Documentos de Académico
Documentos de Profesional
Documentos de Cultura
To cite this Article Hyari, Khalied, El-Mashaleh, Mohammad and Kandil, Amr(2010) 'Optimal Assignment of Multiskilled
Labor in Building Construction Projects', International Journal of Construction Education and Research, 6: 1, 70 80
To link to this Article: DOI: 10.1080/15578771003590284
URL: http://dx.doi.org/10.1080/15578771003590284
Introduction
Attracting and retaining a skilled workforce is one of the greatest challenges facing
the construction industry (Ejohwomu, 2007; Kazaz & Ulubeyli, 2007; Srour, Hass, &
Morton, 2006; Goodrum, 2003; Kashiwagi & Scott, 2002; Rowings, Federle, &
Birkland, 1996). This shortage of skilled labor motivated construction firms to look
for ways to utilize their workforce more efficiently (Ejohwomu, 2007; Srour et al.,
2006; Cai & Li, 2000). Multiskilled labor is one of the strategies that have been
adopted to improve the efficiency of resource utilization in construction firms
(Ejohwomu, 2007; Goodrum, 2003). In this strategy, some workers in the company
are hired or trained to perform many types of work in construction projects. For
example, a marble installer can perform tasks such as brickwork or tiling in addition
70
71
72
K. Hyari et al.
4) increasing the average employment duration of skilled workers at the project site
(McGuinness & Bennett, 2006).
The success of the multiskilling strategy in attaining the above mentioned advantages depends on the ability of project management personnel to assign workers to
appropriate tasks and compose crews effectively during short term planning
(Gomar, Hass, & Morton, 2002; Hass et al., 2001). This task can be very challenging
due to the large number of possible alternatives in assembling crews and the many
constraints that should be taken into consideration in work assignments including
worker skills and availability. Despite the reported advantages of multiskilled workforce in the construction industry, few models were developed to support site management personnel in assigning multiskilled workers to construction tasks. Hegazy,
Shabeeb, Elbeltagi, & Cheema (2000) developed a heuristic for assigning less utilized
multiskilled resources to other construction activities that suffer from insufficient
resources during their shortage period in order to reduce project duration. Tam,
Tong, Cheung, & Chan (2001) developed a genetic algorithm model for multiskilled
workers assignment considering their productivity differences with the objective of
minimizing the total cost of labor. Gomar et al. (2002) developed a linear programming model for assigning multiskilled workers from a pool of workers aiming at
minimizing workers turnover and switching from crew to crew. However, labor
productivity rates are not considered in the model.
This paper presents an alternative model for optimal assignment of multiskilled
labor to construction tasks that should be accomplished in a workday. The model
was designed to support project managers and=or superintendents in their effort
to assign available workers to the needed work in a way that minimizes the total
labor cost of the job. This model is intended for small to medium-sized contracting
firms that possess small groups of laborers that conduct their work as a team. The
present model considers all the practical constraints to this work assignment problem. The following sections of this paper present the development of the model
followed by an application example to demonstrate its capabilities.
Model Development
The primary purpose of this development stage is to formulate a simplified optimization model that supports construction companies in their adoption of the functional
flexibility in human resources strategy (i.e., multiskilled workforce). The objective of
this model, therefore, is to find the optimal work assignment for a multiskilled labor
force while satisfying the practical constraints that are commonly involved in this
assignment problem. To this end, the present model is formulated in three major
steps: 1) formulating the objective function of this resource utilization problem; 2)
defining the practical constraints and considerations that need to be satisfied in
the developed optimization model; and 3) implementing the model using a simple
and easy-to-use optimization tool to ensure the practicality of the developed model.
Model Objective
The objective of this optimization problem is to minimize the cost of labor hours utilized
to perform the work as shown in Equation (1). However, the unit cost of performing
task (i) by worker (j) is a function of the hourly wage (Wj) of worker (j), and his=her productivity (Prij) in performing task (i) as shown in Equation (2). Therefore, Equation (1)
73
can be rewritten as shown in Equation (3). It should be noted that worker (j) will be paid
the same hourly wage regardless of the task (i) performed by that worker. Obviously,
this hourly wage takes into account the multiskilling capability of the worker.
Minimize cost of labor hours utilized
I X
J
X
Xij Cij
i1 j1
where Xij Amount of work in task (i) performed by worker with skill type (j);
Cij Unit cost of work performed in task (i) by worker with skill type (j);
i Construction task (i 1 to I); and j Worker skill type (j 1 to J).
Cij
Wj
Prij
where Wj hourly wage of worker with skill type (j); and Prij productivity rate for
worker with skill type (j) in performing task (i) in units per hour.
Minimize cost of labor hours utilized
I X
J
X
i1 j1
Xij
Wj
Prij
The number of workers assigned from each skill type (Nj) can be obtained by
dividing the amount of work assigned to worker with skill type (j) in each task (i)
over their corresponding productivity in that task as shown in Equation (4).
Nj
I
X
i1
Xij
Prij D
Xij Qi
j1
74
K. Hyari et al.
3) The number of workers assigned from each skill type (j) should be an Integer.
Nj
I
X
i1
Xij
Integer
Prij D
This constraint is intended to ensure the minimization of labor idle times that may occur
when workers in some skill types are assigned less than D hours of work on any day.
This formulation will provide a more fair distribution of work among workers because
idle workers from any skill type can be utilized to help in other tasks.
4) Non-negativity constraint for all variables of the optimization model.
Xij 0 for all tasks (i 1 to I) and all workers skill types (j 1 to J).
5) The number of workers assigned from each skill type should not exceed the
maximum available number of workers. This constraint accounts for a practical
limitation where limited supply of workers in some skill types may exist.
Nj Uj
where Uj upper limit on the supply of workers from skill type (j).
Model Implementation
The simplified formulation of the developed model enables its solution using the mixed
integer linear programming method (MILP). This mathematical optimization method
is an extension of linear programming method, with an additional restriction that
some of the decision variables are constrained to have only integer values at the
obtained optimal solution. MILP is characterized by its small computation complexity
which allows it to be implemented using commonly available solvers. This intended simplicity of the developed optimization problem formulation is needed to ensure the practicality of the developed model in supporting construction personnel in their selection of
short-term work assignments for their workforce (e.g., daily work assignments). The
Spreadsheet Solver that is included as a free add-in program in Microsoft ExcelTM
(i.e., Excel SolverTM) was selected to implement this model due to its availability in
any personal computer that has Microsoft Office installed. The solver has a userfriendly graphical user interface and can be used easily by individuals who do not have
much experience or knowledge of mathematical solvers (Fylstra, Lasdon, Watson, &
Waren, 1998). The solution of an optimization problem using Excel SolverTM is
performed in the following four main steps: 1) specifying the cell which contains the
objective function, and creating a formula for the objective function in that cell; 2) stating the decision variables; 3) specifying the cells that define the constraints; and 4) indicating the type of the model (i.e., linear, nonlinear, integer etc.). The interested reader is
referred to Flystra et al. (1998) for more detailed information on the use of Spreadsheet
Solver. The following section presents an application example that illustrates the use of
the present optimization model and demonstrates its capabilities.
Application Example
This application example optimizes work assignments on a finishes project that consists of five tasks including tiling, brickwork, flooring screed, terrazzo, and marble
75
Table 1. Quantity of work and productivity rates for different tradesmen in different
tasks
(j) Tradesmen
j1
j2
j3
j4
j5
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
i1
Tiling
i2
Brickwork
i3
Flooring
screed
i4
Terrazzo
i5
Marble
fixing
300
400
500
300
310
Wj Hourly
wage ($)
50
40
48
50
60
2
4
2
0
2
0
2.5
5
3
0
0
0
2.5
5
0
2
0
0
0
3
76
K. Hyari et al.
Table 2. Number of different types of tradesman used when the supply is
unlimited (Scenario 1)
Number of workers
Tradesman
Present Model
12
14
12
8
13
59
23,329.03
13
13
13
8
13
60
23,792
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
Total number of workers
Total cost ($)
different trades. As shown in Table 2, the present model yields a total labor cost of
$23,329.03 and a required total number of workers of 59. This obtained result is better
than that obtained by Tam et al., whose model determined the total labor cost to be
$23,792 and the required number of workers to be 60. The improvement in the results
obtained by the current model can be attributed to the utilization of the multiskilling
capability of workers who work less than a full day in their primary area of specialization by assigning them to other tasks for the rest of the working day as shown in
Table 3. The added benefit from this approach is that it ensures a more fair distribution of work among workers from different crafts. The current model produces
the output in less than a second whereas a running time of 10 hours was reported to
be required for obtaining the solution in the previous model (Tam et al., 2001).
Tiling
Brickwork
Flooring
screed
Terrazzo
Marble
fixing
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
Sum output
300
283.4
13.2
0
0
3.4
300
400
0
400
0
0
0
400
500
0
8
480
12
0
500
300
0
0
0
300
0
300
310
3
0
0
0
307
310
77
Present model
10
18
12
8
13
61
23,809.03
10
18
13
8
14
63
24,672
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
Total number of workers
Total cost ($)
Tiling
Brickwork
Flooring
screed
Terrazzo
Marble
fixing
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
Sum output
300
235.4
61.2
0
0
3.4
300
400
0
400
0
0
0
400
500
0
8
480
12
0
500
300
0
0
0
300
0
300
310
3
0
0
0
307
310
of work in each task to the workers from the different crafts. Again the current
model produces better results than those obtained by Tam et al. (2001). The total
labor cost in the current model was $23,809.03 and the required total number of
Present model
10
18
12
13
13
66
25,809.03
10
17
15
12
15
69
27,000
78
K. Hyari et al.
Tiling
Brickwork
Flooring
screed
Terrazzo
Marble
fixing
Tiler
Bricklayer
Screeding layer
Terrazzo layer
Marble layer
Sum output
300
235.4
61.2
0
0
3.4
300
400
0
400
0
0
0
400
500
0
8
480
12
0
500
500
0
0
0
500
0
500
310
3
0
0
0
307
310
workers was 61 while the previous model reported a total labor cost of $24,672 and a
required number of workers of 63.
Scenario 3: Increased Amount of Work in Some Tasks
In this scenario, the amount of the needed terrazzo work was increased from 300 to
500 square meters (m2), in addition to the upper limit of 10 tilers used in the previous
case. As shown in Tables 6 and 7, the current model produces better results than the
previous model with a total labor cost of $25,809.03 rather than $27,000 and a total
number of workers of 66 rather than 69.
Conclusions
A simplified model was developed to solve the problem of assigning multiskilled
workers to daily construction tasks. The model was formulated as a mixed integer
linear programming model that can be solved easily using available solvers. The
model is implemented using Excel SolverTM which is bundled with Microsoft
ExcelTM to ensure the practicality of the model and the ease of its use. An application example was analyzed to demonstrate the capability of the model and compare it to a previously developed model. The present model was found to be
superior to the previous model in terms of: 1) quality of solutions since it utilized
a mathematical optimization method that is guaranteed to produce a global optimal
solution while the previous model utilized a genetic algorithm that yielded good solutions that may not have been globally optimal (i.e., where no proof of convergence
exists); 2) great savings in time needed to yield solutions since the developed model
produces the solution in less than a second, while the reported time of the previous
model was several hours; 3) practicality and simplicity as the current model utilized
widely available solver which comes as a free add-in program to Microsoft ExcelTM;
and 4) eliminating the need for parameter settings that is needed in the previous
model which affects the quality of the obtained solutions. The model is expected
to help superintendents in their endeavor to assign workers to construction tasks
in a way that minimizes the total cost of labor by considering their skills, productivity, and availability. One limitation that this model faces is a property of
79
the multiskilling strategy itself. The limitation is that multiskilling itself would not be
applicable in highly unionized labor markets that do not allow workers to perform
tasks outside their precisely defined trades or crafts.
References
Ballard, G. (2001). Cycle Time Reduction in Home Building. Proceedings of the 9th annual
conference of the international group of lean construction, National University of
Singapore. 68 August 2001.
Brusco, M. J. & Johns, T. R. (1998). Staffing a multiskilled workforce with varying levels
of productivity: An analysis of cross-training policies. Decision Science Journal, 29,
499515.
Burleson, R., Haas, C., Tucker, R., & Stanley, A. (1998). Multiskilled labor utilization
strategies in construction. Journal of Construction Engineering and Management, 124(6),
480489.
Cai, X. & Li, K. N. (2000). A genetic algorithm for scheduling staff of mixed skills under
multi-criteria. European Journal of Operational Research, 125, 359369.
Eitzen, G., Panton, D., & Mills, G. (2004). Multi-skilled workforce optimization. Annals of
Operations Research, 127, 359372.
Ejohwomu, O. (2007). Modeling the supply and demand for construction and building services
skills in the black country. PhD Dissertation, University of Wolverhampton, West
Midlands, UK.
Fylstra, D., Lasdon, L., Watson, J., & Waren, A. (1998). Design and use of the Microsoft
Excel solver. Interfaces, 28(5), 2955.
Gomar, J., Haas, C., & Morton, D. (2002). Assignment and allocation optimization of
partially multiskilled workforce. Journal of Construction Engineering and Management,
128(2), 103109.
Goodrum, P. (2003). Worker satisfaction and job preferences in the U.S. construction
industry. Proceedings of the ASCE Construction Research Congress, Winds of Change:
Integration and Innovation of Construction, March 1921, 2003, Honolulu, Hawaii,
USA.
Hass, C. T., Rodriguez, A. M., Glover, R., & Goodrum, P. M. (2001). Implementing a multiskilled workforce. Construction Management and Economics, 19, 633641.
Hegazy, T., Shabeeb, A., Elbeltagi, E., & Cheema, T. (2000). Algorithm for scheduling multiskilled construction resources. Journal of Construction Engineering and Management,
126(6), 414421.
Kashiwagi, D. T. & Scott, M. (2002). Solving the construction craftperson skill shortage problem through construction undergraduate and graduate education. ASC Proceedings of
the 38th Annual Conference, Virginia Polytechnic Institute and State University,
Blacksburg, VA; 165176.
Kazaz, A. & Ulubeyli, S. (2007). Drivers of productivity among construction workers: A study
in a developing country. Building and Environment, 42, 21322140.
Lill, I. (2002). Who should take responsibility in professional training of construction
workers. Proceedings of the 2nd International Conference of Postgraduate Research
Conference in the Built and Human Environment: International Postgraduate Research
Conference in the Built and Human Environment, University of Salford, 1112 April
2002, pp. 342361.
Loucks, J. S. & Jacobs, F. R. (1991). Tour scheduling and task assignment of a heterogeneous
work force: A heuristic approach. Decision Sciences, 22(4), 719738.
Maturana, S., Alarcon, L. F., & Deprez, M. (2003). Modeling the impact of multiskilling and
concrete batch size in multi-storey buildings. Proceedings of the XI Conference on Lean
Construction, IGLC-11, July 2003, Blacksburg, Virginia, pp. 553566.
80
K. Hyari et al.