Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/238179289
CITATIONS READS
69 193
2 authors:
Some of the authors of this publication are also working on these related projects:
Autonomous Vision-based UAV Systems for Construction Progress Monitoring View project
All content following this page was uploaded by Ming Lu on 05 February 2016.
Elements of a Useful
Estimation Tool
WAI KIONG CHONG, M.ASCE; SANG-HOON LEE, M.ASCE; AND
C
onstruction schedule development is Designers do not have control over many productiv-
a critical process during the design ity variables, such as the means and methods of con-
phase of a highway construction struction and the productivity of equipment and
project. Under- or overestimation of labor. Instead, they take the “big picture” approach,
a project schedule can cripple the seeking an ideal schedule, not an exact one. They
smooth progress of a project; a tight aim to develop a logical and reliable schedule that is
schedule may increase the chances of excessive claims based on the limitations normally faced in highway
and delays, whereas a loose schedule may cause construction projects. The contractor’s schedule is nor-
project idling during construction, increasing the mally more thorough and detailed than the designer’s,
chances of material and equipment damage by as contractors need total control in every construction
bad weather and safety hazards for pedestrians and process. As a result, the designer’s approach to schedul-
drivers. ing can be very different from the contractor’s approach.
provided by the DOTs, and 22% used production from learning (Thomas et al. 1999; El-Rayes &
rates from historical records of previously completed Moselhi 2001). For example, it makes sense to link
projects from other sources. Data used to develop these the productivity of foundation, pipeline, and retaining
systems were normally collected either from survey re- wall construction to soil conditions. Alternatively, the
turns or documentation from projects. Some DOTs learning curve and poor weather conditions have a
and several contractors had simplified their production greater effect on highly repetitive work items such
rate estimates into three values—mean, optimistic, as pavement, multiple drilled shafts, and hot-mix as-
and pessimistic values for each work item—and phalt pour. Thomas et al. (1999) found that changing
designers determined a realistic rate from these values weather, erroneous work, poorly coordinated material
based on their experience. delivery, and frequent equipment relocation were
more disruptive to long and continuous production
processes than to less repetitive work items. Stoppage
PRODUCTIVITY FACTORS to a process slows down work momentum and leads to
In production scheduling, productivity factors are productivity losses of other work, which Thomas et al.
used to establish relationships between fluctuating (1999) described as the “ripple effect.” However, be-
production rates and production work items. Design- cause a highway designer cannot accurately predict the
ers must identify relevant productivity variables they chances of having disruptions during construction,
can rely on to improve their estimates. such disruptions are normally excluded from the
Differing geological locations, site conditions, and schedule at the design phase.
other location variables affect work productivity. Con- Design can significantly affect work productivity
structing a drilled shaft on dry soil may be faster than (Poh and Chen 1998). Constructability has been
constructing one nearer a riverbank, assuming the same shown to increase site productivity, and site conges-
crew is used in both operations. Constructing a road on tion can be avoided with designs that use smaller
a mountain may be much slower than constructing one equipment.
on flat land. A sudden change in the weather can dis- Site condition would normally be considered a
rupt or delay construction operation. Construction productivity factor. Construction on mountainous
work has to stop during snowstorms in the warmer re- regions, in tight work spaces, in extreme cold and heat,
gions (e.g., Florida and Texas) but can continue in the on rough terrain, on congested work zones, and in close
colder regions (e.g., Maine and Washington). Rural proximity to adjacent structures would normally slow
highway projects face fewer traffic problems than high- down workers’ productivity (Koehn and Brown 1985).
way projects in metropolitan areas, which face frequent Wideman (1994) showed that the productivity of
traffic congestion, strict environmental regulations, workers varied during different phases of construction.
risky traffic safety, and right-of-way issues that reduce His study found that workers’ productivity was slow
construction productivity. Rural highway construc- during the early phase of construction and slowly sped
tion, however, may face difficulties in acquiring skilled up as construction progressed. Productivity continued
workers and highly productive equipment, driving to rise and plateau between 25% and 75% of project
down productivity (Koehn and Ahmed 2001). Thus, completion, and then fell until project completion. He
the location of the project is not a dependable produc- attributed the initial growth to workers’ learning
tivity factor; in a work zone located in a metropolitan effect and the fall nearing project completion to re-
area that has ample work space, highway congestion duced work amount. Thus, worker productivity
may affect only material delivery and may have mini- should not be treated as uniform throughout the
mal impact on work operations. construction phase.
availability, job and management conditions, drilling and Shi 2005; Lee 2005; Lee and Arditi 2006). Several
time activity, other time activities, change orders, and attempts have been made to integrate some of these
weather (Zayed and Halpin 2004; Hanna et al. 2002; techniques with standard scheduling software, such
El-Rayes and Moselhi 2001). Furthermore, Thomas as Primavera, Microsoft Project, and SureTrak. Alter-
et al. (1989) emphasized that productivity factors natively, many designers continue to develop project
should be divided into within-project, between- schedules using their own experience. Advanced
project, and regional drivers. Such divisions allow scheduling techniques may improve a schedule esti-
designers to better identify and allocate significant mate’s accuracy, but many departments of transporta-
productivity drivers for different work zone scenarios. tion lack the required infrastructure and information
to use these techniques. In addition, scheduling tools
developed in the past for specific purposes became out-
PRODUCTION RATE DOCUMENTATION AND dated very quickly. Texas DOT’s CTDS (Hancher et al.
ANALYSIS 1992), for example, became obsolete simply because
Developing reliable productivity data documentation there was no way to update the information in the sys-
and analysis processes is critical to the development of tem. Thus, any new scheduling tool should support
a dependable scheduling tool. Data accuracy and popular scheduling software, such as Primavera and
correct analysis are the two critical elements of any Microsoft Project.
successful and reliable information system. Collected Lessons learned from the past have shown that
productivity data should support the following func- schedulers tend to resist a complicated information
tions: (1) identify significant productivity factors, technology system and prefer a flexible system. Be-
(2) examine and establish the relationships between cause most designers use the critical path and Gantt
factors and productivity, (3) develop production rate chart methods to schedule their projects, Primavera
models for scheduling, and (4) study productivity and Microsoft Projects are the most frequently used
improvement methods. Efficiency of data collection scheduling programs. Therefore, production rate esti-
is also critical, though there is a need to balance effi- mation tools should not deviate too much from these
ciency and reliability. Survey forms and existing pro- techniques and software. Production rate information
ductivity databases are the two most common data that is easily integrated with the software and tech-
collection methods. Hancher et al.’s (1992) survey niques will allow designers to spend less time learning
of experienced designers and site personnel in 36 new techniques and software and more time improv-
DOTs indicated that many contractors kept and de- ing the reliability of their estimates. Consequently,
veloped their own production factor and rate informa- any production rate system should be either com-
tion. They gathered most of their productivity pletely independent from or well integrated into
information from daily log books, payment and sched- existing software and estimation techniques.
ule documentation, record books, and information Researchers have successfully applied many tech-
stored on computers. Some contractors required their niques to improve the reliability of production
staffs to input information systematically and kept rate estimates. Some of these techniques include re-
comprehensive records of all their projects. No proce- gression analysis and models, factor analysis models,
dures or standards have been developed for such Monte Carlo simulation, and neural networks. Other,
recording processes; thus, the reliability of the infor- more simplified methods include summing up col-
mation may vary greatly across organizations. Like any lected production rates and averaging them into mean,
research using the survey approach, the reliability of optimistic, and pessimistic values (Hancher et al.
Hancher et al.’s findings cannot be verified, and most 1992). The techniques an organization adopts depend
statistical and simulation techniques to develop a AbouRizk and Wales (1997) illustrated that such a
six-value estimation technique: minimum duration project simulation model should consist of three com-
(optimistic), maximum duration (pessimistic), mean, ponents: (1) a project network that maintains schedule
standard deviation, confidence interval, and probabil- logic, (2) a stochastic and random particles model that
ity. This technique allows estimators to better under- generates uncertain factors, and (3) a productivity
stand the risks and confidence of their estimates and model that relates uncertainty of productivity factors
eliminates the need to guess. Regression analysis is one to generated project conditions. Each activity could be
of the most common methods applied to examine and simulated individually and combined at a later stage
quantify the relationships between productivity fac- (discrete-event continuous simulation). Discrete
tors and rates (Koehn and Brown 1985; Sanders events could be combined for simultaneous and con-
and Thomas 1993; Hanna et al. 2002). Regression tinuous simulation. Their models relied on historical
analysis is useful to illustrate the continuous relation- data such as weather data from meteorological agen-
ships between numeric productivity factors and cies. Construction processes are broken down into
rates. However, it cannot be used to develop relation- individual activities based on their relationships with
ships between nonnumeric (categorical) productivity various productivity factors. The effect of a productiv-
factors and rates. For such factors, Hancher et al.’s ity factor on an individual activity is measured by the
(1992) method is the most common approach; the simulation process and is later combined into a
contractor determines a productivity rate under nor- schedule.
mal conditions (normally, a mean or median) and Dzeng and Tommelein (1997) suggested breaking
two rates for extreme conditions (normally, optimistic down construction projects into “cases” and automat-
and pessimistic). Other statistical techniques are avail- ing the duration estimating process for each case, and
able to handle nonnumeric factors (categorical factors), they found that their schedule estimates were more ac-
such as box plots and longitudinal data analysis tech- curate. The proposed application of neural networks to
niques (commonly used in social science research). The construction scheduling (Adeli and Karim 1997),
more advanced techniques require special treatment of although seeming to help improve the reliability of
data, like creating artificial variables and splitting data construction scheduling through continuous knowl-
into different components. edge learning, has limited application at the design
Peña-Mora and Li (2001) proposed using the stage, especially on improving the reliability of
GERT (Graphical Evaluation and Review Technique) production rate estimates. Adeli and Karim (1997)
diagramming scheme to calculate the probability of stated that neural dynamic models require breakdown
project duration by measuring the variability of differ- of work into tasks, crews, and segments, and logics and
ent branches and loops of each construction activity, constraints between repetitive tasks have to be devel-
their relationships, and overlaps with other nonrele- oped while each task is simulated individually. Exces-
vant activities. They claimed that such a scheme could sive amounts of data are required to ascertain the
better control or even eliminate variability within a usefulness of neural dynamic models since the accuracy
schedule. Park and Peña-Mora (2004) proposed using of the final estimates heavily depends on inputs to the
reliability control to refine activity buffers and applied system. Such requirements of massive and continuous
simulation to measure and reduce buffer variability information inputs make the neural network ineffi-
between activities. However, designers may not have cient and impractical to be applied at the design phase.
sufficient control and may lack the ability to predict Bonnal et al. (2004) believed that fuzzy logic had
buffer variability between activities during design, become sufficiently mature to be applied to project
and they cannot feasibly apply both techniques at scheduling in real life. They claimed that fuzzy logic
(Vanhoucke 2006; Fan and Tserng 2006) requires de- were small, manual calculation helped speed up the
signers to look at different ways of treating their data. estimation process. The feedback also highlighted that
In real life, these variables do not exist, so it may be accurate production rate information was needed that
difficult for designers to understand and appreciate reflected and represented the situations and conditions
their meanings and purposes. These techniques also in different districts and regions, like calendar day and
require extensive algorithms to estimate other resource working day contracts, rural areas, soil conditions, and
constraints (Jiang and Shi 2005). Algorithms are regional profiles. Many designers highlighted that the
needed to clarify relationships between activities units adopted should be in line with the pay units, as
and allow schedulers to better control variability be- this would reduce their work to convert the units
tween activities. Refinement and reliability of the between schedule and payment.
schedules could be enhanced with algorithms. We selected eight Texas DOT designers for inter-
Construction schedules are dynamic and thus need view with regard to their needs for a better scheduling
to be updated frequently to reflect changes during tool. Their feedback included the following seven
construction. El-Shahhat et al. (1995) found that be- points:
tween 29% and 67% of all construction errors oc-
curred at the design phase, while only 12% to 59% 1. They hoped to separate the production rates and
of errors occurred at the construction phase. Errors scheduling tools from their scheduling software
committed at the design phase may be magnified to make the system more flexible and adaptive
at the construction phase and cause problems during to many situations.
construction operations. Thus, designers cannot rely 2. They wanted more realistic rates that reflected
too much on their personal experience. actual site conditions. Many of their rates were
unrealistic and were not collected from the site.
3. They wanted a system that was user friendly
and did not hamper the contributions of the
A REVIEW OF SCHEDULE ESTIMATION designers. The designers needed to have more
PRACTICES AND NEEDS control over the rates and to be able to adjust
A survey conducted by the Texas Department of the rates accordingly so that they would not
Transportation found that scheduling practices varied be forced to accept unrealistic rates.
across districts and regions (O’Connor et al. 2005). 4. They cited the need to know productivity fac-
Most of the practices were driven by the needs of tors at the design stage and to exclude factors
the districts, the top management, and the demands that cannot be predicted at the design stage.
of the engineers in those districts. Some districts even 5. Most agreed that statistical tools can help im-
developed their own system to handle their unique prove the accuracy of their schedules, but they
needs and situations. Each district had authority over preferred these tools to be relatively simple and
its own scheduling methods, and employees were practical. They noted that complicated statisti-
given the option to adopt anything that suited them. cal and simulation techniques make any tool
Scheduling software used in the Texas DOT included unfriendly and yield information that may be
in-house systems like Primavera, Microsoft Project, difficult to interpret.
and SureTrak. The type of software used depended 6. Because it is impossible to establish good rela-
heavily on the preferences of the employees. Several tionships among the many productivity factors
districts did not use the Texas DOT’s standard system for any work item, they suggested that a system
(i.e., CTDS), as they found that it was inflexible and should not be restrictive in establishing these
that the information was not accurate. They preferred relationships.