Automation in Construction 20 (2011) 10411050

Contents lists available at ScienceDirect

Automation in Construction
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / a u t c o n

Optimal schedule adjustments for supplying ready mixed concrete

following incidents
Shangyao Yan , Han-Chun Lin, Yin-Chen Liu
Department of Civil Engineering, National Central University, Chungli 32001, Taiwan

a r t i c l e

i n f o

Article history:
Accepted 5 April 2011
Available online 4 May 2011
Keywords:
Ready mixed concrete
Incident
Schedule adjustment
Truck dispatching
Network ow

a b s t r a c t
In this study, the authors develop a systematic network ow model designed to help RMC carriers effectively
adjust schedules following RMC mixer breakdowns. A timespace network technique is employed to
formulate the production of the RMC and truck eet ows in the dimensions of time and space. A solution
algorithm, incorporating a problem decomposition technique and the use of a mathematical programming
solver, is developed to efciently solve the problem. Finally, the model and solution method are evaluated by
performing a case study. The test results show that, with the application of the model and the solution
algorithm, mixer breakdowns do not lead to much deterioration in the system operating performance. In
addition, the system operating cost is signicantly improved (by 10.75 %) compared to that obtained using the
manual approach. As a result, the model and the solution algorithm could be useful for actual operations.
2011 Elsevier B.V. All rights reserved.

1. Introduction
In the last few years, ready mixed concrete (RMC) production has
become increasingly automated. The process includes procurement of
materials, feed and mixing proportions, discharging, prolonged
agitation during travel, delivery of concrete to the construction site,
and the placement operation at the site. However, if an incident (e.g.,
machine malfunction, electricity problem, etc.) occurs so that some of
the mixers cannot produce concrete, the supply of concrete to the
RMC trucks is interrupted which interferes with the dispatch
schedule. The RMC production and truck dispatching plan has then
to be adjusted. Operating time and overhead costs may increase if the
RMC is not delivered according to schedule. In a worst case scenario,
interruption of delivery leads to the occurrence of cold-joint
problems. Preventive action is needed to ensure construction security.
Cold-joint problems can be handled in several ways given the nature
of the load bearing structure being poured at the placement site. If the
placement site is not in a major load bearing area, the surface of the
concrete can be dug away manually, allowing the placement to
continue after the application of adhesives to the roughed surface.
However, if the placement site is at a key location, to ensure quality,
the site manager must remove all concrete that has already been set.
The plant must then compensate by replacing all the RMC at the
placement site. Under these circumstances, both the plant and the
construction site managers can lose considerable amounts of capital
and resources owing to the delay in work and increased overhead

Corresponding author. Tel.: + 886 3 422 7151x34141; fax: + 886 3 425 2960.
E-mail address: t320002@cc.ncu.edu.tw (S. Yan).
0926-5805/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.autcon.2011.04.005

costs. Various solutions can be adopted to avoid the above scenario.

For example, if a plant has sufcient resources, additional machinery
or trucks could help supply the needed RMC. Alternatively, the rm
may have to purchase RMC from a backup contractor.
Unfortunately, most carriers in Taiwan and other countries still
handle schedule adjustments for RMC production and truck dispatching resulting from temporary mixer breakdowns manually based on
staff experience. The temporary schedule adjustment for RMC
production and truck dispatching is typically carried out using a
trial-and-error process. In particular, the planner adjusts the original
RMC production and truck dispatching schedule considering such
factors as the available number of mixers and production capacity,
available truck eet size, RMC demand at each construction site,
concrete cold joint problems, truck queuing, overtime and so on. The
process of local routing of trucks, RMC production of mixers, adjusting
of the truck dispatch plan within normal and overtime periods, and
supporting trucks between different sites/plants, is repeated by hand
until a satisfactory solution is found. The truck routes/schedules
and the RMC production schedule following incidents are generally modied, possibly with additional supporting truck trips and
overtime periods, depending on the incident characteristics and the
decision maker's adjustment.
This manual method is not efcient for handling temporary mixer
breakdowns, particularly for arranging a large network, and could
possibly result in an inferior feasible solution. To design an efcient
schedule adjustment for RMC production and truck dispatching, the
planner must systematically address both timeliness and exibility in
RMC production and truck dispatching, often needing to consider
normal work hours and overtime to satisfy the plant and construction
site operating constraints. However, owing to the many complex

1042

S. Yan et al. / Automation in Construction 20 (2011) 10411050

factors and constraints involved, systematic optimization approaches

to solving such an integrated problem following RMC mixer
breakdowns have rarely been developed.
Therefore, this study employs a network ow technique to develop
a systematic model that helps RMC carriers effectively adjust the RMC
production and truck dispatching schedules following RMC mixer
breakdowns. The model is formulated as a mixed integer network
ow problem with side constraints, which is characterized as NPhard. To efciently solve realistically large problems, a solution
method is developed based on a problem decomposition method,
coupled with the use of a mathematical programming solver, which
can be used to solve these problems. The validity of the proposed
model and solution algorithm in actual operations is demonstrated
through a case study based on actual operating data from an RMC rm
(called Firm Y) in Taiwan.
The rest of this paper is organized as follows: Section 2 presents a
review of related literature. Section 3 introduces the modeling
approach. Section 4 describes the solution method. The performance
of the model in a case study is discussed in Section 5. Conclusions are
nally drawn in Section 6, along with recommendations for future
research.
2. Literature Review
RMC production and delivery problems are increasingly being
studied by researchers. However, most investigations into RMC rm
operations have mainly concentrated on either the production
process or the dispatching process for regular operations. In relation
to the production process, for instance, Sawhney et al. [15]
emphasized the advanced features of the so-called Petri nets and
described their utilization as a process modeling and analysis tool for
the study of a ready mixed concrete plant. Wang et al. [20] presented a
simulation model based on data collected in Singapore. They included
the spreadsheet add-in software risk, to further analyze the effect of
the arrival pattern on the RMC placing productivity, and to determine
the best RMC trucks' inter-arrival time. Lu et al. [11] developed a
computer system for simulation modeling and analysis of Hong
Kong's RMC production operations.
In relation to the dispatching process, Zayed and Halpin [27]
simulated concrete batch operations by analyzing alternative solutions and resource management. Feng and Wu [3,4] developed a
model based on a genetic algorithm and a simulation method to solve
a near-optimal dispatching schedule for minimizing the total queuing
time of RMC trucks at a construction site and, simultaneously, satisfy
RMC delivery requirements at various construction sites. Feng and Wu
[5] developed a systematic model that incorporates the fast messy
genetic algorithms (fmGA) and the CYCLONE1 simulation method to
identify an optimal dispatching schedule. Graham et al. [6] presented
a neural network approach, comprised of a feed-forward network and
an Elman network, to the delivery problem. Naso et al. [13] proposed a
hybrid genetic algorithm combined with constructive heuristics for
RMC delivery problems. Schmid et al. [16,17] proposed two hybrid
solution procedures for dealing with RMC delivery problems. Park
et al. [14] developed a dynamic simulation model to represent the
generic RMC operation process but its structure and parameters
could be customized to different operation conditions. Lin et al. [12]
developed a multi-objective programming model for the dispatching
operation of RMC trucks as a job shop problem with recirculation.
Note that, in addition to the RMC truck dispatching problems, there
have been many other types of vehicle scheduling/routing/dispatching problems dealt with in various industrial studies. For example,
Erera et al. [2] combined enumeration with a greedy search strategy to
1
CYCLONE (CYCLic Operation NEtwork) developed by Halpin [9] has been widely
used for the design and analysis of construction operations over the last two decades
due to its simple symbols.

develop a solution method that could lead to signicant improvements in solution quality over simpler greedy procedures for largescale resource management problems. Zegordi et al. [28] proposed a
gendered genetic algorithm that considered two different chromosomes with non-equivalent structures for transportation scheduling
problems. However, the above problem characteristics are different
from ours so that these models and solution methods are not
applicable to our problem.
It should be mentioned that, to improve previous models where
the relationship between the production and dispatching processes
has been ignored, Yan and Lai [24] developed an optimization model
that integrates RMC production scheduling and truck dispatching into
the same framework to decide on the optimal RMC supply schedule
for both normal working hours and overtime. However, Yan and Lai's
[24] and the aforementioned approaches are used to deal with RMC
production and/or truck dispatching. They are not applicable to the
problem of schedule adjustment of RMC production and truck
dispatching following incidents, because of differences in the problem
context. There has not yet been any research in literature on optimal
schedule adjustments for the production and dispatching of RMC
following incidents. Therefore, in this study, the authors develop a
systematic model, coupled with a solution method that will help RMC
carriers effectively adjust their production and truck dispatching
schedules following incidents.
3. Modeling Approach
This section introduces the scheduling model. Major elements in
the model include the timespace network, operating constraints, and
mathematical formulation, as described below.
3.1. Timespace Network
A network consists of nodes and arcs. This representation is useful
for modeling a wide range of physical and conceptual situations [1].
Among the network approaches, the timespace network technique,
which has recently received considerable attention in the logistics
eld, is both a natural and exible means of modeling conveyance
routing/scheduling problems. Only a few nodes, links or additional
side constraints must be modied, without adjusting the original
network structure. It should be noted that timespace network ow
techniques have been successfully applied to elds other than
construction management, such as airline eet scheduling/routing,
ight timetable setting, bus routing/scheduling, crew scheduling,
work team scheduling, vessel scheduling/routing and supply chain
management (e.g., see [8,10,18,19,2123,25,26,29]). Yan and Lai [24]
recently used a timespace network technique to develop an
optimization model to help decide on effective RMC production and
truck dispatching schedules/routes. In this study, the timespace
technique is adopted to develop the model. This technique can easily
represent the operation of RMC production and truck dispatching due
to its natural representation of conveyance routing in the dimensions
of time and space. The timespace network developed to formulate
the model is described below (for a detailed description, please see
[18,26]).
As shown in Fig. 1, a node designates a plant at a specic time,
while an arc represents an activity between two different timespace
points, such as the holding of a truck, or the movement of a truck from
a plant to a construction site. The arc ows express the ow of trucks
in the networks. The horizontal axis in Fig. 1 denotes the RMC plant
and construction sites, while the vertical axis represents the time
duration from the starting time when the mixer rst breaks down
(which may add a preparation time), to the ending time of the RMC
plant when the period of overtime ends.
Here, overtime refers to the labor time after an eight hour work
day. The plan shown in Fig. 1 represents the timespace network with

S. Yan et al. / Automation in Construction 20 (2011) 10411050

3.1.1. Assignment Arc

An assignment arc, which connects two nonadjacent nodes
associated with a plant, represents the movement of a truck from
the plant to a construction site and back again at a later time. All
possible assignment arcs from every plant to any construction site
within a reasonable block of time (according to the average traveling
time, the placement time at the construction site, and plus a possible
buffer time added to resolve the delay in real traveling and
placement) are installed into the network. Note that no assignment
arc is allowed to emanate from every node associated with the
broken-down mixer during its broken-down period. The arc cost is
the truck operating cost. The arc ow's upper bound is one, meaning
that the assignment can be served at most once. The arc ow's lower
bound is zero, indicating that no truck serves this assignment.

respect to the recovery time that is before the overtime. However, the
incident time can last until after the overtime period, meaning the
recovery time could be after the overtime. The time gap depends on
the production of the mixer. Assume that each mixer produces 3 m3
concrete per minute and each truck takes, on average, 1.5 min to load
the maximum quantity of 4 m3. With an additional buffer time, a twominute period is taken as the time gap. Each node, except for the
starting node and the ending node, represents the RMC plant or a
construction site at a specic time. The starting node supply and the
ending node demand are the same, which is the number of RMC
trucks in the eet. Based on practical considerations, the network
analysis period is set at one day, including the remaining normal
operating time and an overtime period. Each arc represents an activity
for a truck. There are eight types of arcs, as can be seen below.

Plant 1

Plant 2

Plant a
(3)

(3)

1043

Starting node

(Breakdown)

Site 1

Site 2

Site m

(3)
Starting time
(6)

(4)

(6)

(1)

(4)

(6)

(2)
(2)
(6)
(5)

External
supply node

(5)

Recovery time
(6)
(1)(8)

(6)

Overtime
(7)
(7)

(7)
(7)
Ending time

f
Ending node

(1) Assignment arc

(2) Supporting assignment arc
(3) Initial supply arc
(4) Middle supply arc
(5) Holding arc

(6) Ferry arc

(7) Return arc
(8) External supply arc

Fig. 1. Fleet-ow network.

1044

S. Yan et al. / Automation in Construction 20 (2011) 10411050

3.1.2. Supporting Assignment Arc

A supporting assignment arc, which connects two nodes associated with two plants, represents the movement of a truck from a plant
to a construction site and then to another plant. A supporting
assignment arc is used so that a truck can assist another plant for later
assignments after leaving a plant and completing its own assignment
at a construction site. Thus, trucks can be efciently and exibly used
for different plants in cases when the system is perturbed by
incidents. Note that no supporting assignment arc is allowed to
emanate from any node associated with the broken-down mixer
during its break-down period. The arc cost, upper bound and lower
bound are designed to be the same as that of the assignment arc.
3.1.3. Initial Supply Arc
An initial supply arc, which connects the starting node to another
node associated with a plant at the starting time, denotes the
assigning of a truck into the system for service. The arc ow's upper
bound is the eet size, and the lower bound is 0. The arc cost is zero,
since the assignment of the eet into the system does not incur an
additional operating cost.
3.1.4. Middle Supply Arc
A middle supply arc, which connects the starting node to a node
associated with a plant after the starting time, denotes the assigning
of a truck into the system for service at a time later than the starting
time. Note that this truck has already been assigned to a construction
site before the starting time. It will return to the plant at a later time.
The arc cost is designed to be the same as that of the assignment arc.
To ensure that the truck arrives at the plant at the designated time,
both the upper bound and lower bound are set to be one.
3.1.5. Holding Arc
A holding arc, which connects two adjacent nodes associated with
a plant, represents the holding of trucks which have not completed
their deliveries at the plant in the associated time period. The arc cost
is the cost (e.g., the fuel cost and other related costs) incurred by
holding a truck at the plant for the associated time period. The arc
ow's upper bound is the eet size and the lower bound is 0.
3.1.6. Ferry Arc
A ferry arc, which connects two nodes associated with two plants
at two different times, represents the movement of an empty truck
from one plant to another plant after a time block (according to the
average travel time and plus a possible buffer time added to resolve
the delay in real traveling) for the purpose of assisting with later
assignments. Similar to support assignment arcs, ferry arcs are
designed so that trucks can be exibly relocated and thus more
efciently used in the system following incidents. The arc cost is the
operating cost for the ferry trip. The arc ow's upper bound is one and
the lower bound is 0.

3.2. Operating Constraints

In addition to the traditional node conservation constraints, the
following practical operating constraints must be considered in the
model [24]: 1) RMC production constraints: An RMC production plant's
capacity is determined by the time needed to measure, dispense, and
mix ingredients and to load the material into a truck. 2) Construction
site constraints: The demand (in truckloads) for each construction site
has to be supplied using the designed production and dispatch schedule.
Consequently, a side constraint should be added for each construction
site to ensure that the demand is satised. 3) Cold-joint-preventing
bundle constraints: to avoid the occurrence of the cold joint phenomenon at a construction site, the RMC placement process must be
implemented before the concrete hardens, meaning that the truck
should arrive within the construction site oat time. Such a time
window constraint should be considered in regular operations [24].
However, RMC production may not be sufcient to supply to all
construction sites if an RMC mixer is broken. As a result, there could be
cold joint occurring at some construction sites with a necessary cost to
resolve the cold joint problem. To reect this, the hard time window as
adopted by Yan and Lai [24] is revised to become a soft time window and
design a penalty cost to add to the system if a truck cannot arrive at the
construction site in time, resulting in the cold joint phenomenon.
3.3. Notation and Symbols Used in the Model
Before introducing the model formulation, the notation and
symbols used are listed in Tables 1 and 2.
3.4. Model Formulation
The model is formulated as a mixed integer network ow problem
with side constraints. The objective is to minimize the total system
operating cost, subject to the related operating constraints. Each
equation is briey described below:
Min
p
p
p
m
m
m
p p
Z = cij xij + + e + w
pPf

iN jN

+ v +

mMf

mMf nS

3.1.8. External Supply Arc

An external supply arc, which connects the external supply node to
a node associated with a construction site at a time, denotes the
assigning of a contractor's truck into the system for service at the
associated construction site for that time. The arc ow's upper bound
is one and the lower bound is 0. The arc cost is the price of purchasing
a truckload of RMC from a contractor.

m m
q yn

pPf

mMf nMC

m m
q n

Subject to:
8
< B; if i = v
xij xhi = 0; if i Nfv; f g
:
jN
hN
B; if i = f

xij
i;jAm ASm OSm

iN

i N Rp ; p Pf

xij 1

i;jTDi

3.1.7. Return Arc

A return arc, which connects a node associated with a plant at a
certain time to the ending node, denotes the completion of a truck's
service period, when it leaves the system. The arc ow's upper bound
is the eet size, while the lower bound is 0. The arc cost is zero, since
the assignment of the eet out of the system does not incur operating
cost.

mMf

m m



p
p
tif G 1  if

m Mf

if AR ; pPf

4
5

uif pif xif

if ARp ; pPf


p tvi + G 1  pvi

viAPp VMp ;pPc

uvi pvi xvi

viAPp VMp ;pPc



em tlm
ij G 1  xij

ijAm ASm OSm ;mMf

S. Yan et al. / Automation in Construction 20 (2011) 10411050

Table 1
List of parameters.

1045

Table 2
List of variables.

Symbol

Name/description

xij

Arc (i,j) ow

cij
p,m

p,m

Starting time of plant p(p Pc) and of construction site m (m Mc);

note that the starting times are constant for all plants and construction
sites that are operating before the system plan starts
Ending time of construction site mand plant p
Overtime for plant p
Overtime for plant m
Binary variable associated with return arc (i,f), which equals 1 if the
return arc (i,f) ow is positive and 0 otherwise
Binary variable associated with initial supply arc or middle supply arc
(v,i), which equals 1 if the initial supply arc or middle supply arc (v,i)
ow is positive and 0 otherwise
Binary variable associated with CPBm
n , which equals 1 if the delivery duty
of construction site m has not yet been nished and 0 otherwise
Binary variable associated with CPBm
n , which equals 1 if the delivery duty
of construction site m has been started and 0 otherwise
Binary variable associated with CPBm
n , which equals 1 if the nth coldjoint-preventing constraint associated with construction site m is
not satised and 0 otherwise
Binary variable associated with CPBm
n , which equals 1 if neither the nth nor
the (n + 1)thcold-joint-preventing constraint is satised and 0 otherwise

NRp
N

Arc (i,j) cost

Normal time operation cost (per unit time) of plantpand
construction site m
Overtime cost (per unit time) of plant p subtract p
Overtime cost (per unit time) of construction site m
subtract m
Network ending node and starting node
External supply node and RMC truck eet size
Demand for construction site m
Return arc (i,f)'s departure time associated with node i
Initial supply arc or middle supply arc (v,i)'s arrival time
associated with node i
A very large number (used for modeling)
Return arcxif's upper bound and initial supply arc or middle
supply arcxvi's upper bound
Ferry arc (i,j)'s departure time and arrival time associated
with construction site m
The normal ending time of plantp, equal to the starting
time of the overtime period
The normal ending time of construction site m, equal to
the starting time of the overtime period
First arc's departure time and arrival time in the nth coldjoint-preventing bundle associated with construction site m
The penalty cost for violating the cold-joint constraint at
the mth construction site
The nth cold-joint-preventing bundle associated with
construction site m
The nth queuing-preventing bundle associated with
construction site m
Set of nodes associated with plant p
Set of all nodes, N = NRp fv; f g

APp
AP

Set of initial supply arcs of plant p

Set of all initial supply arcs, AP = APp

VMp
VM

Set of middle supply arcs of plant p

Set of all middle supply arcs, VM = VMp

AH,E
AR

Set of all holding arcs and ferry arcs

Set of all return arcs, AR = ARp

AT
TDi

Set of all initial supply arcs, holding arcs and return arcs
Set of assignment arcs and supporting assignment arcs
emanating from node i
Set of assignment arcs associated with construction site m
Set of plants that have not ended operations before the
system plan starts
Set of all assignment arcs, A = Am
m
Set of supporting assignment arcs associated with
construction site m
Set of all supporting assignment arcs, AS = ASm
m
Set of external supply arcs associated with construction
site m
Set of all external supply arcs, OS = OSm

p
m
f,v
o,B
dm
tif
tvi
G
uif,uvi
m
tlm
ij ,taij

tfp
tfm
m
tlm
n ,tan

CPBm
n
QPBm
n

OS

Mc
Sm,Qm

MCm
I

ijAm ASm OSm ; mMc

10

wp p tf p

pPf

11

mMf

12


m
G m
n + n 1

m
m
xij + ym
n n + n 1

i;jCPBm
n

Gn + tln e

m
m
Gm
n + tan
m
m
m
m
n + 1 yn + yn + 1 2n

xij
i;jQPBm
n
p

nSm ; mMf 13

nS ; mMf

14

nSm ; mMf

15

nMC m ; mMf
nQ m ; mMf

1
m

16
17

mMf ; pPf

18

; 0

mMc ; pPc

19

xij f0; 1g

ijAASOSEVM

20

pif ; pvi f0; 1g



m tam
ij + G 1  xij

v e tf

m
n

m m m
m
n ; n ; yn ; n f0; 1g

Set of plants that have not operated before the system

plan starts, Pc Pf
Set of the construction sites that have not ended operations
before the system plan starts
Set of the construction sites that have not operated before
the system plan starts, Mc Mf
Set of cold-joint-preventing bundle and queuingpreventing
bundle constraints associated with construction site m, each
including a number of assignment arcs, supporting assignment
arcs, and external supply arcs
Set of all bundle constraint pairs, each corresponding to two
continuous cold-joint constraints, for construction site m
Set of integers

Mf

ym
n

xij 0; xij I

Pc

m
n

AS
OSm

m
n

A
ASm

pvi

w ; v ; ; e 0

Am
Pf

em, p
wp
vm
pif

ijAT
nSm ; mMf

21
22

if AR;viAPp VMp ; pPf 23

Eq. (1) denotes minimization of the total system cost, including

the truck operating cost and external supply cost (the rst term), the
normal operating costs, the overtime cost at the plants (the second
and the fourth terms), the normal operating costs, the overtime cost at
the construction sites (the third and the fth terms), the penalty costs
for contravening the cold joint constraint (the sixth term), and
superuous penalty costs deducted for contravening the cold joint
constraint (the last term). Eq. (2) ensures ow conservation at every
node in the network. Eq. (3) indicates the plant production constraint;
the number of trucks which are permitted to depart from each plant
(in normal operations) at each node is less than or equal to one. Note
that because no assignment/supporting assignment arc is allowed to
emanate from those nodes associated with the broken-down mixer
during its break-down period (meaning the mixer does not operate
during this period), this constraint does not apply to these nodes.
Eq. (4) ensures that the sum of all delivery arc ows associated with
each construction site must satisfy its demand. Eqs. (5) and (6)
determine each plant ending time, which is the time associated with
the last truck leaving each plant. Note that the binary variable pif is
designed to indicate whether there are trucks leaving plant p at the
time associated with node i. That is, once the return arc ow is positive
(i.e.,xif N 0), the variable pif will be 1; the ending time of plant p can
then be found.

1046

S. Yan et al. / Automation in Construction 20 (2011) 10411050

Eqs. (7) and (8) determine the starting time of plant p, which is
equal to the time associated with the rst truck entering plant p. The
binary variable pvi is designed to indicate whether there are trucks
entering plant p at the time associated with node i. Once the
assignment arc ow is positive (i.e.,xsi N 0), variable pvi will be 1 and
the starting time of plant p can then be found. Eq. (9) determines the
ending time of each construction site, which is equal to the departure
time of the last truck from each construction site. Eq. (10) determines
the starting time of each construction site, which is equal to the arrival
time of the rst truck at the construction site. Eqs. (11) and (12)
respectively determine the overtime period associated with every
plant and every construction site.
Eq. (13) determines whether there is a positive arc ow in a coldjoint-preventing bundle set, i.e., if a cold-joint-preventing bundle ow
(the sum of the bundle arc ows) is equal to 1. If not, then ym
n is equal
to one, meaning that the penalty for handling the cold-joint problem
in reality is added into the objective. Eq. (14) denotes that there is no
positive arc ow entering a construction site (i.e., no delivery) after its
demand has been satised. On the other hand, there will be a positive
arc ow entering the construction site (a delivery) if its demand has
not yet been satised. Eq. (15) ensures that there is a positive arc ow
entering the construction site once the construction site starts to
operate. Eq. (16) prevents the need to recalculate the cold-joint
penalty cost. If two consecutive cold-joint-preventing bundle constraints are violated, then only one penalty is needed. In this case, m
n
will be equal to one, meaning that one additional penalty needs to be
subtracted to prevent the incurred penalty being double counted.
Eq. (17) prevents the queuing of trucks at every construction site
(i.e., the sum of the arc ows in each queuing-preventing bundle set
for every construction site is less than or equal to 1). Eqs. (18) and
(19) indicate that the time variables are nonnegative. Eq. (20) denotes
that the assignment arc, supporting assignment arc, middle supply
arc, ferry arc and external supply arc ows are binary. Eq. (21)
indicates that the holding arc, initial supply arc and return arc ows
are all nonnegative integers. Eq. (22) ensures that the cold-jointpreventing bundle variables are binary. Eq. (23) ensures that the
variables, pif and pvi are binary.
4. Solution Method
The model is formulated as a mixed integer network ow problem
with side constraints, which is characterized as NP-hard in terms of
optimization [7]. The primary elements related with the problem
complexity mainly include the number of plants, the number of
construction sites and the system planning time. The RMC demand,
the eet size and other parameters are not correlated with the
problem size. The authors adopted the mathematical programming
solver ILOG CPLEX (denoted as CPLEX hereafter), applying a branchand-bound technique coupled with the simplex method to rst solve
the problem. In preliminary tests, CPLEX could not solve a realistically
large-scale problem, with two plants and twenty construction sites,
within one day. In addition, as discussed with experienced staff from
Firm Y manual approaches would require a long time to solve for only
a feasible solution for an even smaller problem, with one plant, two
construction sites and a one-day planning time. This shows that the
problem is so complicated that it is hard to solve. An effective model
with an efcient solution method that would allow carriers to
effectively adjust their RMC production and truck dispatching
schedules following incidents would be very useful. Therefore, the
authors develop an efcient heuristic algorithm which integrates a
problem decomposition and relaxation technique to solve the
problem. The solution procedure is discussed below. By cutting the
time points to divide the planning time and relaxing some of the
constraints, the authors build a series of sub-problems. The rst subproblem, the smallest one, is solved using CLPEX. The second subproblem is created by incorporating an additional period into the rst

sub-problem, where the previously solved assignment arc and

supporting assignment arc values are xed. Then, CPLEX is used to
solve the second sub-problem. These steps are repeated until all daily
demands are satised. Finally, after xing the initial solution of the
last step, and recovering the constraints that have been relaxed,
CPLEX is used to solve the entire problem. The steps are described in
detail below.
Step 1: Modify the original problem to create a sub-problem. The
external supply arcs are not included in the timespace network.
When the original plants could not supply what is needed, an
articial supply arc is added. This step is detailed as follows:
1.1 Cut the original problem with a portion of the planning time to
create a sub-problem. Note that starting from the second subproblem, a new portion of timespace network is added to the
previous sub-problem (with xed assignment and supporting
assignment arc values previously solved) to enlarge the subproblem. Rewrite the objective function as:

m m

Z = cij xij + q yn
mMf nSm

iN jN

+G

m m

q n

mMf nMC m

24

mMf

1.2

Modied Eq. (4)

x

ij

i;jAm ASm

+ d mMf

25

where m represents insufcient demand for the mth construction

site that is supplied by the articial plant.
1.3 To comply with the design for the new objective function (24),
some time-related side constraints are removed, including Eqs.
(5), (6), (7), (8), (11), and (12). Note that to ensure that no RMC
supply is needed after the ending time for each construction site,
and to ensure that RMC can only be supplied after the starting
time for each construction site, constraints (9), (10), (13) and
(14) are retained. To improve the computation efciency, after
the solving of the rst sub-problem add only the last cold-joint
and queuing-prevention bundle constraints (i.e., constraints
(13) to (17)) with a positive arc ow, and the related constraints
of the additional part of the timespace network for the second
and later sub-problems. In addition, if the required demand is
satised for a construction site, then its related cold-joint and
queuing-prevention constraints can be removed next time.
Step 2: Solve the sub-problem using CPLEX.
Step 3: Check if the daily demand is satised for all construction
sites or if the entire timespace network has been added. If yes,
then go to Step 4; else x the previously solved values of the
assignment arcs and supporting assignment arcs, add a new
portion of timespace network (including some nodes and arcs)
and go back to Step 1.
Step 4: Use CPLEX to solve the original problem (including the original
objective function (1), all the original constraints, all the original arcs)
with the xed arc values from the previous steps. Note that the heuristic
nds a feasible solution, because the demand for each construction site
can be best satised by using an external supply at this step.
5. Case Study
The applicability of the proposed model and solution method in
the real world was demonstrated by a case study based on operating
data from Firm Y. The Visual C++ computer language, coupled with

S. Yan et al. / Automation in Construction 20 (2011) 10411050

Table 3
Test results.

1047

Table 5
Internal solution results for the large-scale problem.

Problem scale
# variables
# constraints
Solution method
Computation time
OBJ (NT$)
GAP (%)

Small-scale
2241
4374
CPLEX
12 h
43,552.90
3.99

Problem scale
# variables
# constraints
Solution method
Computation time
OBJ (NT$)
GAP (%)

Large-scale
144,436
268,863
CPLEX

Internal results

Heuristic algorithm
10 s
45,293.20

Heuristic algorithm
331 s
696,382.20

the mathematical programming solver, CPLEX 10.0, was used to solve

the problems and to perform the sensitivity/scenario analyses. The
tests were performed on a Pentium 41.8 GHz with 1 GB of RAM in
the environment of Microsoft Windows XP.

5.1. Model Tests and Results

Firm Y has four RMC plants located in Hsinchu City in northwestern
Taiwan. Actual data are analyzed based on Firm Y's daily production and
truck delivery statements. There is only one RMC mixer at each of the
four plants. They are no farther than 10 km apart and can help each
other. Twenty construction sites, with different demands (including
concrete type and quantity), were served by the four plants, with a eet
of 128 RMC trucks. The system planning time was from 8:00AM to
24:00PM. Moreover, it is assumed that a plant mixer breaks down at
10:00 AM and could be repaired by 12:00 PM. The travel times
(including a buffer time), from each plant to each construction site and
the placement time for each construction site are set referring to real
operating data provided by Firm Y. In addition, the buffer time for each
truck trip was set to be 20% of the travel time, referring to Firm Y, making
the setting of truck travel times conservative in operations. Additionally,
the cold-joint time at all construction sites is set to be an average of
10 min.
The operating costs at a plant or a construction site include those for
labor and the operation of machinery. In this study, only costs related to
production, delivery and placement are used as model inputs in short
term operations. The costs involving RMC production/placement at a
plant/construction site and RMC truck operation (loaded or empty)
following incidents, derived from the actual experience of Firm Y,
include electrical machinery operating costs, labor costs, and depreciation costs of mixer machinery. In addition to the aforementioned costs
and input data, the input also includes data on ordinary dispatch
planning, such as the number of trucks in a plant and the number of
returning trucks. These form the initial supplies and middle supplies.
During the test, CPLEX, with a convergence gap of 3%, was used to
directly solve the problem with the operating data. The results are
summarized in Table 3. The small-scale case (with two plants, four
construction sites, and a system planning time from 8:00 AM to 24:00
PM) solved using CPLEX still requires around 12 h, which is too long to
be applicable for the request for an immediate dispatch, indicating that
directly solving the problem using CPLEX is not efcient and could be
difcult to apply in practice. Note that this also shows that the problem,

Total services
# of truck trip assignments
# of supporting truck trip assignments
# of ferry truck trips
Average plant overtime (min.)
# external supply truck trips
Cold-joint penalty value
Average truck return time (min)
Average plant operating time (min)
Average construction site operating time (min)

438
163
275
187
181.00
0
0.00
330.75
661.00
195.60

which is characterized as NP-hard, is very difcult to optimize using

modern computer techniques, even for a small-scale problem. However,
the proposed heuristic algorithm requires only 10 s to solve the smallscale problem with an error gap of 3.99%, which is acceptable in the real
world. Even for the large-scale case, the algorithm takes only 331 s to
solve the problem, while CPLEX cannot even nd a feasible solution in
24 h, suggesting that the proposed heuristic algorithm could be
applicable for solving realistically large problems.
Furthermore, the performance of the proposed model and
heuristic algorithm is also compared with the performance of
experience based manual schedule adjustment. As shown in Table 4,
the heuristic algorithm out-performs the manual method by 10.75%
(=(771,267696,382.20)/696,382.20), demonstrating that the proposed model and algorithm are signicantly better than the manual
method in practice.
As shown in Table 5, the RMC rm did not purchase any RMC from
contract-rms and no cold-joint problem occurred to incur penalty
cost, indicating that the proposed model and heuristic algorithm
could be used to effectively adjust resources (trucks and plants) to
supply RMC to match demands for all construction sites. The results
also showed that the average plant operating time exceeded the
normal operating time by about 181 min, which is equal to the
average plant overtime. No overtime was found for any construction
site. In addition, there were 163 truck assignments and 275
supporting truck assignments, indicating that the trucks were
efciently and exibly used among different plants when the system
was perturbed by an incident. Moreover, there were 187 ferry truck
trips, implying that trucks were systemically and effectively adjusted
among the plants and construction sites, so as to minimize the total
system cost following the incident.

5.2. Sensitivity/Scenario Analyses

To understand the inuence of the model parameters on the
solution, the authors performed sensitivity/scenario analyses on the
available RMC demand, available eet size, required repair time and
incident starting time which are essential inputs to the model. Other
types of sensitivity/scenario analyses can be similarly performed but are

Table 4
Heuristic algorithm results compared with manual results.
Solution method

Manual method

Heuristic algorithm

GAP (%)

OBJ (NT$)

771,267.00

696,382.20

10.75

Fig. 2. Objective values and computation times for different RMC demands.

1048

S. Yan et al. / Automation in Construction 20 (2011) 10411050

Fig. 5. Average construction site operating times for different RMC demands.
Fig. 3. Average plant operating times for different RMC demands.

not addressed here to save space. The data set mentioned in Section 5.1
and the proposed solution method was used for the analysis.
5.2.1. RMC Demand
To examine the inuence of the RMC demand at construction sites
on the results, this study changes the RMC demand, ranging from 70%,
80%, 90%, 100%, 110%, 120%, 130% and 140% of the original demand. As
shown in Fig. 2, the RMC demand positively affects the objective. In
particular, when the RMC demand decreases to 70%, the objective
value is reduced from 696,382.2 to 498,304.85, a decrement of 39.8%
(=(696,382.2498,304.85)/498,304.85). When the RMC demand
increases to 140%, the objective value is increased from 696,382.2 to
985,374.3 (=(985,374.3696,382.2)/696,382.2), an increase of 41.5%.
Additionally, the solution time increases with rising demand.
However, even when the RMC demand increases to 140% the model
is able to solve the problem in 764 s, indicating that the proposed
model and solution algorithm could be efcient for resolving large
RMC demands that arise in the real world.
Moreover, when the RMC demand is less than or equal to 130%, the
plants can adequately supply RMC to construction sites. However,
when the RMC demand reaches 140%, it is necessary to input RMC
from an outside contractor, because the placement amount exceeds
the maximum for a day's work at some construction sites. It was also
found, as shown in Figs. 3 and 4, that the average plant operating time
and the average plant overtime basically increase as RMC demand
increases. Note that the average plant operating time and the average
plant overtime for a demand of 90% are slightly less than those for a
demand of 80%, mainly due to the convergence error of the solution
algorithm that often occurs for heuristics.
Similarly, as shown in Fig. 5, as the RMC demand increases, the
average construction site operating time increases, meaning that
increasing RMC demand will increase average operating time at
construction sites. Additionally, as shown in Fig. 6, the average truck
return time (i.e., average truck service time including travel time,
pouring time and waiting time) increases as the RMC demand
increases, indicating that increasing RMC demand will lengthen the
average truck service time. Finally, as shown in Fig. 7, the number of
ferry truck trips increases as the RMC demand increases, indicating
trucks should be more exibly relocated to plants/construction sites
so as to systematically and effectively supply RMC from plants to
construction sites to meet the demand.

Fig. 4. Average plant overtimes for different RMC demands.

5.2.2. Available Fleet Size

This study examines the sensitivity of the truck eet size, ranging
from 80, 96, 112, 128, 144 and 160 on the results. The computation
times are similar for different sizes. As shown in Fig. 8, as the eet size
increases from 80 to 112, the objective value decreases by 4.13% (=
(720,331.3690,592.5)/720,331.3). The objective value changes
slightly when the eet size is greater than 112. Note that when the
eet sizes increase from 144 to 160, the objective value increases
slightly, mainly due to the convergence error from the solution
algorithm. In total, the objective value shows a tendency to decrease
as the eet size increases, indicating that the more the RMC trucks, the
more efcient the transportation of RMC from the plants to the
construction sites and, consequently, the lower the operating cost.
In addition, as shown in Fig. 9, the average truck return time
decreases with increasing eet size. Moreover, increasing the eet
size from 80 to 128 (the planned size), decreases the average truck
return time from 458.85 to 330.75 min, a decrement of 27.2% (=
(458.85330.75)/458.85). Increasing the eet size from 128 to 160
decreases the average truck return time from 330.75 to 288.48 min, a
decrement of 12.78% (=(330.75288.48)/330.75). Briey, the average plant operating time decreases slightly from 696 to 650 min, but
there is not a signicant change because the demand remains
constant. The average construction site operating times are similar
for different eet sizes, because the demand required at each
construction site remains the same. Finally, the number of ferry
trips decreases slightly from 200 to 190, meaning that the more the
trucks, the fewer the number of ferry trips needed to adjust the eet
schedule/route to satisfy the demand of all construction sites.
5.2.3. Required Repair Time
During the repair time for an RMC mixer following an incident the
plant cannot produce RMC. A longer repair period would have a more
signicant inuence on RMC production and truck dispatching plans.
Overhead costs may arise because of reduced production. Therefore,
taking the same test case in Section 5.1, with the other data remaining
unchanged, four repair times are designed to examine the inuence of
required repair time on the results as follows:
Scenario A: no time required for repair (meaning no incident occurs)
Scenario B: one hour required for repair
Scenario C: two hours required for repair
Scenario D: three hours required for repair

Fig. 6. Average truck return times for different RMC demands.

S. Yan et al. / Automation in Construction 20 (2011) 10411050

Fig. 9. Average truck return times for different eet sizes.

Fig. 7. # ferry truck trips for different RMC demands.

As shown in Table 6, when the required repair time increases, the

objective value increases, implying that a lower production of RMC will
result in higher operating cost to produce and transport RMC from plants
to construction sites. In addition, the objective value increases as the
required repair time increases. The objective difference between scenarios
A and D (a three hour difference) is 3.84%. The maximum difference
between two scenarios with a one-hour difference is 1.71% (i.e., for
scenarios C and D). These show that the required repair time length does
affect the system cost. Moreover, as the required repair time increases, the
average plant operating time and the average plant overtime signicantly
increase. This is because when a plant is incapable of producing RMC for an
extended period, more RMC must be supplied by other normally
functioning plants. Note that the average construction site operating
time and the average truck return time do not change signicantly,
implying the breakdown of a mixer does not affect them very much. Note
that the computation times are similar for different scenarios.
It should be mentioned that to apply the model in practice, the
repair time, which has to be estimated in advance, is a model input. In
real operations, if the estimated repair time is not correct at the
beginning and has to be revised later, then the repair time can be
updated and the model with the updated problem parameters (i.e.,
the updated allocation of trucks, the remaining RMC demand, etc.) can
be rerun to obtain a new schedule.
5.2.4. Incident Starting Time
To examine the inuence of the incident starting time on the
results, a scenario analysis is performed with three incident starting
times. The same test case as discussed in Section 5.1 was used while
the other data remained unchanged. The three incident starting times
with the repair time of 2 h are designed as follows:
Scenario A: starting from 09:00 (AM.)
Scenario B: starting from 10:00 (AM.)
Scenario C: starting from 11:00 (AM.)
As shown in Table 7, the objective value decreases with increasing
starting time, implying that the later the incident occurs, the less the
system is affected, and consequently, the lower the operating cost
needed to produce and to transport RMC from plants to construction
sites. In addition, there is a signicant objective difference between
two starting times, e.g., 11.58% (=(787,625.4696,382.2)/787,625.4)
between scenarios A and B (a one hour difference), showing that the
starting time does have a signicant effect on the objective value.

Moreover, as the starting time increases, the average plant operating

time and the average plant overtime decrease. This is because the later
an incident occurs (the less the system is perturbed), the more the
RMC that has already been supplied, and therefore the shorter the
average plant operating time (including overtime) needed at each
plant. Note that the average truck return time increases as the starting
time increases, because the later the system is perturbed, the less
efciently the eet can be adjusted, and consequently the longer the
average service time needed for each truck. Finally, the number of
ferry trips decreases from 380 to 113, showing that the less the
remaining RMC demand, the fewer the number of ferry trips that need
to be used to adjust the eet schedule/route in terms of system
operation. Note that the computation times are similar for different
starting times.
6. Conclusions
Effectively and efciently delivering Ready Mixed Concrete (RMC)
to construction sites when there are temporary breakdowns of
machinery, is an important issue to an RMC batch plant manager.
Effective and efcient adjusting of schedules for the production and
dispatching of ready mixed concrete (RMC) following incidents is an
important issue to an RMC batch plant manager, which is however not
an easy task in practice. Currently, this adjustment is manually
determined by staff based on experience, without optimization from a
systemic perspective. There has not yet been any research in literature
on optimal schedule adjustments for RMC production and truck
dispatching following incidents. This study develops a timespace
network model, which takes practical operating constraints and
objective function into account to efciently and effectively adjust the
RMC production and truck dispatching schedules following RMC
mixer breakdowns. The model is formulated as a mixed integer
network ow problem with side constraints, that is characterized as
NP-hard in terms of optimization. A solution algorithm, incorporating
a problem decomposition technique and the use of a mathematical
programming solver, is developed to efciently solve the problem.
The model and the solution algorithm are evaluated using a case
study, based on real operating data from an RMC rm in Taiwan. In
preliminary tests, it is found that CPLEX cannot be directly used to
approach a near optimal solution within one day. However, the

Table 6
Results for different required repair times.
Scenario
OBJ (NT$)
Difference of OBJ (%)
Average plant operating
time (min)
Average construction
site operating time (min)
Average truck return time (min)
# ferry truck trips
Average plant overtime (min)

Fig. 8. Objective values for different eet sizes.

1049

Note: c = (b a) / a.

Internal result
681,731.1a 690,078.5b 696,382.2 708,274.0
1.22c
0.91
1.71
637.0
646.5
656.5
688.5
191.2

194.2

196.8

196.2

330.06
158
157

336.45
204
226.5

331.56
210
296.5

332.69
194
388.5

1050

S. Yan et al. / Automation in Construction 20 (2011) 10411050

References

Table 7
Results for different incident starting times.
Scenario
OBJ (NT$)
Difference of OBJ (%)
Average plant operating time (min.)
Average construction site operating
time (min.)
Average truck return time (min.)
# ferry truck trips
Average plant overtime (min.)

Internal result
787,625.4a
11.58%c
759.00
219.00

696,382.2b
9.09%
656.50
196.80

633,063.6

308.42
380
279

331.56
210
176.5

358.64
113
161

641.00
163.80

Note: c = (a b)/a.

proposed model and solution algorithm can both effectively and

efciently solve realistically large problems, indicating that the
proposed model and the solution algorithm could be helpful planning
tools for determining appropriate RMC production and truck
dispatching schedules in practice. Sensitivity/scenario analyses are
also performed in order to understand how the essential parameters
affect the solution. Finally, the development of the model, the solution
algorithm, the case study, and the sensitivity/scenario analyses could
all be useful as reference material for RMC carriers. In summary, this
study proposes a useful framework, including a basic system
perturbation model and a solution method, that can serve as the
basis for future modication/improvement of the model and/or the
solution method from which to develop more suitable ones for other
carriers in other applications.
Although the test results show that the proposed model and the
solution method are potentially useful for RMC production and truck
dispatching following RMC mixer breakdowns, especially for Firm Y,
more testing and case studies should be conducted, so that carriers
may better grasp the effectiveness of the model and the solution
method, and their limitations. In addition, the proposed solution
method works well for realistically large problems with up to 144,436
variables and 268,863 constraints in size. The solution method may
not be as efcient for larger size problems in practical applications.
Thus, the proposed solution method may be further suitably modied
or new solution methods may be developed by employing, for
example, Lagrangian relaxation, column generation, or modern metaheuristic techniques (e.g., the tabu search method, threshold accepting method or genetic algorithms) to efciently solve larger-scale
problems. Finally, the extension of the model to the operations of
multiple truck types that are sometimes used in the real world could
be a direction of future research.
Acknowledgements
The authors would like to thank the National Science Council of
Taiwan for their nancial support of this research under grant NSC98-2221-E-008-077-MY3. The authors would also like to thank the
ready mixed concrete rm for kindly providing the test data and their
valuable opinions. Finally, the authors would like to thank the three
anonymous reviewers for their helpful comments and suggestions
that have helped improve the presentation of the paper.

[1] R.K. Ahuja, T.L. Magnanti, J.B. Orlin, Network ows, theory, algorithms, and
applications, Prentice Hall, Englewood Cliffs, 1993.
[2] A. Erera, B. Karack, M. Savelsbergh, A dynamic driver management scheme for
less-than-truckload carriers, Computers and Operations Research 35 (11) (2008)
33973411.
[3] C.W. Feng, T.M. Cheng, H.T. Wu, Optimizing the schedule of dispatching RMC
trucks through genetic algorithms, Automation in Construction 13 (3) (2004)
327340.
[4] C.W. Feng, H.T. Wu, Using genetic algorithms to optimize the dispatching
schedule of RMC trucks, Proceedings of the 17th International Symposium on
Automation and Robotics in Construction, Taipei, Taiwan, 2000, pp. 927932.
[5] C.W. Feng, H.T. Wu, Integrating fmGA and CYCLONE to optimize the schedule of
dispatching RMC trucks, Automation in Construction 15 (2) (2006) 186199.
[6] L.D. Graham, D.R. Forbes, S.D. Smith, Modeling the ready mixed concrete delivery
system with neural networks, Automation in Construction 15 (5) (2006)
656663.
[7] M.R. Garey, D.S. Johnson, Computers and intractability: a guide to the theory of
NP-completeness, W. H. Freeman & Company, San Francisco, 1979.
[8] I. Steinzen, V. Gintner, L. Suhl, K. Natalia, A timespace network based approach
for integrated vehicle and crew scheduling in public transport, Transportation
Science 44 (3) (2010) 371377.
[9] D.W. Halpin, CYCLONE method for modeling job site processes, Journal of
Construction Division, ASCE CO3 (1977) 489499.
[10] N. Kliewer, T. Mellouli, L. Suhl, A timespace network based exact optimization
model for multi-depot bus scheduling, European Journal of Operational Research
175 (3) (2006) 16161627.
[11] M. Lu, M. Anson, S.L. Tang, Y.C. Ying, HKCONSIM: a practical simulation solution to
planning concrete plant operations in Hong Kong, Journal of Construction
Engineering and Management 129 (5) (2003) 547554.
[12] P.C. Lin, J. Wang, S.H. Huang, Y.T. Wang, Dispatching ready mixed concrete trucks
under demand postponement and weight limit regulation, Automation in
Construction 19 (6) (2010) 798807.
[13] D. Naso, M. Surico, B. Turchiano, U. Kaymak, Genetic algorithms for supply-chain
scheduling: a case study in the distribution of ready-mixed concrete, European
Journal of Operational Research 177 (3) (2007) 20692099.
[14] M. Park, W.Y. Kim, H.S. Lee, S. Han, Supply chain management model for ready
mixed concrete, Automation in Construction 20 (1) (2010) 4455.
[15] A. Sawhney, O. Abudayyeh, T. Chaitavatputtiporn, Modeling and analysis of
concrete production plant using petri nets, Journal of Computing in Civil
Engineering, ASCE 13 (3) (1999) 178186.
[16] V. Schmid, K.F. Doerner, R.F. Hartl, G.J.J. Salazar, Hybridization of very large
neighborhood search for ready-mixed concrete delivery problems, Computers
and Operations Research 37 (3) (2010) 559574.
[17] V. Schmid, K.F. Doerner, R.F. Hartl, M.W.P. Savelsbergh, W. Stoecher, A hybrid
solution approach for ready-mixed concrete delivery, Transportation Science 43
(1) (2009) 7085.
[18] R.W. Simpson, Scheduling and Routing Models for Airline Systems, Flight Transportation Laboratory, Report R-68-3, Massachusetts Institute of Technology, USA, 1969.
[19] D.Z.W. Wang, H.K. Lo, Multi-eet ferry service network design with passenger
preferences for differential services, Transportation Research 42B (9) (2008) 798822.
[20] S.Q. Wang, C.L. Teo, G. Ofori, Scheduling the truckmixer arrival for a ready mixed
concrete pour via simulation with risk, Journal of Construction Research 2 (2)
(2001) 169179.
[21] S. Yan, J.C. Chang, Airline Cockpit Crew Scheduling, European Journal of
Operational Research 136 (3) (2002) 501511.
[22] S. Yan, C.H. Chen, Coordinated ight scheduling models for allied airlines,
Transportation Research Part C 15 (4) (2007) 246264.
[23] S. Yan, C.J. Chi, C.H. Tang, Inter-city bus routing and timetable setting under
stochastic demands, Transportation Research 40A (7) (2006) 572586.
[24] S. Yan, W.S. Lai, An optimal scheduling model for ready mixed concrete supply with
overtime considerations, Automation in Construction 16 (6) (2007) 734744.
[25] S. Yan, Y.L. Shih, A timespace network model for work team scheduling after a
major disaster, Journal of Chinese Institute of Engineers 30 (1) (2007) 6375.
[26] S. Yan, H.F. Young, A decision support framework for multi-eet routing and
multi-stop ight scheduling, Transportation Research 30A (5) (1996) 379398.
[27] T.M. Zayed, D.W. Halpin, Simulation of concrete batch plant production, Journal of
Construction Engineering and Management 127 (2) (2001) 132141.
[28] S.H. Zegordi, I.N. Kamal Abadi, M.A. Beheshti Nia, A novel genetic algorithm for
solving production and transportation scheduling in a two-stage supply chain,
Computers and Industrial Engineering 58 (3) (2010) 373381.
[29] T. Zhang, L. Li, The dynamic pricing model with timespace network based on
uncertain demands, Control and Decision Conference (2-4) (2008) 19561959.