Application of queuing theory in construction

industry
Vjacheslav Usmanov 1*, Čeněk Jarský 1

1
Department of Construction Technology, FCE, CTU Prague, Czech Republic
* Corresponding author (usmanov@seznam.cz)

Purpose Each production process in construction is closely connected with the question of costs and deadlines. In
every project an investor or customer, as well as the construction company, has to meet the planned completion date
and the estimated costs associated with the construction. In practice, determining the duration of construction at mini-
mum costs is still not based on the reliable calculation, and in the planning of costs, the connection between terms and
financial costs is rarely taken into account. Method The queues theory examines systems with operating channels,
where the process of queues formation takes place and subsequent servicing of the customers by servicing centers. The
main objective of the queues theory is to determine the laws under which the system works, and further to create the
most accurate mathematical model that takes into account various stochastic influences on the process. The entire con-
struction process can be examined from the point of view of a customer who is waiting in the queue and is interested
primarily in the waiting time, as well as from the point of view of servicing centers. A waiting element decides if you join
the queue, or to go to another system entirely. In terms of servicing centers, the priority is to determine the occupancy of
the channel and the probability of failure, including the time of repair. A servicing center should also reliably identify the
time per customer service, taking into account the current construction task. Results & Discussion The present study
demonstrates that it is possible to simulate the complex process of construction, containing hundreds of individual con-
struction processes, mathematically and technically, with a number of simplifications, and then perform various calcula-
tions and changes for effective and long-term planning of construction. The mathematical simulation should show that
some variants of machines combinations fail to accomplish the task under the given conditions, some will not be optimal
in terms of costs or other parameters, other variants will be optimal in the view of costs required to fulfill the construction
task. The simulation software allows a look at the results in graphical form or to export data to other programs. Applica-
tion of the queues theory allows the introduction into the system waiting time the servicing elements and to approximate
the mathematical model to a real working tasks on site.

Keywords: queues theory, optimal choice, machines combination, construction, simulation

INTRODUCTION the optimal choice of machinery is still in its early

Each production process in construction is closely stages of development, and a software, which would
connected with the question of costs and deadlines. allow an easy optimal choice of machinery in terms
An investor or customer, as well as construction of minimizing labour content and costs, fuel con-
companies themselves in every project have to meet sumption, time of construction, environmental im-
the planned completion date and the estimated costs pact, etc., including relevant data, does not exist on
associated with the construction. In practice, deter- the internet basis.
mining the duration of construction at minimum cost
is still not founded on a reliable calculation, and in THE QUEUING THEORY
the planning of costs, the connection between terms The current practice does not allow construction
and financial costs is rarely taken into account. companies to perform detailed time-consuming cal-
At present, some programs are available for the culations of optimization during formulation of their
analysis of building production and for scheduling offers. An offer is usually focused on the contract
with the help of network graphs, timetables and their price, which has to match the situation in the con-
optimization. A further supporting facility to optimize struction market. Optimization steps are therefore
the prices is so called internal corporate guidelines; made only after the contract is signed. This study is
construction companies produce them depending on devoted to the creation of a technical and mathemat-
their own experience and statistical values. Some of ic model and to searching the methods leading to the
these documents are then used to optimize the use optimization of construction processes (minimization
of machinery and to create building plans. Other of labor and costs, fuel consumption, time of con-
similar information instruments, especially for the struction, environmental impact, etc.) by means of
optimization of construction processes, do not yet special simulation software.
exist, or they are produced only in a small and evi- The queuing theory examines systems with operat-
dently insufficient volume. In particular, the sphere of ing channels, where the process of queues formation
takes place and subsequently the servicing of the
customers by servicing centers. The main objective
of the queuing theory is to determine the regularities
under which the system works, and further to create
the most accurate mathematical model that takes
into account various stochastic influences on the
process. The entire construction process can be
examined from the point of view of a customer who Fig. 2. Basic structure of the queuing system
is waiting in the queue and is interested primarily in
the waiting time, as well as from the point of view of
A closed queuing system is shown in Figure 3.
servicing centers. A waiting element decides in what
queue to be included or whether to go to another
system entirely. In terms of servicing centers, the
priority is to determine the occupancy of the channel
and the probability of failure, including the time of
repair. A servicing center should also reliably identify
the time of customer servicing, taking into account
the current construction task.
Fig.3. Closed queuing system
As a result of the application of the queuing theory, a
mathematical model should provide the data regard-
For the optimization of a construction process, a
ing the optimal design of servicing centers and, at
closed system is more convenient, where customers
the same time, determine the number of customers
given a certain time after the service return back into
taking into account the optimization parameters. The
the system and go to the queue again. Under a
parameters, under which the construction process
closed process there is understood a situation,
will be optimized, may be the following: time, number
where the source of requirements is final. The queue
of failures, fuel consumption, financial costs, envi-
length is limited and the processing of customers’
ronmental impact etc.
requirements is done according to the FIFO method
The queuing theory appeared in the early 20th cen-
(first in - first out).
tury. Fundamentals of the theory were developed by
As an example of the application in use there can be
Danish mathematician Agner Krarup Erlang (1878-
given an optimization of construction machinery at
1929), who examined the development of call cen-
the stage of "earth works", where the role of servic-
tres. According to D.G. Kendall6, any system of the
ing centers is performed by the loaders or excava-
queuing theory can be classified according to the
tors (number C) and the role of customers is played
following combination of letters and numbers:
in the system by trucks or dumpers (number D). We
investigate a system where D is greater than C.
For each construction process, a time unit can be
determined according to the depth of view on the
Fig.1. Classification of the queuing theory mathematical model of optimization, for example,
minute, hour, shift, week, month etc. For a proper
where A – describes the input stream of elements, B functioning of the mathematical model of the queuing
– describes the probability distribution during service theory application, the following range of conditions
time, C – describes the number of service lines, D – has to be met5:
specifies the maximum number of elements in the  input of an element into the queue can occur at
system, E – describes the queue discipline (finite, any moment of time;
infinite, FIFO, LIFO, etc.).  the number of inputs during the time interval
The parameters "A" and "B": in place "A" and "B" depends on the length of the interval and the
may be presented by the following symbols: type of distribution of a servicing centre per-
 M – for exponential distribution, formance (e.g. uniform, power-series, falling or
 D – for constants (deterministic intervals), rising) and the scheme of a servicing machine
 KK – for Erlang distribution of k-type, performance is given by the parameters of the
 G – any distribution. construction task; these are determined before
Basic structure of the queuing system is illustrated in the mathematical simulation and do not change
Figure 2. during the mathematical modelling;
 the probability, that in the interval of the length
δT occurs, more than one input converges to
zero more quickly than the length of the interval
δT;
 the average number of inputs per the unit of mathematical model we can perform a variety of
time is equal to λ. optimization tasks and propose an optimal choice of
To calculate the characteristics of the system, the machines combination for a particular construction
following formulae have to be used. The service process according to various criteria.
intensity can be determined as follows:
SIMULATION SOFTWARE

 , (1) For the mathematical modelling and simulation, spe-
C cial software was used1. The software allows an
easy input of additional more precise parameters
where: λ is a parameter of the exponential distribu-
and coefficients into the model. A mathematical
tion, which characterizes the time spent by an ele-
modelling was performed by the example of a con-
ment outside the operating system, for example,
struction phase "earth works". On the basis of the
removal of soil to landfill and return of the truck back
simulation results, an optimal design of machinery
to the servicing centre; µ is a parameter of the expo-
(excavators) for this task will be chosen from three
nential distribution, which characterizes the time
possible options, as well as the optimal number of
spent by an element during service, for example, soil
trucks for removal of soil. The decisive factor is the
loading on a truck by a loader.
task fulfilment within a certain time with minimal
The probability of Pk function, that in the time interval financial costs. The volume of the task is constant
of the length T number of elements k= 1, ..., C enter and does not change. Other machinery, in this case
the system, can be expressed as follows: the type of trucks (dumper), is predefined. The num-
k ber of channels is based on the geographical condi-
1  tions of the construction site, and, in our example, is
Pk      P0 (2)
k!    ,
given, N = 1. For simplicity, we will always choose
the same kind of excavators for each channel.
where P0 is determined as follows:

1
P0  k C
(3)
C
1  1  1   D C
      
  C!   
k  0 k!   

 
  
1 
,
After that we can easily calculate other properties of
the system:
The average number of customers in service:


ES   1  PD  (4)
 ,
Fig. 4. Mathematical model of the system

The average number of customers in the queue: The whole mathematical model can be divided into
D C
three parts: input, core of the model, and output. See
EL  l  P
l 1
C l (5) Figure 4. The mathematical model contains several
important subsystems for the calculation of machin-
,
ery failure and for the determination of economic
The average number of customers in the system: parameters of the system; see Figure 5.
Further, the input parameters of the mathematical
EK  ES  EL , (6) model will be described:
 construction process: excavation, construction
We can easily derive and use formulas for calcula-
pit, figure 1;
tion of further parameters of the system, for exam-
ple, system use, the average waiting time of an ele-  volume of the task: 6 000 m3, workability class:
ment in the queue inside and outside the service 3, loosening coefficient Kn = 1.25, the total vol-
system. Different probabilities of elements’ failure ume of soil Qexc = 6 000 x 1.25 = 7 500 m3;
and idle times can be identified. After inclusion of  parameters of auto truck (dumper) are shown in
other parameters in the mathematical model we can table 1.: dumper, body volume – 15 m3, the av-
calculate further properties of the system, such as erage travel time from site to landfill and back is
cost characteristics, duration of construction pro- 30 min., financial costs of machine operation:
cesses, environmental impact, fuel consumption, etc. fixed costs – 3000 CZK, variable costs – 1 000
After the introduction of all of the regularities in the CZK /hour, the maximum number of dumpers –
 15; the failure rate is 2%/day; the average repair  Working shift TS = 8 hours per day, i. e. 40 hours
time – 2 hours; time of delivery to servicing ma- per week.
chine is 2 minutes.  Construction task has to be performed maximum
in 2 weeks (TM = 80 working hours, that is 4 800
minutes).
 The input parameters of proper excavators are
described in Table 27. Most of the input parame-
ters of machines are given by the manufacturer
or calculated and averaged on the basis of ob-
servation and monitoring. An appropriate output
per hour of an excavator, without the influence
of random disturbances, according to9 is greater
than
Papp= =7 500 / 80 / 1 = 93.75 m3/hour (7)

 Basic time unit is a minute.

Parameter Variant 1 Variant 2 Variant 3

Bucket volume, [m3] 2 3 4

Fig. 5. Subsystems of the mathematical model in SW1 Average duration of
0.83 0.67 0.625
working cycle, [min]
Average hourly out-
100 120 150
put, [m3/hour]
Fixed costs, [CZK] 5 000 7 500 15 000
Variable costs,
2 000 3 000 4 000
[CZK/hour]
Probability of failure,
2 4 3
[%/day]
Average time of re-
60 80 90
pair, [min]
Table 2. Input parameters of excavator7

The total average time of the truck working cycle for

each variant without the influence of random factors,
according to9 is equal to:

TC1 = 40 + 2 + 15 x 60 / 100 = 40 + 2 + 6 = 48 min.

TC2 = 40 + 2 + 15 x 60 / 120 = 40 + 2 + 5 = 47 min.
TC3 = 40 + 2 + 15 x 60 / 150 = 40 + 2 + 4 = 46 min.

Fig. 6. Subsystems of the mathematical model in SW1 Hourly output of the truck for each variant according
to9 is the following:

Parameter Value Ptruck 1=Vbucket/TC1x60 = 15 / 48 x 60 = 12.50 m3/hour

Ptruck 2=Vbucket/TC2x60 = 15 / 47 x 60 = 12.77 m3/ hour
Body volume, [m3] 15 Ptruck 3 =Vbucket/TC3x60 = 15 / 46 x 60 = 13.04 m3/ hour
Average travel time, [min] 40
Average time of maneuvering and The minimum number of trucks for maximum occu-
2
unloading, [min]
pation of the excavator for each variant without the
Average loading time by excavator
15 x 60 / P influence of random factors, according to9 is:
performance, [min]
Fixed costs, [CZK] 3 000
PV1 = N x Pexc 1 / Ptruck 1 = 100 / 12.5 = 8,0 => 8
Variable costs, [CZK/hour] 1 000
dumpers
Average failure rate, [%/day] 2
PV2 = N x Pexc 2 / Ptruck 2 = 120 / 12.77 = 9,4 => 10
Average repair time, [min] 60
dumpers
Maximum number of dumpers, [-] 15
PV3 = N x Pexc 3 / Ptruck 3 = 150 / 13.04 = 11,5 => 12
Table 1. Input parameters of the auto truck7
dumpers
Other calculations for determining the random effects
will be performed in the simulation software1.
After the introduction of random values (machine
failures) in the mathematical model, according to the
simulation, the time of machine service extends:

TN1 = 6 + 0.058 = 6.058 min.

TN2 = 5 + 0.084 = 5.084 min.
TN3 = 4 + 0.059 = 4.059 min.

In the same way the intervals between inputs of Fig. 7. Calculation the number of units for completing
elements into service will increase: task

After that we will calculate the basic characteristics

TV = TV1 = TV2 = TV3 = 42 + 0.052 = 42.052 min. of the system. According to the calculation, that to
fulfil the excavation Q = 7 500 m3 within up to 80
The total average time of the truck working cycle for working hours (4 800 minutes) for the first variant 11
each variant with the influence of random factors dumpers will be needed, the second requires 9, and
(failures) is equal to: the third requires only 8 dumpers. The initial period
of the working cycle will be greater due to waiting in
TC1 = 42.052 + 6.058 = 48.11 min line for the service and will continually extend; the
TC2 = 42.052 + 5.084 = 47.14 min efficiency of the entire transport system with an in-
TC3 = 42.052 + 4.059 = 46.11 min creasing number of dumpers does not grow linearly,
see Figure 5. In the last step of selecting the optimal
The described calculation is arranged in Tables 3, 4, system we will focus on the assessment of the cost
5 for each variant. Working cycles of the excavator parameters for each variant. The optimal solution is
and trucks are interdependent at the service place. a variant with the lowest total costs.
Due to the random intervals between arrivals of ve- Calculation of the cost characteristics of the system
hicles, a queue to service will arise here. It is there- is shown in Tables 3 and 4 for each variant. In Figure
fore possible to examine the dependences between 7 the optimal set of machinery is evaluated. The
the transport system and an excavator with the use calculation and the selection of an optimal system
of the theory of waiting lines9. include the operational costs for the excavator and
trucks. The total working time does not exceed 80
To make the optimal choice of machinery it is neces- working hours.
sary first of all to determine other parameters of the In Table 3 it is shown that the total costs of trucks for
queuing theory. The intensity of service is equal to: the third variant are the lowest of all options and are
equal to 664,000 CZK, respectively 89 CZK per 1 m3
µ1 = 1 / TN1 = 1 / 6.058 = 0.1650709 min-1 of excavation. Table 4 shows, that the total costs of
µ2 = 1 / TN2 = 1 / 5.084 = 0.1966955 min-1 the excavator by the first variant are the lowest of all
µ3 = 1 / TN3 = 1 / 4.059 = 0.2463661 min-1 and are equal to 165,000 CZK, respectively 22 CZK
per 1 m3 of excavation.
The intensity of the input of elements into the queu- On the basis of 8 we choose the second variant. For
ing system: the task fulfilment, the second variant is chosen:
excavator with output of 120 m3/hour and 9 trucks for
λ = λ1 = λ2 = λ3 = 1 / TV = 1/42.052 = 0.023780 min-1 the transport of soil.

The intensity of the system operation:

ρ1 = λ / µ1 = 0.023780 / 0.1650709 = 0.14405974

ρ2 = λ / µ2 = 0.023780 / 0.1966955 = 0.12089794
ρ3 = λ / µ3 = 0.023780 / 0.2463661 = 0.09652335
 means to avoid unnecessary delays in construction,
Parameter Variant 1 Variant 2 Variant 3
as well as unnecessary additional costs due to a
Number of trucks, [-] 11 9 8 poor technology or wrong mechanization. The aim is
Fixed costs, [CZK] 3 000 3 000 3 000 to minimize these problems or even entirely elimi-
Total fixed costs, nate them. The one of the main aim of our research
33 000 27 000 24 000
[CZK] is to eliminate unnecessary delays and additional
Variable costs, [CZK costs. The mathematical modelling in the special
1 000 1 000 1 000
/hour] simulation software proved the applicability of the
Total variable costs, queuing theory for construction processes. In the
880 000 720 000 640 000
[CZK] case of the introduction of additional parameters into
Total costs, [CZK] 913 000 747 000 664 000 the system, the mathematical model will be closer to
3
Costs per 1 m of an actual construction process in reality. The model
3 122 100 89
excavation, [CZK /m ] can include a variety of random factors, including
Table 3. Calculation of costs characteristics of auto climatic and geographic conditions. On the basis of
trucks system
the simulation results, a construction manager can
justify a decision on choice of machinery according
Parameter Variant 1 Variant 2 Variant 3 to various criteria.
The mathematical simulation should show that some
Number of excava-
1 1 1 variants of machines combination fail to fulfil the task
tors, [-]
under the given conditions, and some will not be
Total fixed costs,
5 000 7 500 15 000 optimal in terms of costs or other parameters, while
[CZK]
Variable costs, [CZK other variants will be optimal from the view of costs
2 000 3 000 4 000 required to fulfil the construction task. The mathe-
/hour]
Total variable costs, matical simulation of the basic example showed that
160 000 240 000 320 000 the first option fails to perform the task under the
[CZK]
Total costs, [CZK] 165 000 247 500 335 000 given conditions; the second variant proved to be
Costs per 1 m3 of optimal in terms of the costs of the construction task
3 22 33 45 fulfilment. The simulation software1 allows us to look
excavation, [CZK /m ]
Table 4. Calculation of costs characteristics of excava- at the results in a graphical form or to export the data
tor system to other programs. The application of the queuing
theory allows us to introduce into the system a wait-
ing time for the servicing elements and to approxi-
Parameter Variant 1 Variant 2 Variant 3
mate the mathematical model to a real working task
Total costs of the auto on site.
913 000 747 000 664 000
trucks, [CZK] The application of queuing theory in the construction
Total costs of the exca- Industry is well known in the world and very wide
165 000 247 500 335 000
vator, [CZK] and well described in the technical literature. But the
Total costs, [CZK] 1 078 000 994 500 999 000 first time, we use a mathematical modelling in the
3
Costs per 1 m of exca- simulation software1. The present study demon-
3 144 133 134
vation, [CZK /m ] strates that it is possible to model mathematically
Table 5. Calculation of cost characteristics of the sys- and technically the whole complicated construction
tem „excavator + auto trucks“ process containing hundreds of constituent construc-
tion processes, with a number of simplifications, and
then to perform various calculations and changes for
an effective, efficient and long-term planning of con-
struction. Of course, we cannot state that our model
is flawless and absolutely accurate. It is therefore
important to verify the results of the study under real
conditions.

References
1. Simulation SW Matlab Simulink v. R2010b.
2. Liška, V., Macroeconomics, 2.vyd. Praha: Pro-
fessional Publishing, 2004.
Fig.8. Calculation of cost parameters of the system
3. Přikryl, P., Numerical methods of mathematical
analysis. Praha: SNTL, 1985.
SUMMARY AND CONCLUSIONS 4. Rektorys, K., Overview of Applied Mathematics.
To concentrate on reducing the costs of construction Praha: Prometheus, 2000.
5. Kendall, D.G., “Stochastic processes occurring
in the theory of queues and their analysis by the
method of the imbedded Markov chain“, The
Annals of Mathematical Statistics, Vol. 24, pp.
338-354, 1953.
6. The theory of queues,
www.mti.tul.cz/files/oa/obsluha/index.html
7. Building machinery KOMATSU, www.kuhn-mt.cz
8. Klvaňa, J., Operation Research 2. Praha: FSv
ČVUT, 2000.
9. Jarský, Č., On to the mathematical modelling of
construction processes. Disertační práce. Praha:
FSv ČVUT, 1981.
10. Jarský, Č., Automated Preparation and Man-
agement of Realisation of Project. Kralupy n.Vlt:
CONTEC, 2000.