061/(ASCE)AE.1943-5568.0000289. © 2017 American Society of
Civil Engineers.
Author keywords: System dynamics (SD); Modular construction; Energy consumption (EC); Value stream mapping (VSM); Discrete
optimization.

Introduction preassembly, modularization, and off-site fabrication techniques and

processes are all effective in reducing waste of materials and energy
An estimated loss of $1 million per project occurs in the construc- (McGraw Hill Construction 2010).
tion industry due to wasteful practices, such as over production, The demand for modular construction has been on the rise. For
poor handling, incorrect storage, incorrect ordering, design change, example, modular hotel construction in the United States increased
manufacturing defects, and rework (Dajadian and Koch 2014). by 25.7 and 31% in 2013 and 2014, respectively (Modular Building
The U.S. National Research Council (NRC) suggested five break- Institute 2015). Energy consumption (EC) became an essential part
throughs as potentially having the most significant impact on con- of the costs of production and operation of modular construction,
struction industry efficiency and productivity (Modular Building especially with automation in the manufacturing processes.
Institute 2010). These breakthroughs were (1) to deploy and imple- Therefore, the production growth of modular construction made the
ment interoperable technology applications, such as building in- need for efficient EC a priority. Although modular construction has
formation modeling (BIM); (2) to improve jobsite productivity many benefits, it also has some limitations in implementation,
through more effective integration of people, processes, materi- including (1) limited design options, e.g., limited amount of mate-
als, equipment, and information; (3) to increase the use of prefab- rial options and space layout possibilities; (2) reduced resale value,
rication, preassembly, modularization, and off-site fabrication e.g., there is a stigma that modular buildings are of lower quality;
techniques and processes; (4) to promote innovative use of dem- and (3) difficult financing, e.g., banks are usually unfamiliar with
onstration installations; and (5) to drive efficiency and support modular construction processes (Josephson 2016). In addition, there
innovation through effective performance measurement. Among is deficiency in research for the manufacturing process of modular
the potential breakthroughs, modular construction for prefabrication, construction (Gann 1996; Li et al. 2014). The production, preassem-
bling, and installation processes of modular construction are differ-
1
ent from conventional and in situ construction. Hence, the measure-
Associate Professor, Dept. of Technology, College of Applied ment and control of modular construction should be different from
Science and Engineering, Illinois State Univ., Normal, IL 61790-5100.
conventional construction.
(corresponding author). E-mail: hxie@[Link]
2
Graduate Student, Dept. of Technology, College of Applied Science This research focused on the manufacturing process of modular
and Engineering, Illinois State Univ., Normal, IL 61790-5100. E-mail: construction, particularly on reducing the cycle time (CT) of the
mmchowd@[Link] operation to make modular construction more energy efficient. The
3
Holland Professor, Rinker School of Construction Management, purpose of this study is to identify the nonvalue-added tasks based
Univ. of Florida, Gainesville, FL 32611-5703. E-mail: raymond-issa@ufl on CT and possible causes of wasted CT to make the EC process of
.edu modular construction more efficient. The objectives of this research
4
Executive Director, Capstone Global, Inc., Chicago, IL 60621. are (1) to identify the principal components of production effective-
E-mail: gatorwei@[Link]
ness in modular construction and (2) to explore whether the rela-
Note. This manuscript was submitted on March 8, 2017; approved on
August 22, 2017; published online on December 5, 2017. Discussion pe- tionships between these components and energy efficiency can be
riod open until May 5, 2018; separate discussions must be submitted for modified by the implementation of the value stream mapping
individual papers. This paper is part of the Journal of Architectural (VSM) technique. This study built a framework for analysis and
Engineering, © ASCE, ISSN 1076-0431. measurement of dynamic EC in modular construction. Although

© ASCE 04017034-1 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 theoretical evidence abounds to support the perceived benefits of Table 1. Drivers and Barriers of Modular Construction (Data from Pan
the Lean Six Sigma methods, very little empirical work has been et al. 2005 and Pan et al. 2012)
conducted to quantify its actual use and the associated costs and Driver Barrier
benefits in the modular construction industry (Pan et al. 2012;
Taghaddos et al. 2014). Very few studies and empirical work have Addressing the skill shortages Higher capital cost
been performed on the optimization of energy in modular construc- Ensuring time and cost certainty Difficult to achieve economies of scale
tion (Chen et al. 2010; Domínguez et al. 2012; Mao et al. 2013; Achieving high quality Complex interfacing between systems
Soršak et al. 2014). The research contributions of this study include Minimizing on-site duration Unable to freeze the design early on
the following items: (1) quantitative methods to identify the current
investigatory themes and (2) optimization of production configura-
CT ¼ Processing Time þ Inspection Time þ Wait Time
tion in modular construction.
In this research, published literature was reviewed, case studies þ Move Time (1)
were analyzed, modular construction companies were visited, and
industry representatives were interviewed to collect data and build
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

A building module is made of many components, and the num-

the measurement framework for energy efficiency in modular con- ber of processes involved is large. Fig. 1 shows a flow model of pro-
struction. National survey data (Robert Snow Means Company duction, in which moving, waiting, processing, and inspection are
2016) and variations in equipment units were also taken into con- all part of a value stream. The results of inspection include rework,
sideration. Realistic assumptions were made based on U.S. con- move on to the next process, or scrap.
struction industry cost data (Robert Snow Means Company The combined application of both Lean and Six Sigma principles
2016). Because there might be substantial differences in the con- could maximize the benefits of these technologies to the industry,
figurations of modular production, a prototype residential build- especially for modular construction (Koskela 1992). By working in
ing was designed and configured, which took into consideration unison, Lean and Six Sigma create a potent framework in eliminat-
many possible and practical work-break-down (WBD) structures ing process variation. Breyfogle et al. (2001) noted that after their
for that type of building. Later sections of this paper include the combination that Lean tools created a set of standards for problem
details of the scope of this type of building. The building will be a solving, whereas Six Sigma tools investigated and resolved any var-
single-story, multiunit, student dormitory. These parameters have iation from the set of standards. From the perspective of waste
an impact on production cycles and efficiency. reduction, Lean identifies waste and Six Sigma eliminates waste.
Moreover, according to Hammer and Goding (2001), Six Sigma
focused on the problems that were hard to find but easy to fix, which
Literature Review were better addressed using Lean production tools (Hammer and
Goding 2001). Lean construction alone does not clearly show the
Modular Construction underlying mechanism of how to measure the level of defects in the
current work processes (Howell and Ballard 1994; Howell 1999).
Modular construction is also referred to as off-site construction or Yu et al. (2013) developed and implemented a production system
production, manufactured construction, and industrialized construc- for effective application of Lean tools in building component pre-
tion. Pan et al. (2012) defined off-site production (OSP) as the man- fabrication. They chose a communication shelter production line as
ufacture and preassembly of building components, elements, or their case study and started with a very common Lean technique
modules before installation into their final locations. Recent called 5S (sort, straighten, shine, standardize, and sustain) to show
research on the benefits of modular construction found that, com- the immediate improvement to the workers and middle managers.
pared with traditional methods, the implementation of modular con- Then they implemented the VSM method to eliminate the root
struction yielded benefits in terms of reduction in time, defects, causes of wastes and increased the percentage completion rate of
risks, costs, and environmental impact. As a result, there were modules. They found that the low completion rate of modules was
increases in quality, consistency, predictability, productivity, per- due to variations of workloads caused by different module
formance, and profitability (Goodier and Gibb 2005; Blismas and configurations.
Wakefield 2009; Pan et al. 2012). Low adoption levels of modular The VSM was a tool demonstrating both value-added and
construction still existed despite the increased demand for imple- nonvalue-added activities in the processes required to bring a prod-
mentation of such technology by builders (Pan et al. 2005; 2012). uct from the raw material state into the hands of the customer, bring
The main driver for modular construction is the shortage of skilled a customer requirement from order to delivery, and bring a design
labor, whereas the main barrier is industry concerns about the high from concept to launch (Manrodt et al. 2008). As shown in Fig. 2,
capital costs of implementing such technology (Table 1). Nadim Yu et al. (2013) suggested in their case study three measures,
and Goulding (2011) attributed the reasons for the low adoption of including workload leveling, restructuring work and in-station qual-
modular construction in the industry to (1) the difficulty to ascertain ity, and increasing process reliability and achieving a 6-h takt time
the benefits, (2) the negative image regarding prefabricated/modu- (Tk). The basic idea of workload leveling was to meet varying cus-
lar buildings, (3) poor quality, (4) poor aesthetics, (5) limited tomer demand (a mix of modules with variations) without workload
choices, and (6) previous failures. Quantitative data to show the dif- fluctuation in the manufacturing process. Workers at each station
ferences of cost or the energy savings between modular and conven- were assigned based on the configuration of the module at that sta-
tional construction would help encourage more implementations. tion. The responsibility of deciding the production sequence was
moved from sales to production line managers. The purpose of
restructuring work was to balance the production line so that the
Lean Six Sigma Principles and Tools to Optimize
overall CT of each station could be as close to the Tk as possible. In
Production
4 months after the implementation of 5S, the company improved
The CT is the total time required for a piece of material to traverse labor efficiency by 10%, the labor cost was reduced by 18%, and
the production flow and is calculated using the overtime was reduced from 20 to 5%.

© ASCE 04017034-2 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Flow model of production (Adapted from Koskela 1992)

Anderson and Kovach (2014) described how a construction optimum insulation level, which was greater than those values rec-
company applied Lean Six Sigma methodology to reduce welding ommended by the energy-efficient building regulation in Quebec.
defects in turnaround projects. The company implemented the Discrete optimization is used in this research to help improve the
action research method of inquiry involving employees and performance of modular construction.
researchers and the define-measure-analysis-improvement-control
(DMAIC) method to solve the welding defects problem. By imple-
menting windshield standards, training welders, and instituting eye- Research Design
sight tests for welders, the repair rate decreased by more than 25%,
which translated into a savings of $90,000 for this company. The Overview of Methodology
project team identified, through multiple brainstorming sessions
using a five-why analysis, the potential root causes of high butt This study used a quantitative research methodology, as shown in
weld repair rates. Fig. 3. It implemented Six Sigma tools to identify the root causes of the
problem of energy efficiency in modular construction and used VSM
as a Lean technique to quantify the root causes. In addition, the identi-
Discrete Optimization in Modular Construction
fied root causes were verified and confirmed through the implementa-
Production
tion of loosely coupled multiunit (LCMU) systems. After the confirma-
Table 2 compares the reasons for choosing the leveraged techniques tion of the root causes, the next step was to focus on possible areas of
among other available alternatives. In this research, the system sim- improvement. Those areas were quantified and verified through the
ulates the allocation of workers in the manufacturing process of application of a system dynamics (SD) model. The discrete optimiza-
modular construction. Both workers and modules need to be finite. tion process was successfully applied to optimize the system.
In addition, the eigenvalue matrix of the system is symmetric. The This paper provides original reviews of past practice in the
system design uses vector iteration or the power method. improvement of production efficiency, presents information of cur-
Discrete optimization has been used in many construction- rent interest in modular construction and energy simulation, and
related calculations and engineering improvements. Zawidzki and probes a new field by developing a method for simulating the EC in
Nishikawa (2010) used discrete optimization in their Truss Z sys- modular manufacturing. The following steps explain the methodol-
tem to minimize the total number of modules used. They also ogy used in the simulation and optimization model. The step num-
achieved the best alignment to the given paths’ reduced network bers are consistent with Fig. 3.
distance, and maximized the network flow. A discrete optimization
process using the firefly algorithm (FA) was applied in scaffolding
Step 1: Identify Problem in Modular Construction
modular construction to optimize the scaffolding schedule. The
implementation generated improved scaffolding designs, enhanced To report research results, the authors designed a typical construc-
erection and dismantling schemes, shortened construction sched- tion project suitable for the implementation of modular construction
ules, and reduced resource requirements and estimated cost (Liu considering planning, analysis, design, construction, and manage-
et al. 2014). Hamelin and Zmeureanu (2012) presented the optimi- ment of the building. The experiment data are based on literature
zation of a single-family house using two objective functions, i.e., review, practical applications [such as RSMeans: Building con-
lifecycle primary energy use (LCE) and lifecycle cost (LLC). They struction cost data 2016; Robert Snow Means Company (2016)],
created a dynamic simulation model using discrete optimization to and field observations of three building-manufacturing plants. The
simulate the annual heating and cooling EC for the climate of purpose of using the experiment data is to verify the design of the
Sainte-Agathe-des-Monts, Quebec. As a result, they obtained an simulation method.

© ASCE 04017034-3 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Future VSM (Adapted from Yu et al. 2013)

Table 2. Comparison of Available Alternatives

Item Applicable areas Algorithm examples Advantages Disadvantages

Discrete optimization Searches for an optimal Vector iteration (power Easy to implement Might be slow
(Arbenz and Kressner solution in a finite or method) (Dashti et al. 2010; Adaptable to extend many algo- Often computes the largest
2012) countable and infinite set Lau et al. 2011) rithmic variants of solutions of eigenvalue of a matrix
of potential solutions generalized eigenvalue problems
to improve the accuracy
Can be used to incorporate addi-
tional constraints
Simultaneous vector or Low storage requirement Nearly always too slow; for
subspace iterations (Kelley most problems they do not
1999; Saad and van der Vorst converge at all; do not take
2000) advantage of the situation
when the equations are
symmetric
Continuous optimization Variables used in the Genetic algorithm Multiple local optima; the objec- No guarantee of finding global
(Lee and Geem 2005; objective function are tive function is not smooth; can maxima; long time taken for
Storn and Price 1997) continuous variables handle multiple parameters; the convergence
objective function is noisy or
stochastic

Data Resource 1. Selection of task: Select tasks for analysis purposes.

In this research, initial data were collected from a building costs 2. Crew information: Gather information about the number of
data book (Robert Snow Means Company 2016). Examples of the workers that are needed to complete a unit for every task.
tasks, crew information, and labor hours are shown in Table 3. 3. Labor hours: Determine the amount of labor required to per-
Generally, the following procedure was followed in conducting form one unit of work. This step records the labor hours for
data collection: selected tasks.

© ASCE 04017034-4 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Research methodology; note that the step numbers are used in the paper to explain the steps in the simulation and optimization model

Table 3. Tasks, Crew, and Labor Information of One Sample

Task ID
Parameter 1 2 3 4
Description Frame wall Install doors and window Complete interior wall finish Complete ceiling finishes
Number of workers 2 2 2 2
CT (min) 290 215 593 394
EC (kW) 1,015 752.50 2,075.50 1,379
Completion (%) 96 80 71 93
Processing rate (%/min) 0.33 0.37 0.12 0.24
Note: The EC (kW) values are calculated and explained in the subsection Findings and Results of the Data Analysis section.

Term Definitions was the amount of time one station has to wait on its previous
Based on the cost information collected and the design of the room station’s work to be completed.
module, the following items were used in the measurement of EC: 7. Actual working time for station (i) without waiting on the
1. Number of workers = Worker per unit  Number of units completion of station (i − 1) = CTi − 100 − % of completioni − 1/
needed Processing ratei − 1
2. CT ¼ Labor hours per unit  Number of units needed 8. Actual processing rate of station (i) = % of Completioni/
3. Completion Rate ð%Þ  70%. An assumption was made that Actual working timei
no station completes less than 70% of its total tasks within the 9. Actual time needed for station (i) to complete its remaining
scheduled CT. Random numbers for the completion rates unfinished work = 100 − % of completioni/Actual processing
within the assumed data range are generated. ratei
4. Processing rate ð%=minÞ ¼ Completion rate=CT:
10. m = Actual time needed for station (i) to complete its remain-
Measurement of EC: equipment energy consumption (kW · h)
ing unfinished work/Number of workers assigned to station (i)
(OMNITECH 2017).
11. Reduced CT after adding workers = Original CT − ( m 
5. EC ¼ Equipment EC  CT
Number of workers added)
6. The amount of time that station (i) needs to have for incom- 12. Reduced EC after adding workers = Reduced CT 
pleted work = 100 − % of completion/Processing rate, which Equipment EC

© ASCE 04017034-5 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 the other hand, it is assumed that the factor People is not directly
related to the factor Methods, but it is directly related to the factors
of Measurement and Machines. Hence, the factor People has an
indirect effect on the factor Methods. For example, if the employees
of a company are not well trained, they will not be able to use
machines properly, which will eventually affect the production pro-
cess of the company. Therefore, their effects are important to calcu-
late the eigenvalue associated with the root cause.
The matrix [K] power method (Wilkinson 1965) is used to calcu-
late the dominant eigenvector for the relationships shown in Fig. 4.
An initial approximation (X0) of one of the dominant eigenvectors
of matrix [A] ([K  K] matrix) was chosen, where vector [X0] [
Rn = 0. A X0 = l X0, where l is the eigenvalue of [A] with corre-
Fig. 4. Modular construction as a loosely coupled system
sponding eigenvector [X0].
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

For example, to demonstrate the eigenvalue model, a dormitory

Step 2: Lean Six Sigma for a WBD Structure module room 3.048 m (10 ft) long, 2.4384 m (8 ft) wide, and
2.4384 m (8 ft) tall in size was designed (Fig. 5). Each module
Use Lean Six Sigma to identify factors causing problems in opera- includes a door and one to two windows. It uses wood for the floor
tions. Particularly, the purpose of this step is to build a WBD struc- and wall frames.
ture of the production process of modular construction.
Step 5: Use VSM to Quantify the Time Value of Factors
Step 3: Build a LCMU System
This step is to generate a list of the major tasks in module manufac-
Build the operation process into a LCMU system operation to iden- turing processes based on literature review and site observations.
tify the nonvalue-added tasks and possible causes of low energy The operation of the tasks is under the conditions shown in Table 3.
efficiency.
Step 6: Use Discrete Optimization to Calculate EC in
Step 4: Calculate Eigenvalues for Possible Areas of Production
Improvement
Little’s Law (Little 1961) in queuing theory is able to calculate the
Based on the result of the LCMU system, use matrices to calculate average number of customers in a queuing system. The next step in
eigenvalues and find possible areas (or factors) of improvement and this research is the discrete optimization, which applies Little’s Law
quantify the factors. The following example shows how energy in a VSM system. The VSM system consists of several stations
might be wasted in modular construction. Suppose a company has used to produce modules for prefabricated buildings. Eq. (2) shows
four stations running concurrently to produce a module, which must the calculation of the work in process (WIP)
complete four tasks in four stations in a sequential order. The equip-
ment ECs per hour at Stations 1–4 are 100, 150, 250, and 200 k · W/h, WIP ¼ TH  CT (2)
respectively. The ECs of the stations depend on the amount of time
spent in completing the tasks. If Station 2 has a product defect where WIP = parts of the modules not yet completed in the factory;
found, for example, workers need to spend extra time to fix it, TH = average quantity of modules (nondefective) produced per unit
which will exceed the planned production time. It will, conse- time; CT = elapsed time from the module’s creation start to comple-
quently, increase the EC of Station 2. Moreover, the tasks of tion, including processing time, transportation time, queue time,
Station 1 will have to wait until the problem at Station 2 is fixed hold time, and so forth.
because the resources of labor and equipment are still occupied at Eq. (3) calculates CT using Eq. (2)
Station 2. This leads to additional energy usage because the equip-
ment in Station 1 keeps running. To solve this problem and to CT ¼ WIP=TH (3)
reduce the EC, workers from other stations (i.e., who are standing
idle or doing nothing) are moved to Station #2 in which Task 2 and Therefore, CT can be reduced (or improved) by reducing WIP
the unfinished tasks of Station 1 are being completed. This will and/or increasing TH.
increase the processing rate. By increasing the processing rate, The third step is to reduce CT and increase TH. The definition of
more work will be done in less time, which will, consequently, critical WIP (W0) indicates the WIP level in which a production line
reduce EC. without congestion would achieve maximum TH (i.e., bottleneck
In modular construction, the work at one station affects the work rate) with minimum CT (i.e., T0). The following equation calculates
at subsequent stations. Thus, an activity at a station is loosely critical WIP:
related (or tightly interdependent) with other activities. If two sta-
tions or tasks have a number of variables in common, compared W0 ¼ Rb  T0 (4)
with other variables influencing the activity, they are defined as
loosely coupled sets. A loosely coupled system depends on its envi- where W0 = critical WIP; Rb (bottleneck rate) = rate of the process
ronment and complexity. In a loosely coupled system, any modifi- center, which has least the long-term capacity with highest long-
cations or changes in an activity cannot affect the entire system. term utilization; T0 = raw processing time, minimum CT, and sum
Fig. 4 shows how modular construction is loosely coupled. Among of long-term average processing time at each workstation in a pro-
the six causes shown in Fig. 5, assume that the Machines, Material, duction line.
and Environment are all involved with the process Methods and The calculations of CT and TH for different W (WIP) values
that they have strong direct effects on the production process. On are shown next. The following conditions show that increasing

© ASCE 04017034-6 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. (a) Floor plan view of a dormitory project and module room and (b) elevation view

bottleneck rate (Rb) can improve both TH and CT when WIP distributed. The quality check (QC) task was placed at every station
levels are greater than W0 before bottleneck operation to reduce the number of defects. To
reduce the change over time (COT) in case of equipment break-
CT ¼ T0 ; if W ( W0 down, replacement parts and repair personnel (RP) must be ready
CT ¼ W=Rb; if W > W0 and as close to the St4 as possible.

Hence, Step 7: SD for Verification of Optimization

TH ¼ W=T0 ; if W ( W0 This step is to use SD software to calculate and verify the EC in
production. The system dynamics (SD) is a set of perspective and
In this case, increase bottleneck rate (Rb), if W > W0, to improve conceptual tools that enable the understanding of the structure
TH. In this research, it is assumed that the processing rates of differ- and dynamics of complex systems (Sterman 2002). The SD deals
ent stations are X1, X2, X3, and X4 (Fig. 6), where X1 < X2, X2 = X3, with the interaction of the various elements of a system in time
X3 > X4. and captures their dynamic aspects by incorporating concepts,
Hence, the bottleneck is Station 4 because it has the lowest proc- such as stocks, flows, feedback, and delays, providing insights
essing rate (X4) among all other stations. The entire system of mod- into the dynamic behaviors of systems over time (Tang and Vijay
ule production is shared among four stations, which are operated 2001). Implementation examples of SD models include a mea-
based on the production schedule. surement model of urban-EC in Beijing (Feng et al. 2013) and an
Fig. 6 shows that the remote management interface (RMI) is estimating model of what the EC would be on a monthly or yearly
releasing the material into the process at the rate of the first opera- basis for different types of housing systems (Quinn 2008).
tion. Station 1 (St1) processes X1 parts per minute and places them Generally, in a SD model, stocks represent materials, knowledge,
in WIP1, in which they wait for St2. Meanwhile, St2 processes X2 people, money, and so forth. Flows in an SD model represent
parts (which is greater than X1) per minute. Because St2 processes change rates, which show how stock values change and define the
more parts than St1 on average, no parts accumulate in WIP1. St2 system’s dynamics. For this study, based on literature review and
feeds parts to St3 at the rate of X2 per minute, and St3 processes site observations, the Machines, Materials, Measurements, Methods,
them at the same rate (because X2 = X3). Therefore, no work accu- People, and Environment factors are considered as stocks in the SD
mulates in WIP2 as well, given that there is no variability in CT. model shown in Fig. 7. The stocks represent the locations in which
The output of St3 waits in WIP3 to be processed by St4. Because operations take place and make values change. The flows include
St4 can process less parts than St3, X3 – X4 parts accumulate in COT and CT, which affect the values of the variables of defects, CT,
WIP3, which leads to increase in CT. completion rate, and Tk, which in turn impact EC.
To increase the Rb and to make sure that no parts would wait at Fig. 8 shows the SD palette with two stocks added, Material
WIP3 to be processed by the next station, more people (L) were Equipment (ME) and EC. Then the flow of CT from the stock of
assigned to St4, from another station during unused time. In reality, ME to the stock of EC is added. In the next step, the parameters and
if the modules have different dimensions and configurations, the dependencies as discussed previously are defined. After that, two
production team will carefully analyze the demand and prepare the parameters, called initial EC [denoted as InitialEngCon (IEC)] and
production schedule, and then the schedule of each station will be duration of CT [denoted as CTduration (CTd)], are added to the

© ASCE 04017034-7 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 Fig. 6. Discrete optimization based on Little’s Law
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Potential causes of longer CT (using Six Sigma methodology)

system and are set as dependencies of stocks and flows. In the fourth the principal components of production effectiveness in modular
step, the formula for the flow of CT was defined. In the fifth step, construction. The statistical analysis in the next section shows the
the model was executed for approximately 15 s to show the change details of the calculation.
of EC due to CT for every station. The time used in the fifth step is
only for demonstration of the SD system. The EC value is the indi-
cator for energy efficiency. When EC is greater than a certain
threshold value, waste factors should be analyzed and improve- Data Analysis
ments would be proposed. The calculation of the threshold value
will be discussed in the following section.
Analysis for Eigenvalue (Step 4 of Research Design)
Following the DMAIC approach, after identifying the waste fac-
tors in the manufacturing process of building modules, the next step Table 4 shows a list of the major tasks necessary when manu-
would be to identify root cause(s) of the factors responsible for facturing room modules. The data were collected from an indus-
waste in EC. This analysis focuses on changes implemented to try standard cost data book (Robert Snow Means Company 2016).
improve the process. As shown in Fig. 7, the potential root causes of Under the first task of Framing Walls at Station 1, there were four
waste in each category include machines (i.e., equipment), methods different activities. The total hours for this task were broken down
(i.e., how work is done), people (i.e., human elements), environ- into labor hours needed for each individual activity.
ment (i.e., buildings, logistics, space), materials (i.e., raw materi- Table 4 shows the eigenvalue of the activity Frame Wall. The
als), and measurement (i.e., calibration, standard). following calculation defines Frame Wall activity as matrix [B]
!
Step 8. Use Statistical T-Test to Check 20 20
½B ¼
Optimization Results 220 30
The statistical tests help to answer whether there is a significant dif-
ference in terms of energy savings after applying the Lean Six According to the characteristic equation of a 2  2 matrix, the
Sigma technique to the production system. The results can identify authors found the following equation:

© ASCE 04017034-8 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. The SD model of EC

Table 4. Room Module Schedule

Quantity in Quantity in Labor hours Total labor

Description of task SI unit Imperial unit per unit hours (min) Eigenvalue
Station 1: Frame walls 10.97 m 36 linear feet (L.F.) 0.133 290 91.50
Material staging — — — 20 —
Layout wall plates — — — 20 —
Assemble wall frames on working surface — — — 220 —
Assemble wall frame sections into a room module — — — 30 —
Station 2: Install doors and windows — — — 215 123.25
Apply water-resistant barrier 7.43 m2 80 sq ft. 0.002 10 —
Install window and door flashing 12.19 m 40 L.F. 0.008 20 —
Install doors 1 each 1 each 1.23 75 —
Install windows 2 each 2 each 0.9 110 —
Station 3: Interior walls and finish — — — 593 271.25
Install wiring for stitches and outlets 4 each 4 each 0.468 112 —
Install insulation 26.76 m2 288 sq ft 0.006 100 —
Install and finish 15.875-mm fire-rated drywall 26.76 m2 288 sq ft 0.017 295 —
Prime and paint drywall 26.76 m2 288 sq ft 0.005 86 —
Station 4: Ceiling finishes — — — 394 199.27
Install wiring for ceiling lights 3 each 3 each 0.468 84 —
Install insulation 7.43 m2 80 sq ft 0.017 81 —
Install and finish 15.875-mm fire-rated drywall ceiling 7.43 m2 80 sq ft 0.021 100 —
Install spray applied acoustical ceiling treatment 7.43 m2 80 sq ft 0.027 129 —

λ2  λ ð20 þ 30Þ þ ð20  30  220  20Þ ¼ 0 The data collected from all tasks in Table 4 show a significant time
¼> λ2  50λ  3; 800 ¼ 0 variance between the stations, indicating the importance of balanc-
¼> λ1 ¼ 91:50; λ2 ¼ 41:5; λ1 ¼ 91:50 ing the total time in each station. Consequently, significant varia-
tions exist in terms of energy consumed by each station in Fig. 9,
is the dominant eigenvalue. which shows that the EC is higher at Station 3, in which CT (593
In Table 4, the eigenvalue defines that the Interior Walls and min.) is the highest among all the stations. Therefore, the optimiza-
Finish activity have the highest value. This is most likely responsi- tion process should focus on reducing the CT of Station 3 by man-
ble for the root cause of a longer CT, which needs to be improved. aging the related labor units and moving them from different sta-
tions to Station 3 when they are needed.
The following calculation explains the bottleneck rate of Station
Analysis Using Discrete Optimization on EC (Step 6 of
3 when the processing rate of Station 3 (X3) is less than the process-
Research Design)
ing rate of Station 2 (X2). The data in Table 3 show that the process-
The machine used is a wood PVC door panel–making machine with ing rate of Station 3 (0.12% per minute) was less than any other sta-
an EC (per hour) of 210 kW (=3.5 kW/min) (OMNITECH 2017). tion; therefore, it was considered as a bottleneck. Under the

© ASCE 04017034-9 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 Current scenario of CT and EC in different stations
2500.00
2075.50
2000.00

1500.00 1379.00

EC
1015.00
1000.00 752.50
500.00

0.00
St 1 St 2 St 3 St 4
CT

Fig. 9. Current situation of CT and EC

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

previously mentioned situation, Station 3 was the bottleneck Verification Using SD (Step 7 of Research Design)
because it had the lowest processing rate and also the highest EC
Fig. 11 shows a screen shot of the simulation using SD to determine
(2075.50 kW). From the data in Table 3, the completion rate of tasks
the change in EC of Stations 1–4, respectively, after running the sim-
at Station 2 was 80.113% within the scheduled CT. Therefore, the
ulation for CT of 290, 215, 593, and 394 min, respectively. They
amount of time that workers at Station 2 spend on waiting for
show the change of EC from the initial phase (MaterialEquipment)
Station 3 to be available for accepting WIP from Station 2 is calcu-
to the later phase of EC due to a change of CT (ECchange). The val-
lated as follows: ues from all the stations demonstrate the effect of CT on EC and
100  80:113 prove the functionality of the model. The flows that will cause more
¼ 53:749 min EC are the main factors causing the inefficiency problem in EC. The
0:37
minus value in the model indicates that energy, material, or effi-
where 0.37 is the processing rate of Station 2. ciency of equipment is being consumed when the value of CT keeps
If Station 3 can synchronize with Station 2, then the time left for changing. The assumption is that the initial EC (InitialEngCon) is
Station 3 = 593 – 53.75 = 539.25 min. Thus, the actual processing zero.
rate of Station 3 (Pr 3) is calculated as follows:
Statistical Analyses (Step 8 of Research Design)
71%
Pr 3 ¼ ¼ 0:1317%=min Two sample t-tests based on 55 observations were used to compare
539:25 EC before and after the application of Lean Six Sigma in the same
system. The statistical t-tests were used to validate the simulation
From Table 3, the remaining work of Station 3 is (100 – 71) =
results and optimization results. Hence, the same modules are simu-
29%. Therefore, the amount of time that Station 3 had to spend to lated multiple times to fulfill the optimization process and to arrive
complete its remaining work is calculated as follows: at the expected confidence level of the t-test. The following are the
29% two hypotheses for the t-tests:
¼ 220:26 min 1. Null hypothesis (H0): There is no difference in terms of energy
0:1317%=min
savings after applying the Lean Six Sigma technique to the pro-
duction system.
Because a total of two workers are involved in Station 3, the CT 2. Alternate hypothesis (Ha): There is a significant difference in
per worker is calculated as follows: terms of energy savings after applying the Lean Six Sigma
technique to the production system.
220:26 min
¼ 110:13 min=worker Table 5 shows the results for the t-tests. The sample group had
2 workers 54 degrees of freedom because it contained 55 observations. As
shown in Table 5, the Pearson correlation was 0.724, which indi-
From the previous calculation, for every additional worker in
cates that the two variables were highly correlated. The value of the
Station 3, CT could be reduced by 110.13 min. Fig. 10 shows how
calculated t statistic is higher than the value of t critical for one-tail
EC at Station 3 is reduced when CT is reduced after adding addi-
(2.62 > 1.67). Thus, there is a significant difference between the
tional units of labor from other stations.
two samples. Because the calculated p value of the one-tail test is
Fig. 10(a) shows that at Station 3, CT could be reduced from 593
less than the alpha level (0.006 < 0.05), Ha is accepted with 95%
to 482.87 min after adding one unit of labor. Consequently, EC is
confidence.
reduced from 2075.50 to 1690.05 kW. Fig. 10(b) shows that at
Station 3, CT could be reduced from 482.87 to 372.74 min after
adding two units of labor. Consequently, EC is reduced from Findings and Results
1690.05 to 1304.60 kW. Fig. 10(c) shows that at Station 3, CT could
be reduced from 372.74 to 262.61 min after adding three units of Based on the preceding statistical analysis, there is a significant dif-
labor. Consequently, EC was reduced from 1304.60 to 919.15 kW. ference in terms of energy savings after applying the Lean Six
Because the units of labor were added from other stations, the total Sigma technique to the production system. This research revealed
possible number of units of labor that could be added to Station 3 the following findings. (1) Lean Six Sigma is able to identify the
was limited. Per the calculations, this is the optimized situation. causes of EC inefficiency and select the critical one to optimize. (2)

© ASCE 04017034-10 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. (a) Future scenarios of EC after adding one worker at Station 3 (St 3), (b) future scenarios of EC after adding two workers at Station 3, and
(c) future scenarios of EC after adding three workers at Station 3

The EC is an integral indication of machine hours, tool usage, its immediate predecessor station (Station 2). The completion rate
indoor HVAC and lighting needs, and any equipment that requires at Station 3 was slower than that of Station 2. Therefore, it could
energy. The identified causes are valuable to managers and engi- not completely synchronize with Station 2, which was processing
neers of modular construction to consider when designing the pro- and feeding the tasks at a faster rate. As a result, some tasks at
duction layout. (3) The calculated optimization suggestion is able to Station 2 had to wait for a time difference between Stations 2 and
improve both EC and CT. The enhancement of CT will in turn 3, until the crews at Station 3 were ready to accept the tasks.
increase productivity of modular construction. (4) For the manufac- Station 3 had a longer CT and higher EC, with 593 min and
turers of modular construction, which have a stable workforce, ma- 2075.50 kW, respectively. The CT and EC at Station 3 were
terial supply, and manufacturing locations, their workforce needs higher than any other stations in the system. Therefore, a new
are different from those that are temporary and project based. (5) method was implemented to optimize the existing process to
Financial performance should not be the only criterion for the eval- reduce its CT and, consequently, EC.
uation of the production of modular construction. The findings and This research uses EC as a comprehensive indicator of any
calculation results will help to decide on the support for the opera- equipment that requires energy. Managers and engineers can use
tions of modular construction and the improvement solutions on the the indicator to seek an optimum solution when designing the pro-
efficiency of production cycles. duction layout. For the verification of the SD model, the system ran
In this research, Lean Six Sigma is able to identify the causes of for the duration of the CT. The EC at Station 1 was –1,015 kW in
EC inefficiency and select the critical one to optimize. The applica- Fig. 11(a). The ECs was –752.50, –2075.50, and –1,379 kW for
tion of Lean Six Sigma, however, does not reduce the EC of the Stations 2–4, as shown in Figs. 11(b–d), respectively. In the discrete
equipment in the dynamic system. Instead, the implementation optimization process, the first focus was on Station 3 because its
determines which factor is the main cause of the inefficiency with processing rate was the lowest, which was 0.12%/min. Its CT and
cross-functional work-process diagnostics and effective strategies. EC were also the highest among other stations. The results showed
With the verification of a LCMU system, the largest eigenvalue of that for every additional unit of labor added at Station 3, CT can be
Interior Walls and Finish activity was 271.25. It was selected to be reduced by 110.13 min, leading to a savings of 385.46 kW (110.13 
improved first. In this research, the tasks at all the stations were 3.5) per unit of labor. Figs. 9 and 10 show that after adding up to 3
modeled sequentially. As indicated in Table 3, the processing rate at units of labor one at a time, CT was reduced from 593 to 482.87 mins,
Station 3 was 0.12%/min, which was lower than any other station in from 482.87 to 372.74 min, and from 372.74 to 262. 61 min, respec-
the process. It led to longer CT due to an increase in waiting time tively. Thus, the EC was reduced from 2075.50 to 1690.05 kW, then
and amount of WIP between Stations 2 and 3. The reason for the to 1304.60 kW, and finally to 919.15 kW, respectively, at Station 3.
increase of WIP was that Station 3 had a lower processing rate than The data analysis results indicated that the data samples were

© ASCE 04017034-11 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. (a) The ECs of Station 1, (b) ECs of Station 2, (c) ECs of Station 3, and (d) ECs of Station 4

statistically significant and the application of Lean Six Sigma tech- Table 5. T-Test Results
nique could greatly increase the processing rate at Station 3, which,
consequently, reduced the process CT and EC. In summary, the calcu- Statistical result Before application After application
lated optimization suggestion is able to improve both EC and CT. The Mean 1,310.781818 1,177.810545
enhancement of CT will in turn increase productivity of the modular Variance 254,686.97 259,525.2333
construction. Observations 55 55
For the manufacturers of modular construction, which have a Pearson correlation 0.724722238 —
stable workforce, material supply, and manufacturing locations, Hypothesized mean difference 0 —
their workforce needs are different from those that are temporary Degree of freedom 54 —
and project based. The implementation results of the Lean Six t Stat 2.620940195 —
Sigma technique showed that moving idle workers from different P(T <= t) one-tail 0.005682756 —
stations to Station 3 helped to balance the production line and t Critical one-tail 1.673564906 —
increased the processing rate of that station identified as bottleneck. P(T <= t) two-tail 0.011365511 —
The increase was from 0.12 to 0.13, which, consequently, reduced t Critical two-tail 2.004879288 —
the wait time, reducing the CT time at Station 3. As EC increases
with the increase in the time spent in the production process, reduc-
ing the CT of Station 3 helped to reduce the overall EC of the pro- trades. For example, carpenters can install floors, walls, ceilings,
cess. The results showed that after applying VSM as the Lean tool doors, and windows of a building module. In addition, worker train-
and DMAIC as the Six Sigma technique, and adding up to 3 units of ing programs provided by unions offer classes covering the skills of
labor gradually at Station 3, CT measurements were reduced by a wide spectrum of trades. Nevertheless, the AEC industries gener-
18.6, 37.14, and 55.7%, respectively. Because EC and CT changed ally do not have omnipotent workers. In this research, the layouts of
at the same rate, the reductions of EC were also 18.6, 37.14, and the stations under study were carefully analyzed to make the
55.7% when adding 1, 2, and 3 workers, respectively. assumption of interchangeable workers practical. This assumption,
When moving idle workers around to balance the production together with the observations of the simulation and optimization
line and increase the processing rate of the station identified as a processes, generates an innovative suggestion to the AEC indus-
bottleneck, the assumption is that workers are allocable and can per- tries. Multiskilled, interchangeable workers may have better market
form each other’s jobs. This assumption is an ideal situation and competitiveness in the production of modular construction than the
arguable under the current conditions of the architecture, engineer- workers who have single, particular, and fixed skillsets. With the
ing, and construction (AEC) industries. Interchangeable workers growth of modular construction, the demands of labor market are
are usually able to devote themselves to the tasks of one or several changing, which is noteworthy to the AEC industry practitioners

© ASCE 04017034-12 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 and educators. Moreover, the simulation and optimization model construction. The research also provides a new angle to study Lean
developed in this research can help further discussions on how to construction.
optimize the production process of modular buildings under the This research optimized the manufacturing process of modular
assumption of limited interchangeable workers. construction. Compared with the workers who cannot be reallo-
In modular construction, savings on project duration, material cated, this research suggests that the workers with interchangeable
usage, and labor hours; improvement on project quality and work- skillsets are preferred in the Lean improvement of energy effi-
ers’ job satisfaction; and financial performance should all be consid- ciency. The application results may bring long-term benefits to
ered as evaluation criteria of the production of modular construc- modular construction. Because the demand for modular buildings is
tion. For the companies that perform modular construction on the rise, application of this model would not only reduce the
temporarily, their challenges of modular manufacturing include overall production cost in terms of energy saving, but also result in
labor availability, fluctuations of labor costs, and uncertainties of faster construction time and reduction of the cost of waste manage-
the cost of leasing large warehouses and material costs. In this ment. Another benefit includes the identification and reduction of
research, the modular manufacturer under consideration has perma- nonvalue-added activities with quantitative proofs. In addition, with
nent production and storage facilities, a stable labor group, and rela- less use of energy, it would contribute to the reduction of the CO2
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

tively steady material supply. Under the stabilized conditions, the emission to the environment and help tackle the climate changes.
following evaluation of financial performance is mainly on how The limitations of this research include the following items. (1)
much the contribution of energy is to the labor cost of modular man- There were only four workstations in the simulation and analysis of
ufacturing. Fig. 11(b) shows that when adding two workers to the system. Modular construction in the real world may have more
Station 3, the CT is relatively balanced. Therefore, there are two complicated situations. (2) The US construction industry cost data
workers at Station 1, two workers at Station 2, four workers at (Robert Snow Means Company 2016) published machine specifica-
Station 3, and two workers at Station 4. The average price of elec- tions (OMNITECH 2017), and observations at a local modular con-
struction company were used in this research. Individual companies
tricity to industry customers is 6.8 cents/kW · h in July 2017 (EIA
may consider adjusting the data used in the simulation process of
2017). For a small size factory (i.e., 10 workers), the saving of
the system. For example, future research needs to monitor energy
$1402.87/day has considerable significance and impact on the
use of individual stations and overall energy use of factories to
research among the overall financial model of modular operations:
determine actual EC of use in a factory environment. Further, EC
1. Saved EC at Station 3: 2075.5 – 1304.6 = 770.9 kW
for machines is one of many energy-consuming factors including
2. Saved CT: 593 – 372.74 = 220.26 min
heating and cooling factory space, transportation energy for work-
3. Saved energy: (6.81/100)  (593/60)  770.9 = $518.86 per
ers, deliveries, and so forth. It may be that equipment energy is not a
cycle
significant contributor in the overall energy use in factories.
Suppose a worker’s salary is $50/h:
Future research should look at other types of modular buildings
1. Saved labor cost: 10 workers  (220.26/60) h  $50/h =
and facilities in addition to the residential type used in this research,
$1835.5
explore the effects of a more complex production system with more
2. Additional workers’ salary: 2 workers  $50/h  (372.74/60) =
than four stations, and collect real-time data from the modular con-
$621.23
struction industry rather than use general industry cost databases.
3. Overall saved energy and labor: (518.86 þ 1835.5 – 621.23) 
Another research direction is a study on how to optimize the pro-
(8  60/593) = $1402.87/day
duction process of modular buildings under the assumption of lim-
ited interchangeable workers. In addition, the studies of different
Conclusions and Recommendations business models and how they affect modular construction will be
helpful to the AEC industries.
The purpose of this study was to design and verify a measurement
and optimization system to calculate the EC quantitatively in the
References
manufacturing process of modular construction. The EC measure-
ment and optimization system was able to identify nonvalue-added Anderson, N. C., and Kovach, J. V. (2014). “Reducing welding defects in
tasks based on long CTs and possible causes of unnecessary long turnaround projects: A lean Six Sigma case study.” Qual. Eng., 26(2),
CT in modular construction to make the EC process more efficient. 168–181.
The model would help the modular construction industry in reduc- Arbenz, P., and Kressner, D. (2012). Lecture notes on solving large scale
ing EC. It allowed for the quantitative verification of the effective- eigenvalue problems, ETH Zurich, Zurich, Switzerland, 2.
ness of Lean Six Sigma in manufacturing for modular construction. Blismas, N., and Wakefield, R. (2009). “Drivers, constraints and the future
This research for the first time identified waste factors and their of offsite manufacture in Australia.” Constr. Innovation, 9(1), 72–83.
Breyfogle, F. W., III, Cupello, J. M., and Meadows, B. (2001). Managing
influences quantitatively on the manufacturing process of modular
Six Sigma: A practical guide to understanding, assessing, and imple-
construction. Reducing EC on nonvalue-added activities is impor- menting the strategy that yields bottom-line success, Wiley, New York.
tant for the modular building industry to keep the cost of the produc- Chen, Y., Okudan, G. E., and Riley, D. R. (2010). “Decision support for
tion as low as possible. This research used several methods to iden- construction method selection in concrete buildings: Prefabrication
tify possible energy efficiency improvements in the modular adoption and optimization.” Autom. Constr., 19(6), 665–675.
manufacturing processes. For example, VSM was used to identify Dajadian, S. A., and Koch, D. C. (2014). “Waste management models and
the areas with process wastes and bottlenecks. The DMAIC process their applications on construction sites.” Int. J. Constr. Eng. Manage.,
of Six Sigma methodology was used to identify the root causes for 3(3), 91–98.
Dashti, H. T., Siahpirani, A. F., Wang, L., Kloc, M., and Assadi, A. H.
waste and bottlenecks and provide suggestions. These factors were
(2010). “Parallel evolutionary computation in very large scale eigen-
confirmed and verified using the LCMU system and the SD process, value problems.” ArXiv, 1008.5391.
respectively. To verify the significance of the data collected, a two- Domínguez, S., Sendra, J. J., León, A. L., and Esquivias, P. M. (2012).
sample paired t-test was conducted. The computation details of the “Towards energy demand reduction in social housing buildings:
results from the data analysis of this research shed light on EC in Envelope system optimization strategies.” Energies, 5(7), 2263–2287.

© ASCE 04017034-13 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034

 EIA (U.S. Energy Information Administration). (2017). “Electric Power McGraw Hill Construction. (2010). “Prefabrication and modularization:
Monthly: Table 5.6.A. Average price of electricity to ultimate customers Increasing productivity in construction industry.” h[Link]
by end-use sector.” h[Link] .[Link]/sites/default/files/documents/el/economics/Prefabrication
_grapher.php?t=epmt_5_6_ai (Jul. 14, 2017). -[Link]
Feng, Y. Y., Chen, S. Q., and Zhang, L. X. (2013). “System dynamics mod- (Feb. 17, 2017).
eling for urban energy consumption and CO2 emissions: A case study of Modular Building Institute. (2010). “Improving construction efficiency and
Beijing, China.” Ecol. Modell., 252, 44–52. productivity with modular construction,” Charlottesville, VA.
Gann, D. M. (1996). “Construction as a manufacturing process? Similarities Modular Building Institute. (2015). “Permanent modular construct-ion
and differences between industrialized housing and car production in 2015 annual report.” h[Link]
Japan.” Constr. Manage. Econ., 14(5), 437–450. /2015/08/[Link] (Mar. 8, 2017).
Goodier, C. I., and Gibb, A. G. F. (2005). “Barriers and opportunities for Nadim, W., and Goulding, J. S. (2011). “Offsite production: A model for
offsite in the UK.” Proc., Systematic Innovation in the Management of building down barriers: A European construction industry perspective.”
Project and Processes, Helsinki Int. Joint Symp., Abdul Samed Kazi, Eng. Constr. Archit. Manage., 18(1), 82–101.
ed., VTT Technical Research Centre of Finland, Espoo, Finland. OMNITECH. (2017). h[Link]
Hamelin, M.-C., and Zmeureanu, R. (2012). “Multi-objective life cycle -casegoods-router/#tab-id-3i (Nov. 17, 2017).
Pan, W., Gibb, A., and Dainty, A. (2012). “Strategies for integrating the use
Downloaded from [Link] by University Of Florida on 12/26/17. Copyright ASCE. For personal use only; all rights reserved.

optimization of a single-family house envelope.” Proc., eSim 2012:

The Canadian Conf. on Building Simulation, International Building of off-site production technologies in house building.” J. Constr. Eng.
Performance Simulation Association (IBPSA), Ottawa. Manage., 10.1061/(ASCE)CO.1943-7862.0000544, 1331–1340.
Hammer, M., and Goding, J. (2001). “Putting Six Sigma in perspective.” Pan, W., Gibb, A. G., and Dainty, A. (2005). “Offsite modern methods of
Quality, 40(10), 58. construction in housebuilding–Perspectives and practices of leading UK
Howell, G., and Ballard, G. (1994). “Implementing lean construction: house builders,” Buildoffsite, London.
Reducing inflow variation.” Proc., 2nd Annual Conf. on the Int. Group Quinn, J. D. (2008). “Modeling the resource consumption of housing in
New Orleans using System Dynamics.” Ph.D. dissertation, Massachusetts
for Lean Construction, Lean Construction Institute, Arlington, VA.
Institute of Technology, Boston.
Howell, G. A. (1999). “What is lean construction–1999?” Proc., 7th Annual
Robert Snow Means Company. (2016). RSMeans: Building construction
Conf. of the Int. Group for Lean Construction, Lean Construction
cost data 2016, 74th Ed., RS Means, Rockland, MA.
Institute, Arlington, VA, 1–10.
Saad, Y., and van der Vorst, H. A. (2000). “Iterative solution of linear sys-
Josephson, A. (2016). “The pros and cons of prefab homes.” hhttps://
tems in the 20th century.” J. Comput. Appl. Math., 123(1–2), 1–33.
[Link]/mortgage/the-pros-and-cons-of-prefab-homesi (Mar. 8,
Soršak, M., Leskovar, V. Ž., Premrov, M., Goricanec, D., and Pšunder, I.
2017).
(2014). “Economical optimization of energy-efficient timber buildings:
Kelley, C. T. (1999). Iterative methods for optimization, Society for
Case study for single family timber house in Slovenia.” Energy,
Industrial and Applied Mathematics, Philadelphia.
77(Dec), 57–65.
Koskela, L. (1992). “Application of the new production philosophy to con-
Sterman, D. J. (2002). System dynamics: Systems thinking and modeling for
struction.” Tech. Rep. TR072, Center for Integrated Facility Engineering, a complex world, McGraw-Hill, Boston.
Stanford Univ., Stanford, CA. Storn, R., and Price, K. (1997). “Differential evolution–A simple and effi-
Lau, L. C., Ravi, R., and Singh, M. (2011). Iterative methods in com- cient heuristic for global optimization over continuous spaces.” J.
binatorial optimization, Vol. 46, Cambridge University Press, Global Optim., 11(4), 341–359.
Cambridge, U.K. Taghaddos, H., Hermann, U., AbouRizk, S., and Mohamed, Y. (2014).
Lee, K. S., and Geem, Z. W. (2005). “A new meta-heuristic algorithm for “Simulation-based multi agent approach for scheduling modular
continuous engineering optimization: Harmony search theory and prac- construction.” J. Comput. Civ. Eng., 10.1061/(ASCE)CP.1943-5487
tice.” Comput. Methods Appl. Mech. Eng., 194(36–38), 3902–3933. .0000262, 263–274.
Li, Z., Shen, G. Q., and Xue, X. (2014). “Critical review of the research on Tang, V., and Vijay, S. (2001). “System dynamics: Origins, development,
the management of prefabricated construction.” Habitat Int., 43(Jul), and future prospect of a method.” Proc., ESD.83, Fall 2001 Research
240–249. Seminar in Engineering Systems, h[Link]
Little, J. D. C. (1961). “A proof for the queuing formula: L = l W.” Oper. /notebook/[Link] (Nov. 17, 2017).
Res., 9(3), 383–387. Wilkinson, J. H. (1965). The algebraic eigenvalue problem, Vol. 87,
Liu, J. J., Hou, L., and Wang, X. Y. (2014). “A discrete firefly algorithm for Clarendon Press, Oxford, U.K., 570–644.
the scaffolding modular construction in mega projects.” Proc., 31st Yu, H., Al-Hussein, M., Al-Jibouri, S., and Telyas, A. (2013). “Lean trans-
ISARC, International Association for Automation and Robotics in formation in a modular building company.” J. Manage. Eng., 10.1061
Construction (IAARC), Bratislava, SK-Slovakia. /(ASCE)ME.1943-5479.0000115, 103–111.
Manrodt, K., Vitasek, K., and Thompson, R. H. (2008). “Lean practices in Zawidzki, M., and Nishikawa, I. (2010). “Discrete optimization of mod-
the supply chain.” Jones Lang LaSalle, Chicago. ular truss network in a constrained network.” Proc., 6th China-
Mao, C., Shen, Q., Shen, L., and Tang, L. (2013). “Comparative study of Japan-Korea Joint Symp. on Optimization of Structural and Mechanical
greenhouse gas emissions between off-site prefabrication and conven- Systems, Japan Society of Mechanical Engineers, Tokyo, Japan, and
tional construction methods: Two case studies of residential projects.” International Society for Structural and Multidisciplinary Optimization,
Energy Build., 66(Nov), 165–176. Baltimore, MD.

© ASCE 04017034-14 J. Archit. Eng.

J. Archit. Eng., 2018, 24(1): 04017034