Professional Documents
Culture Documents
Son2021 Article OptimizationOfConstructionMate
Son2021 Article OptimizationOfConstructionMate
Received 12 August 2020 Managing a construction project is challenging because of cost, time, safety, and quality
Revised 6 January 2021 considerations. In the most projects, the cost of construction is one of the most critical aspect
Accepted 17 February 2021 because material cost alone accounts for significant ratio of the total project. Therefore, the
Published Online 12 April 2021 cost of construction materials should be controlled. In this study, we proposed the use of
material requirements planning (MRP) to control the cost of construction materials. After
KEYWORDS determining the demand for the materials required for construction, we estimated both the
quantity of materials required and time taken to deliver the materials to the construction site.
Construction logistics planning Although economic order quantity models have been applied to analyze construction material
Optimizing material costs, they do not accurately reflect concerns related to material cost. Therefore, we used the
Inventory control material supply chain model (construction logistics planning) to analyze material costs. To
Dragonfly algorithm
optimize MRP according to the current progress of a project, a novel approach combining the
PSO algorithm
dragonfly algorithm (DA) and particle swarm optimization algorithm (PSO) was proposed. To
Optimize delivery schedules
verify the advanced searchability of the DA–PSO algorithm, the algorithm was compared with
Time to import construction materials
the gray wolf and the genetic algorithms.
CORRESPONDENCE Nguyen Huynh Chi Duy nhcduy007@gmail.com Construction Engineering & Management Department, Ho Chi Minh City University of Technology
(Bach Khoa University), Vietnam National University, Ho Chi Minh 700000, Vietnam
ⓒ 2021 Korean Society of Civil Engineers
KSCE Journal of Civil Engineering 2351
the CLP model will have additional financial costs – the (t) is built as a fixed-time ordering system (FOP) that replaces
cost the contractor spends in advance to buy large quantities of inventories at the beginning of fixed intervals when there are
materials therefore capital is tied up in materials inventory new orders to meet the demand for the next period.
(Mitsel et al., 2017). FOP is divided into 2 main types:
3. Stock-out Cost (SC): The cost of delay (Panova and FOP = 1: The trend of day-to-day supply of materials is equal
Hilletofth, 2018). to the demand for the daily use of materials, i.e., the materials are
4. Layout Cost (LC): This is the cost of transporting materials ordered just in time. With FOP = 1, the inventory will be discarded
in the construction site. Therefore, we can see that construction because each day the material is provided by demand.
logistics planning (CLP) reflects correctly and thoroughly all FOP > 1: The trend of ordering materials within a particular
kinds of costs during the management of construction time in advance. With FOP > 1, the inventory will appear.
materials, which helps in striking an appropriate balance
between the different types of expenses (ordering costs, 3.2 Implementation Process
financial costs, stock-out cost and layout cost for storage and This section presents a detailed description of the search and
construction site). Using the construction logistics planning analysis of effective material requirements planning to optimize
model to analyze the material cost (objective function) is material costs. The construction logistics planning (CLP) model
entirely reasonable and objective. Therefore, we conduct is used to review and analyze material costs with input parameters
searching and optimization of material demand plan at and constraint conditions. Input parameters and constraint conditions
different stages with the CLP model and the DA-PSO include construction progress, quantity of materials used for each
algorithm. task, material unit price, and exchange rate coefficient. The DA-
Fixed Ordering Period (FOP): Planning the procurement of PSO algorithm will then be used to determine the material
supplies and transports in the model is done by determining the requirements plan that will minimize the material cost.
optimal ordering time for each material that changes flexibly to
consider the demand for materials used throughout the project. 3.2.1 Simulation Process
As shown in Fig. 1, the procurement of materials in each phase Step 1: Determine the minimum required material use per day.
Fig. 1. Types of Fixed Ordering Period (FOP) (Said and El-Rayes, 2011)
Before developing a suitable material requirements plan, we cohesion; f is the food factor; e is the enemy factor; w is the
need to determine the minimum daily need for materials. In this inertial weight.
paper, we use the Last Planner System (Khanh and Kim, 2016) The avoidance of the static crashing of an individual into other
to determine the minimum daily need of materials by following individuals in the neighborhood, which is known as separation, is
these steps: calculated by Eq. (4):
1. Based on the construction schedule, we list all tasks that N
will be accomplished using materials whose use needs to Si = −∑ ( X − X j ) , (4)
be optimized (Olivieri et al., 2019). j =1
2. We determine the duration of each task (in the same unit: where X is the position of the current individual, Xj is the position
day, month or year). of the j-th neighboring individual, and N is the number of
3. We determine the quantities of materials used for each task. neighboring individuals.
4. We determine the daily quantities of materials for each Alignment, which refers to the matching of the velocity of an
task. individual to the velocity of others in the neighborhood, is computed
From the approved construction progress, we proceed to the by Eq. (5):
above-mentioned steps. We will accurately determine the minimum N
required material use per day at the construction site as in Table 1. ∑
j=
Xj
1 , (5)
Table 1 shows an example of the process of listing all project Ai =
activities and the amount of material used for each task. From N
there we can determine the exact amount of material used for the where Xj shows the velocity of the j-th neighboring individual.
project on each construction day, which will be done for each Cohesion, which is the propensity of individuals to be attracted
material in this project. towards the center of mass of the neighborhood, is given by Eq. (6):
Step 2: Declare the input parameters. N
The necessary data for the required CLP model are ∑
j=1
Xj
, (6)
1. OC costs: Material unit price; transportation cost; Ci = −X
2. FC cost: Material interest; N
3. SC costs: Late schedule penalties; where X is the position of the current individual, N is the number
4. LC costs: Cost of transport equipment; change in site of neighborhoods, and Xj is the position of the j-th neighboring
layout. individual.
Constraints: Fixed time limits between material needs (FOP); Attraction towards a food source, computed by Eq. (7), should be
warehouse size limit. the main objective for any swarm to survive:
Step 3: Initialize the first population. Initializing a population
F = X+ − X , (7)
by a random method of the algorithm helps ensure the objectivity i
of the population. The population in the study is the amount of where X is the position of the current individual and X+ shows
material in each material requirements plan during the period. the position of the food source.
Populations are directed to more energetic areas if they find Distraction of an enemy, which is computed by Eq. (8), is
potential prey and, conversely, they will look for new search another survival objective for the swarm:
areas if the current prey does not look promising enough.
E = X− − X ,
i
(8)
Each population will be checked for conditions of warehouse
size and material inventory. Each population is calculated to find where X is the position of the current individual and X − shows
the best population in each loop as shown in Eq. (1): the position of the enemy.
PSO:
CLC = ∑(OC + FC + SC + LC) . (1)
Vi (t ) = w(t ) ×Vi (t − 1) + C1r1 ×[X iL − X i (t − 1)] + C2 r2 ×[X iG − X i (t −1)]
Step 4: Improve the population. During the optimization (9)
Xi (t ) = Vi (t ) + X i (t − 1)
process, the populations are constantly changing and updating
based on the optimal parameters and equations of the DA-PSO 1. i = 1, 2, ..., P where P is the total number of individuals in
algorithm. The DA-PSO algorithm is a development algorithm the population
combining the PSO (Jia and Guo, 2016) and DA algorithms as 2. t = 1, 2, ..., T where T is the number of limited generations
shown in Eqs. (2) and (3). (or number of iterations)
DA Algorithm: 3. X Li = { X Li1, X Li 2, …, X LiM } represents the best position the
individual achieves after (t − 1) loop (called pbest)
ΔX +1 = (sS + α A + cC + fF + eE ) + wΔX , (2) G G G G
t i i i i i t
4. X i = { X i1, X i2, …, X iM } represents the best position of
the whole population after (t − 1) loop (called gbest)
X +1 = X + ΔX +1 , (3)
t t t
5. C1 and C2 are the learning factors
s is the separation weight; α is the alignment weight; c indicates the 6. r1 and r2: changes from 0 to 1 are random standards
2354 P. V. H. Son et al.
Fig. 3. Relationship between the Material Cost and Warehouse Size in Fig. 4. Relationship between the Material Cost and Warehouse Size in
Phase 1 Phase 2
the warehouse increases, and vice versa. This phenomenon is construction site. Therefore, the warehouse selection, cost, and
reasonable. To obtain a high discount from the supplier, the FOP are critical. The relationship diagram enables managers to
contractor buys larger quantities of material than actually required, take optimal decisions.
which results in more inventories. Therefore, large warehouses Currently, for most projects, contractors have to spend money
are necessary. on construction first and then claim the money from investors.
The intersection of the warehouse-size–FOP and material Because the cost of a construction project is high, contractors
cost–FOP relationship curves is considered optimal outcome and typically borrow from a bank. Because contractors have to buy
provides optimal FOP results, warehouse size, and corresponding materials in large quantities to avail a high discount from suppliers,
costs. The optimum research point is the equilibrium position considerable money is invested in the material inventory. Therefore,
corresponding to the minimum values of the material cost and unreasonable revenue and expenditure plans are risky and may
warehouse size. However, the decision-maker may choose different result in considerable amounts of losses for or even bankruptcy
optimal goals according to the perspective of the manager or the for contractors if the purchased inventory is not utilized. Therefore,
situation at the site. For example, if a warehouse area larger than in addition to OC, other cost-related factors should be controlled
the construction site is not a feasible solution, the decision-maker to optimize construction costs.
should consider a reasonable FOP to achieve the most beneficial The optimal FOP results in phases 1, 2, and 3 are 4, 4, and 3,
results based on the relationship curve. respectively. The FOP is in the range 3 – 4, which indicated that
The relationship curve chart indicates that the further a point materials were ordered every 3 or 4 days. This approach presented
is from the optimal point (increasing FOP), the lower is the cost the following advantages:
reduction and the higher is the increase in the warehouse The proposed method approach minimized capital spending
capacity. For example, in phase 1, the FOP increased from 4 to 5. on inventory, which reduced the amount of money borrowed
Therefore, for a material cost reduction of 46.4 million, the from the bank and resulted in the availability of additional money
warehouse capacity must increase by 19.8 tons; in phase 2, the for investment in other areas, such as efficient construction methods,
FOP increased from 11 to 14. Therefore, for a material cost to improve project quality and expedite development; thus, the
reduction of 7.7 million, the warehouse size must increase to brand of the company was enhanced. Furthermore, the cash flow
38.8 tons. However, this approach is not optimal because the plan of the project was controlled.
warehouse size increases excessively, whereas the cost benefit is This approach eliminates the necessity of large storage spaces.
negligible. Therefore, the approach facilitates the use of small warehouses
Purchases and shipping (OC) accounts for the highest proportion in construction sites. Currently, most construction projects in big
(90%) of material costs, SC accounts for the second-highest cities (Ho Chi Minh, Hanoi, Da Nang, and Hai Phong) have
proportion (5%), and FC and LC account for the rest. Therefore, limited construction sites. Thus, small warehouse size is the
to reduce material management costs, contractors focus on OC primary constraint in the site layout. The FOP value that corresponds
but not on other costs. To minimize OC, managers negotiate and with a small warehouse is selected to address this constraint.
purchase materials in large quantities to obtain the highest Furthermore, with advantages such as reduction in bank loan
discount from the supplier. When the volume purchased is larger costs, a short FOP is critical for construction enterprises in
than the volume required, it leads to an inventory surplus. setting up material supply plans for the construction site.
Therefore, a large warehouse is required to store this inventory Moreover, a short FOP maintains low surplus of inventory
surplus. However, the size of the warehouse is a problem that the because new materials are purchased when required. The quality
contractor cannot control because it depends on the size of the of such fresh products remains superior to those stored for a long
KSCE Journal of Civil Engineering 2357
term. When stored for too long, the quality of the material may guarantee the optimal solution, which is the major drawback of
be easily degraded by the surrounding environment. In many the GA because the population that could have been closer to a
cases, the quality of the material has to be retested because better position is inadvertently disturbed to obtain optimal
materials have been in stock for too long. Reinspection of quality results. Therefore, the convergence chart of the GA tends to have
is not recommended because it directly affects the progress of the many long horizontal lines because the optimal position could
project. not be determined due to this disturbance (Li et al., 2020).
Therefore, decision makers should be flexible and consider In the GWO algorithm, optimization is performed based on a
factors, including objective and subjective factors, before taking hierarchy of alpha (best), beta (second best), and delta (third best)
final decisions. solutions. The next position is updated based on the positions of the
The results of the analysis indicate that the approach was three solutions. Thus, alpha, beta, and delta estimate the optimal
reasonable and practical. To verify the feasibility of these results, positions and other solutions update their positions randomly
we compared results of the dragonfly algorithm (DA)–particle around the current optimal positions. Therefore, the GWO algorithm
swarm optimization (PSO) model with the GA and the GWO exhibits a strong advantage in the exploitation stage compared
algorithm. with other algorithms. The expansion of the search area relies
only on the vectors in the optimal equation, which results in a
Comparison of DA–PSO with GWO and GA
lack of diversity of the population; the limited expansion of the
We considered phase 2, the phase with the most demand for
search area is a drawback of the GWO algorithm. Therefore, for
materials, for the comparison.
a problem with large search spaces, the GWO algorithm easily
According to Table 7, the DA–PSO algorithm outperformed
falls into local optimization (Hatta et al., 2019).
the GWO and GA algorithms in terms of cost and warehouse
The DA can easily expand its search space. The algorithm is
optimization. The optimized material cost obtained using the
based on the behavior of a dragonfly, which often changes
DA–PSO algorithm was less than those obtained using the GWO
direction suddenly. On combining with the Levy algorithm, the
and GA algorithms, which were 4.1 million and 10.2 million
search area can be increased rapidly, rendering the search space
dollars higher than the DA–PSO value, respectively.
of the DA algorithm larger than those of other algorithms.
Figure 6 displays that with 500 loops, the DA–PSO algorithm
However, the optimization capability (exploitation) of the DA is
provides better solutions than those of the other two options
extremely slow and easily leads to local optimization. The PSO
The reasons for the superior performance of the DA–PSO
algorithm was combined with the DA algorithm to overcome the
algorithm are discussed.
shortcomings of the DA. The PSO algorithm is an excellent
Because of hybridization, selection, and mutation, the GA
spatial exploitation algorithm because it regularly updates the
exhibits an extensive search area. However, scrambling by a
best location of the population, directs the individuals to the best
random mechanism and not following specific rules does not
location, and updates the best location of the population in each
loop based on the optimal function. Thus, in the DA–PSO model,
Table 8. Result of the Project
the DA is used to expand the search (exploration) space, whereas
Algorithm
Material cost Warehouse the PSO algorithm is used to explore the space that has been
(million VND) (ton) searched by the DA. Therefore, the DA–PSO model outperforms
DA-PSO 13,516.8 85.3 the GA and the GWO algorithm.
GWO 13,520.9 87.5
GA 13,527 88.4 5. Conclusions
options should be carefully considered; the relationship diagram Eberhart R, Kennedy J (1995) A new optimizer using particle swarm
is used obtain optimal conditions. The proposed approach theory. Proceedings of the sixth international symposium on micro
provides flexibility in decision-making. Project managers should machine and human science (MHS'95), October 4-6, Nagoya, Japan
Eberhart RC, Yuhui S (2001) Particle swarm optimization: Developments,
consider constraint conditions and other factors, including
applications and resources. Proceedings of the 2001 congress on
objective and subjective factors, for determining the optimal evolutionary computation (IEEE Cat No01TH8546), May 27-30,
material cost for the available storage space in each phase of the Seoul, Korea, 81-86
construction. Georgy M, Basily SY (2008) Using genetic algorithms in optimizing
Optimal results are obtained for short FOPs (approximately construction material delivery schedules. Construction Innovation
3 – 4 days) because of the following reasons: First, contractors 8:23-45, DOI: 10.1108/14714170810846503
borrow less money from banks and cash flow can be easily Hatta NM, Zain AM, Sallehuddin R, Shayfull Z, Yusoff Y (2019) Recent
controlled. Second, a large storage space is not required. Therefore, studies on optimisation method of grey wolf optimiser (GWO): A
review (2014-2017). Artificial Intelligence Review 52:2651-2683,
the site layout need not be too big. Because most construction
DOI: 10.1007/s10462-018-9634-2
projects in major city centers have limited space, a short FOP is Hsu P-Y, Angeloudis P, Aurisicchio M (2018) Optimal logistics planning
suitable. Thus, a short FOP addresses construction limitation for for modular construction using two-stage stochastic programming.
managers. Furthermore, a short FOP limits material inventory Automation in Construction 94:47-61, DOI: 10.1016/j.autcon.2018.
surplus because materials are regularly purchased and updated 05.029
continuously. Therefore, high-quality materials are guaranteed. Jaśkowski P, Sobotka A, Czarnigowska A (2018) Decision model for
Thus, construction enterprises should prioritize short FOPs for planning material supply channels in construction. Automation in
the construction site. Construction 90:235-242, DOI: 10.1016/j.autcon.2018.02.026
Jia Q, Guo Y (2016) Hybridization of ABC and PSO algorithms for
Furthermore, the solution provided by the proposed DA–PSO
improved solutions of RCPSP. Journal of the Chinese Institute of
model was compared with the solutions provided by the GA and Engineers 39:727-734, DOI: 10.1080/02533839.2016.1176866
the GWO algorithm to evaluate the suitability of the DA–PSO Khanh HD, Kim SY (2016) A survey on production planning system in
model in solving single-target problems with a large and dependent construction projects based on last planner system. KSCE Journal of
search space (MRP). The DA–PSO model outperformed the GA Civil Engineering 20(1):1-11, DOI: 10.1007/s12205-015-1412-y
and PSO algorithm. Therefore, DA–PSO algorithm is suitable Li Y, Huang Z, Xie Y (2020) Path planning of mobile robot based on
for achieving optimal results in limited time. improved genetic algorithm. 2020 3rd international conference on
This study can be extended in many directions in the future. electron device and mechanical engineering (ICEDME), May 1-3,
Suzhou, China, 691-695
The cost of construction materials is not constant but changes
Mao H, Cheng P (2010) Design of material delivery system based on
randomly. Therefore, future studies should consider the uncertainty lean construction. International conference of logistics engineering
of the unit price of materials. Furthermore, the supplier capacity and management (ICLEM), October 8-10, Chengdu, China, 1793-
is a high-risk factor and should be considered. 1799
Marini F, Walczak B (2015) Particle swarm optimization (PSO). A
Acknowledgments tutorial. Chemometrics and Intelligent Laboratory Systems 149:153-
165, DOI: 10.1016/j.chemolab.2015.08.020
Not Applicable Min W, Sui Pheng L (2005) Economic order quantity (EOQ) versus
just-in-time (JIT) purchasing: An alternative analysis in the ready-
mixed concrete industry. Construction Management and Economics
ORCID 23(4):409-422, DOI: 10.1080/01446190500041339
Mirjalili S (2016) Dragonfly algorithm: A new meta-heuristic optimization
Pham Vu Hong Son https://orcid.org/0000-0002-9788-0627 technique for solving single-objective, discrete, and multi-objective
Nguyen Huynh Chi Duy http://orcid.org/0000-0002-2201-955X problems. Neural Computing and Applications 27:1053-1073, DOI:
Pham Ton Dat https://orcid.org/0000-0003-2343-1084 10.1007/s00521-015-1920-1
Mitsel AA, Kritski OL, Stavchuk LG (2017) An inventory model with
random demand. Journal of Physics: Conference Series 803:012099,
References DOI: 10.1088/1742-6596/803/1/012099
Nolz PC (2020) Optimizing construction schedules and material deliveries
Ala-Risku T, Kärkkäinen M (2006) Material delivery problems in in city logistics: A case study from the building industry. Flexible
construction projects: A possible solution. International Journal of Services and Manufacturing Journal, DOI: 10.1007/s10696-020-
Production Economics 104:19-29, DOI: 10.1016/j.ijpe.2004.12.027 09391-7
Behera P, Mohanty RP, Prakash A (2015) Understanding construction Olivieri H, Seppänen O, Alves TdCL, Scala NM, Schiavone V, Liu M,
supply chain management. Production Planning & Control 26:1332- Granja AD (2019) Survey comparing critical path method, last planner
1350, DOI: 10.1080/09537287.2015.1045953 system, and location-based techniques. Journal of Construction
Bose S, Goswami A, Chaudhuri KS (1995) An EOQ model for deteriorating Engineering and Management 145:04019077, DOI: 10.1061/(ASCE)
items with linear time-dependent demand rate and shortages under CO.1943-7862.0001644
inflation and time discounting. Journal of the Operational Research Panova Y, Hilletofth P (2018) Managing supply chain risks and delays
Society 46:771-782, DOI: 10.1057/jors.1995.107 in construction project. Industrial Management & Data Systems
KSCE Journal of Civil Engineering 2359