Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Dynamic Modeling of Delay and Disruption Claims

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5
At a glance
Powered by AI
The key takeaways are that the system dynamics approach can be used to model disruption on construction projects and quantify productivity losses. It allows simulation of how disruption propagates through the project network.

The system dynamics approach takes a comprehensive view of the project and focuses on feedback processes. It can be used to simulate a project's performance over time including the impacts of changes. The difference between the actual and simulated performance without changes quantifies the impacts.

The seven main steps to establish a system dynamics model are: 1) Identify scope of each activity 2) Assign original workhours 3) Establish activity network 4) Identify causal relationships and assign coefficients 5) Calculate affected hours 6) Sum affected hours in each iteration 7) Repeat until impacts are negligible.

T ECHNICAL A RTICLE

System Dynamic Modeling of


Delay and Disruption Claims
William Ibbs and Min Liu

have occurred in the absence of changes.


The difference between "what happened"
and "what would have happened" is the full
impact of those changes and the disruption
[1, 7].
There are four purposes of a SD
model:
1.

ABSTRACT: The construction environment is uncertain, dynamic and complex in nature. The
idea that small disruptions and delays can ripple through the project and mushroom in larger,
more serious consequences is well established. Nevertheless, it is hard to identify and accurately
prove the consequences of such disruptions. The system dynamic (SD) method is presented in
this article as a tool that can show the logic link between cause and effect; that is,
disruption/delay/acceleration and productivity loss. This in turn helps quantify productivity loss.
Because the SD technique recognizes and models the interaction of work activities and graphically illustrates the mechanism by which disruption occurs, the results are accurate and persuasive. This article describes how to establish and use the SD model for acceleration, delay, and
disruption purposes with a simple example. The importance and sensitivity of the coefficients
in the SD model are also discussed.

2.
3.

4.

Demonstrate causalityshow what


events triggered a disruption and how
the triggers propagated through the
project.
Assign responsibilityidentify the
party responsible for these impacts.
Quantify the impactshow that the
events that caused acceleration, delay
or disruption, and created a specific
time and cost overrun in the project.
Convince the project stakeholders of
points 1, 2, and 3.

KEY WORDS: Acceleration, construction claims, delay, disruption, and labor productivity

cceleration, delays and disruptions


are frequently encountered on
construction projects and one of
the main reasons for productivity loss.
Uncertainties surrounding disruption not
only endanger contractor profit margins
but also make parties less amenable to
settlement. This consequently delays the
resolution of disputes and increases costs
for all concerned.
Owners prefer to know the full cost
impact of any proposed change prior to its
authorization. Contractors, on the other
hand, prefer to claim impact costs based on
a single, overall calculation oftentimes
submitted after project completion. A
stalemate may result, poisoning change
negotiations and driving up costs.
The system dynamics (SD) approach
is a tool for quantifying such complicated
impact costs. By recognizing the specific
characteristics of and interactions between
schedule activities, the SD model can
simulate disruption throughout the entire
network.
Because it is a graphical
illustration of the project and the changes
to the project, SD is useful for visualizing
and understanding these project triggers,
and the acceleration, delay, and disruption
that results. This in turn helps pinpoint
responsibility for the change and estimate
the quantum of the change.
The purpose of this article is to
illustrate how to use the SD method by
modeling a simple example. One of the
most important parameters in establishing
the SD model is the coefficient that

12

There are seven principal steps to


establishing
a SD model:
quantifies the correlation between two
adjacent schedule activities.
The
importance of accurately estimating those 1. Identify the scope of each work
activity.
coefficients is also discussed in this article.
2. Assign the original workhours to each
activity.
SYSTEM DYNAMIC APPROACH
3.
Establish a network to model the
The SD approach is based on a
relationships between the work
comprehensive view of the project and
activities.
focuses on the feedback processes that take
4.
Identify the causal relationship
place within the project system. It offers a
between each work activity and assign
rigorous method for the description,
the coefficients. Some predecessors
exploration, and analysis of complex
might have substantial impact (high
project
systems
comprised
of
coefficient value) to the successor,
organizational elements, the project work
some might have very slight impact
packages
and
the
environmental
(low coefficient value), and some
influences [8].
might have no impact at all
Kenneth G. Cooper [1, 2] used the SD
(coefficient is zero).
approach in the Ingalls Shipbuilding case
5.
Calculate
the accelerated, delayed or
against the US Navy in the late 1970s. This
disrupted
hours to an activity by
case involved owner (US Navy) delays,
multiplying
the coefficients of all its
mainly stemming from design changes.
predecessor(s)
against its work hours.
The total settlement was $447 million.
6.
Add
the
total
affected hours for all
The model was the basis for at least $200activities
in
this
iteration.
300 million of this settlement [2]. Since
7.
Repeat
step
6
until
the impact of an
this important precedent, the method has
iteration
is
negligible.
Sum the
been used in litigation, though to-date still
affected
hours
of
all
activities
in all
outside of construction. Part of this lack of
iterations.
acceptance is because of its retrospective,
rather than contemporaneous nature.
In essence the SD model provides a
simulation of project performance as the
project is built, including both customer
changes (e.g., design changes) and varying
project conditions (such as the labor
market, regulatory changes, etc). The
numerically validated model can then be
used to simulate the project as it would
Cost Engineering Vol. 47/No. 6 JUNE 2005

There are some problems with SD that


prevents its widespread use and acceptance
by courts. Howick, for example, noted that
the SD models are frequently drawn in too
limited a fashion [5]. Disruptions on an
adjacent project may not be captured when
modeling a second project. Similarly,
modeling work of one subcontractor may

exclude impacts associated with other


subcontractors.
Susan Howick [6] also explored
criteria that should be used to identify
whether a situation could be modeled
using SD. Her criteria are:
1.

2.
3.

Does the situation contain feedback


loops and can those loops explicitly
model the situation?
Is there sufficient, reliable information
and data available?
Will SD modeling explain the project
more clearly than some other
modeling technique?

One other criterion is whether the


dispute justifies the amount of time and
effort required to model the situation.
APPLICATION FOR DELAY
CLAIMAN EXAMPLE
An example from Colin Eden is
adapted here to illustrate how to quantify
labor productivity loss caused by owner's
delay in decision-making [3].
An owner hired a contractor to make
modifications to his house. The contract is
fixed price and involves modifications to a
bedroom, kitchen, and bathroom.
Half way through the contract the
carpenter was working in the bedroom
laying a new wooden floor. The plumber Table 1 Delay Calculation With the Original Coefficients

Figure 1 The SD Model for Disruption and Delay Analysis


Cost Engineering Vol. 47/No. 6 JUNE 2005

13

Table 2 Security Analysis of the Coefficients

was working in the kitchen moving the sink


and installing a new waste disposal unit.
The carpenter realized that the owner had
not yet told the contractor (his employer)
where the owner wanted his new power
outlets and telephone sockets. As a result
the electrician was unable to lay the
cabling which would allow the carpenter to
close the flooring.
So, the carpenter decided that the best
thing was to work around the problem by
laying most of the floor. That carpenter
then moved to installing the kitchen
cabinets early, in order to keep working.
But it got crowded in the kitchen with the
carpenter and plumber fighting for space,
and their work was slowed because of the
congestion.
The carpet layer arrived at the jobsite
on time to start work in the bedroom,
which meant the carpenter had to get the
electrician to complete the cabling. This
resulted in an interruption to the
electrician, who had been installing the
power supply to the new shower in the
bathroom. However he changed tasks and
made his best guess on the location of
outlets, which would allow the carpenter to
lay the floor. This meant that the carpenter
stopped working on the kitchen cabinets.
While all these activities were being
worked on by the electrician and carpenter,
the carpet layer waited. Unfortunately,
after the electrician and carpenter finished
working on the floor, the owner notified
them to relocate the power outlets and
telephone sockets. (The carpenter's guess
about the sockets was wrong and rework
resulted.) This delayed the carpet layer.
And even though the electrician is
competent and experienced, he mistakenly
omitted part of his work in the bathroom
when he went back to the previous task.
An activity-level SD model for this
small example is presented in Figure 1.
From steps 1 and 2, the scope and work

14

The total delay time in this example is


hours of each activity are identified; see
177.3 hours. The original total labor hours
Table 1.
needed to finish these six activities was 210
In step 3 the logic sequences of 21 hours. The results show that owner's delay
links are established. In this model each affected the labor productivity significantly,
work activity represents a separate joint as by 84.4%. Table 1 also shows that the
in a structural frame. For example, arrow 1 indirect labor productivity loss is much
shows the owner-caused delay to more than the direct labor productivity
electrician when he was working in the loss: 7 hours vs. 170.3 hours. In practice
bedroom.
Arrow 2 shows that the this indirect, ripple impact is usually hard
electrician could not finish cabling and to prove and very arguable. The graphical
nature of the SD model helps to
diverted to work in the shower.
demonstrate such ripple effects.
In step 4 the causal coefficients are
SENSITIVITY ANALYSIS
developed according to the relationship of
OF THE COEFFICIENTS
the two activities and the "stiffness" (to
borrow terminology from network methods
One of the most important tasks in
of structural analysis) of the successor
activity. If an activity has no disruptive establishing the SD model is to estimate
Coefficients
effect on its successor activity, the the coefficient values.
coefficient of the causal relationship represent how much disruption ripples
linking the two activities will be 0. Though through the network. The analyst has a
a more complicated situation, the choice of how to define these coefficients.
coefficients can be allowed to change as They can be based on either the
the model cycles through different predecessor or successor activity's
iterations [4]. Step 5 computes the total workhours. In this example they are based
delay of the current iteration. The process on the successor activity's workhours
is continued until further iterations result because a delay to the successor activity is
highly dependent on the characteristics of
in minimal impact; see Table 2.
the successor activity. For example, the
The results are presented in Table 1 material, equipment, working space, and
for three iterations. For this example the supervision constraints of an activity have
delays are 125.5, 48.1, and 3.7 hours in the significant influence on its own
first, second and third iteration, productivity performance. In other words,
respectively, meaning a grand total of 177.3 when the coefficient of A-B = 0.2 in the
first iteration, the B (50 workhours) activity
hours.
is delayed by 10 (0.2*50=10) hours.
These coefficients are a function of the
From Table 2 it can be seen that
activity G is the most seriously delayed. linked activities' locations, times of
Even though the carpet layer arrived on performance, types of work, crew sizes and
time, he could not start to lay the carpet compositions, supervisors, and numerous
until the wooden floor was finished. So in other factors. Specifically, productivity
the first iteration 30 extra hours were spent impacts can be caused by factors such as
to solve all the problems associated with work area congestion, resource diversion,
skill dilution, stacking of trades, waiting for
the owner's delay in decision-making.
shared equipment, and dilution of

Cost Engineering Vol. 47/No. 6 JUNE 2005

supervision.
When estimating these
coefficients we need to collect information
that includes each of these factors. More
than the two directly linked activities may
need to be considered. For example, when
the coefficient for C-G was estimated, we
considered all the activities delayed before
activity G could start, because it represents
the time the carpet layer had to wait. For
the first iteration, 30 extra hours were
wasted before the carpet layer received the
notice to start. For the second iteration
another 30 extra hours were spent.
Estimating the coefficients is not easy,
yet very important to accuracy of the SD
model. To underscore the importance of
accurate coefficients, Table 2 is included to
show what happens when the coefficients
are changed by just +/- 5%. When the
coefficient for link C-G1 increases 5%,
from 1.5 to 1.575, the total delayed hours
increase from 177.3 to 180.5. Figure 2
illustrates how the total amount of delay
hours change as this one coefficient
changes. Any SD analysis should include
such a sensitivity analysis.
D modeling is one effective tool for
quantifying productivity loss. The
concept and process proposed in
this article can be used in acceleration,
delay, and disruption claims.
It is
especially useful to quantify and portray
indirect (ripple) productivity losses and to
determine which activity causes the largest
amount of delay and which activity is
delayed most in a particular project.
The example is presented to show the
SD methodology. The values derived in
this analysis are strictly hypothetical. The
importance of coefficient estimation is also
discussed in this article. In the simple
example presented in this paper, even a 5%
increase or decrease in a single coefficient
causes a notable difference of the result
from the SD model.
Further research is still needed to
develop criteria that help quantify these
coefficients. The authors are pursuing
such.

Figure 2 Sensitivity Analysis of the Coefficients

REFERENCES
1. Cooper, Kenneth G, Naval Ship
Production: A Claim Settled and a
Framework Built, Interfaces, 10, no. 6,
(1980): 30-36.
2. Cooper, Kenneth G, and Thomas,
Mullen, Swords and Plowshares: The
Rework Cycles of Defense and
Commercial Software Development
Projects. American Programmer, 6,
no. 5, (1993): 41-51.
3. Eden, Colin, Williams, Terry,
Ackerman Fran. and Howick Susan,
On the Nature of Disruption and
Delay (D&D) in Major Projects,
Journal of the Operational Research
Society, 51, (2000): 291-300.
4. Finke Michael R., A Better Way to
Estimate and Mitigate Disruption.
Journal of Construction Engineering
and Management, 124, no. 6, (1998):
480-497.
5. Howick Susan, Using System
Dynamics to Analyze Disruption and
Delay for Litigation: Can the
Modeling
Purposes
Be
Me?.
Management
Science
Theory,
Method & Practice, 21, (2001): 1-24.
6. Howick Susan, Should System
Dynamics be Used to Model
Disruption and Delay for Ligation? An
Exploration of Criteria. Management
Science Theory, Method & Practice,
1, (2002): 1-26.
7. Rodrigues Alexandre, and John
Bowers. System Dynamics in Project
Management: a Comparative Analysis
With Traditional Methods. System
Dynamics Review, 12, no.2, (Summer
1996): 121-139.

Cost Engineering Vol. 47/No. 6 JUNE 2005

8.

Wolstenholme, Eric F. System


EnquiryA System Dynamics
Approach. New York : Wiley, 1990.

ABOUT THE AUTHORS


William Ibbs and Min Liu are with the
Department of Civil & Environmental
Engineering, University of California,
Berkeley. You can contact Professor Ibbs by
e-mail at: DRCWIbbs@aol.com.
Technical Articles - Each month, Cost
Engineering journal publishes one or more peerreviewed technical articles. These articles go
through a blind peer review evaluation prior to
publication. Experts in the subject area judge the
technical accuracy of the articles. They advise the
authors on the strengths and weaknesses of their
submissions and what changes can be made to
improve the article.

NOW AVAILABLE

The 2005
AACE International
TRANSACTIONS
This must-have reference contains
presentations featured at the 2005
Annual Meeting in New Orleans,
LA. The CD-ROM is fully searchable.
To order your copy of the 2005
Transactions, visit the AACE
International Online Bookstore at
www.aacei.org/bookstore/welcome.s
html

15

You might also like