Cost Optimisation of the Design of Reinforced Concrete Flat

Slab to BS8110

S O A Olawale1, M A Tijani1,*, M A Kareem1, A M Ogungbire2 and O Alabi3

1
Department of Civil Engineering, Osun State University, Osogbo, Nigeria
2
Department of Electrical/Electronic Engineering, Osun State University, Osogbo,
Nigeria
3
Department of Physics, Osun State University, Osogbo, Nigeria

*Corresponding author: murtadha.tijani@uniosun.edu.ng

Abstract. The optimum design of reinforced concrete flat slab could reduce its construction
cost because it is usually employed in large floor area without any structural framing beams.
The design consideration is based on the provision of BS8110 separating the slab into middle
and column strips. The optimisation is to find the optimum slab thickness with least cost. The
use of the in-built genetic algorithm function of MATLAB software was employed to
minimise the design at various steel ratios. The constraints applied were non-linear and thus
required a lengthy iteration cycles before convergence. The objective function for either strip
is treated separately without any coupling in line with the design philosophy of BS8110. The
optimum thicknesses for the middle and column strips are 130mm and 140mm respectively.
The reinforcement ratios for the optimum design were also established to be 2% and 3.5% for
the middle and column strips respectively.

1. Introduction
A flat slab is a reinforced concrete plate supported by columns only without any framing beams [1].
The columns can be provided with heads to improve punching shear resistance where there is
occurrence of high shear [2]. Flat slab systems are quite popular and can be found around residential
buildings as well as public buildings. With the absence of beams in flat slabs, lower storey heights are
made possible. These lowered stories height ultimately leads to cost reduction. The most important
aspects of any structural design are safety and cost. Generally, in the practical realm, economics is
given little consideration. Suitable sections are designed as long as they satisfy the given conditions.
Therefore, it is important to introduce the concept of optimisation to the design of structural members.
These members are optimized within the actual constraints at both ultimate and service levels of the
locally used design codes. To optimise any system, there are many approaches such as genetic
algorithm, fuzzy logic, and neural network among many others. However, for structural members,
optimisation, evolutionary genetic algorithms (EGA) presents a modern technique of optimisation.
EGAs are subsets of evolutionary computation, a generic population-based meta-heuristic optimisation
algorithm. This method leans on to the biological process which are easily programmed using any high
level language.
Genetic Algorithm (GA) is a type of evolutionary algorithm and are global optimisation techniques.
Various biological processes of reproduction and natural selection was imitated in order to solve for
the ‘fittest’ solutions in [3]. Several of the genetic algorithms processes are random just like in

1
evolution. However, with the genetic algorithm, there is an ability to control the algorithm technique
and level of randomization [4]. Genetic algorithm provides better solution than both exhaustive and
random search algorithms without the need for extra information on the given problem. This makes
them the perfect fit in finding solutions to different problems where other techniques and algorithms
have failed. Thus, the current work studied the structural optimisation in reinforced concrete
structures, with the focus on the cost optimisation of the design of reinforced concrete flat slabs.
Efforts have been made to provide state of the art software that incorporates genetic algorithm as
demonstrated in MATLAB.
Reinforced concrete flat slabs are highly versatile elements widely used in building construction
allowing flexible column grids. They generally transfer the loads directly to the supporting columns
which are spaced suitably around the slab. The finishing of the slab obtained not only enhances the
aesthetics of the building, but also helps to diffuse light better and reduce their vulnerability to fire.
Furthermore, the absence of beams from flat slabs reduces the overall height of the entire building,
thus saving some amount of cost. In addition to this, extra fittings to the building such as auto
sprinkler become much easier to incorporate. There is however, limitation to the span and thickness of
the slab in addition to their unsuitability for the support of brittle (masonry) partitions.
Galeb and Atiyah [5] worked on the optimum design of reinforced concrete waffle slabs using
genetic algorithms. Two case studies were compared; the first is a waffle slab with solid heads, while
the other is a waffle slab with band beams along the column centrelines. A MATLAB program was
written for the optimisation and the result shows that the cost of formwork was reduced by 30%. It
was also further discovered that a top slab of 62mm – 72mm was the most economic for slabs with
solid heads, while most texts set their practical limit between 75mm and 125mm. Matthias et al. [6]
worked on a multi-objective optimisation in which both cost and mass minimization of the structure is
being done concurrently. Non-dominated Sorting Genetic Algorithm (NSGA – II) and meta-model
was employed in getting the optimum solutions. The design variables are the cross-sectional
dimensions of the structure to get optimum cost and weight. While there are several optimisation
techniques to solve the constrained multi-objective problems, appropriate techniques depend on the
type of problem and NSGA – II is a quite popular method especially for strength which was the reason
for its choice. The entire procedure for the optimisation was implemented on MATLAB.
Ghandi et al. [7] presented a Cuckoo Optimisation Algorithm (COA) model for the cost
optimisation of the one-way and two-way reinforced concrete (RC) slabs according to the ACI code.
Its objective function is the cost of concrete and reinforcing steel. Constraints were also developed to
conform to the requirements of ACI code. The model can be applied in practical designs so as to
reduce project cost. This is also the first application of Cuckoo Optimisation Algorithm to the
optimisation of reinforced concrete slabs. The result of the COA was compared with neural dynamics
model, in which the COA achieved better results.
Sudarshana and Ramesh [8] demonstrated the design of short columns under biaxial bending
making use of artificial neural network (ANN) and genetic algorithm (GA). It was demonstrated that
the hybrid model does not perform significantly better than ordinary genetic algorithm. An
optimisation model for the design of rectangular reinforced concrete beams subject to a specified set
of constraints was developed by [3]. Genetic Algorithm was used to solve this, and the results were
compared with a mathematical programming technique that deals with the non-linear equations of the
existing model. The results showed that GA is quite efficient in optimisation. The current work is
pinned on the use of evolutionary genetic algorithm (EGA) in MATLAB. The aim of the paper is to
determine the most economical design variable for the construction of reinforced concrete flat slabs.

2. Methodology
The study was segmented into two stages. The first stage involved the determination of the design
constraints and cost/objective functions, while the second stage involves the preparation and
implementation of the suitable genetic algorithm to optimize the cost function for the construction of a
reinforced concrete flat slab using MATLAB. The design variables are the various variables which are

2
expected to be optimized for the objective function. For this project, a mono-variable of the thickness
(h) of the flat slab was chosen. This was designated as 𝑥1 to obtain the optimum range of the thickness
using the genetic algorithm optimisation process. The aim of the research is to determine the most
economical design variable for the construction of reinforced concrete flat slabs. This follows that the
objective function has to be a function that will estimate the total cost for the design of reinforced
concrete flat slabs. This function depends on the cost of concrete, formwork and reinforcement. For
the flat slab, this was further divided into the cost of the middle strip and the cost of the column strip.
The resulting moment on both strips of the slabs will be different due to the difference in the moment
coefficients as expressed in the BS8110 part 1. This therefore results the problem to a multi-objective
optimisation problem. The positions of the middle and column strips are shown in Figure 1. This was
evaluated as follows.
𝐶𝑐𝑜𝑙 = 𝑄𝑐 𝐶𝑐′ − 𝑊𝑠 𝐶𝑟 ′ + 𝐴𝑓 𝐶𝑓 ′ (1)

′
𝑙2 𝑙2 𝑙2
𝐶𝑐𝑜𝑙 = 𝑥1 𝑙1 ( ) 𝐶𝑐′ − γ(𝑥1 ) ( ) (𝑙1 )(𝜌𝑠 )𝐶𝑟 ′ + ( ) (𝑙1 )𝐶𝑓 (2)
2 2 2
′
𝑙2 ′ 𝑙2 𝑙2
𝐶𝑚𝑖𝑑 = 𝑥1 𝑙1 ( ) 𝐶𝑐 − γ(𝑥1 ) ( ) (𝑙1 )(𝜌𝑠 )𝐶𝑟 ′ + ( ) (𝑙1 )𝐶𝑓 (3)
2 2 2

Figure 1. Positions of middle and column strips

2.1 Concrete
The popular concrete mix of 1:2:4 was taken into consideration which costs N35, 000 for a cubic
metre of concrete in Nigeria.
Unit weight of concrete cost, 𝐶𝑐′ = 35
𝑙2
𝑐𝑜𝑠𝑡 𝑜𝑓 𝑐𝑜𝑛𝑐𝑟𝑒𝑡𝑒 = 𝑥1 𝑙1 ( )𝐶𝑐 ′
2
2.2 Formwork
The cost of the formwork is dependent on the total surface area covered by the formwork so as to
support the concrete. The cost of a unit area of formwork was determined to be N9, 000.
Cost of unit area of formwork, 𝐶𝑓′ = 9
𝑙 ′
Cost of formwork = ( 2 )(𝑙1 )𝐶𝑓
2

3
2.3 Reinforcement
The cost of reinforcement depends on the weight of reinforcement required for the construction. The
percentage of reinforcement for the slab was varied from 0% to the maximum percentage of 4%. The
weight of the resulting reinforcement is calculated by first determining the area of the reinforcement.
This is followed by estimating the volume by multiplying the area with the span. The weight was then
established by multiplying the volume with the reinforcement density φ.
The cost of a tonnage of reinforcement was obtained to be N240, 000.
Cost of unit area of reinforcement, 𝐶𝑟′ = 240
𝑙
Cost of reinforcement = 𝑝ℎ𝑟(𝑥1 )( 2 )(𝑙1 )(𝜌𝑠 )𝐶𝑟 ′
2

2.4 Design constraints

The objective functions are subjected to the various constraints for the design of reinforced concrete
flat slab according to BS8110 part 1 [9] as discussed below.

2.5 Minimum depth

The minimum thickness of a reinforced concrete flat slab should be taken as described in section 3.7.1
of BS8110 part 1 which states that in no case should the thickness of the slab be less than 125mm.
𝑔1 = 0.125 − 𝑥1 ≤ 0 (4)

2.6 Serviceability constraints

The minimum reinforcement area for steel as specified by table 3.25 of BS 8110, which
specifies 0.13%𝑏ℎ.
γ ≥ 0.13 (5)
𝑔2 = 0.13 − γ ≤ 0 (6)
The maximum reinforcement area for steel as specified by table 3.12.6 of the BS 8110, which
specifies 4%𝑏ℎ.
γ ≤ 4.0 (7)
𝑔3 = γ − 4 ≤ 0 (8)

2.7 Flexural constraints

The design moments are distributed across each flat slab panel. All nominal flexural strength at the
middle (positive), and two ends (negative) of the middle strips, Mn, should be greater than the ultimate
design moment on the middle strips (mid) at each of these parts of the slab Mu- and Mu+.
𝑀𝑢 − ≤ 𝑀𝑛 − (9)
𝑔4 = 𝑀𝑢 − − 𝑀𝑛 − ≤ 0 (10)
𝑀𝑢 + ≤ 𝑀𝑛 + (11)
𝑔5 = 𝑀𝑢 + − 𝑀𝑛 + ≤ 0 (12)
In the equations above, Mu is calculated as described below using the simplified BS 8110 part 1
approach which requires that the conditions stated in section 3.7.2.7 of the code is satisfied.
𝑀𝑢 = 𝑘𝑓 𝛽𝐹𝐿 (13)
Where kf and β are coefficients obtained from tables 3.18 and 3.12 of BS 8110 part 1
respectively.
The nominal moments resistance of the slab can also be obtained as described below
𝑀𝑛 = γ(𝑥1 )(𝑙2 )𝑓𝑦 𝑧 (14)
Where z is the lever arm for the strips of the slabs and is obtained as follows
𝑧 = 𝑑 (0.5 + √0.25 − 𝑘⁄0.9) ≤ 0.95𝑑 (15)

4
 𝑀𝑢
𝑘= (16)
𝑙2 (𝑥1 )2 𝑓𝑐𝑢
𝑘≤
𝑘 ′ (0.156) (17)
𝑔6 = 𝑘 − 0.156 ≤ 0 (18)

2.8 Constant parameters

These are the final requirements for the optimisation problem. They are the given parameters that are
not decision variables. For the purpose of this study, the various parameters and values were taken as
described in the Table-1 below.

Table 1. Constant parameters

Parameter Value
l1 4.0 m
l2 4.2 m
Load on slab 14.0 kN/m2
Fcu 30.0 N/mm2
Fy 460.0N/mm2

3. Results and Discussion

The MATLAB built-in genetic algorithm comes with different plot functions to explore the fitness of
the design variable. This plot functions were included to show the variations of the fitness of cost
function at various reinforcement ratio. The value of the reinforcement ratio, γ was incremented by
0.5% between 0 and 4%. The software outputs for various reinforcement ratios are displayed for the
middle and the column strips in the following sections.

3.1 Middle strip optimisation

Figure 2 shows the plots of various optimisation outputs such as the best fit of the objective function,
the range of score of variable population and the optimal depth for the slab for the middle strip design
with 0% reinforcement ratio. It is evident from the plots that the score of the variables is zero even at
optimal depth of 130mm. No single member of the variable population shows any sign of optimisation
because the plain concrete cannot survive under the loading in tension.

Figure 2. Plots of optimisation statistics for 0% reinforcement

5
 Figure 3 shows the same optimisation statistical outputs for 1% reinforcement. It can be seen
clearly that the score for the variable populations has jumped from zero to 1.1778 × 105 . The various
score range distribution is a fairly even with the highest number of individual having the same score
value being 20. The optimisation statistical outputs for 1.5% is shown in Figure 4. At this
reinforcement percentage, the score for the variable populations moved from the 1.1778 × 105
obtained from the 1% reinforcement percentage to 1.2192 × 105 . Here, the range of the scores is well
distributed among four values with the highest number of individuals being 33. The same optimisation
statistical outputs for 2% reinforcement is shown in Figure 5 below. Here, the score range distribution
shows very little variation with 48 individuals of the 50 population converging. The score range value
was also recorded to be 1.2606 × 105 .

Figure 3. Plots of optimisation statistics for 1% reinforcement

Figure 4. Plots of optimisation statistics for 1.5% reinforcement

Figure 6 shows the statistical outputs for the 2.5% reinforcement. Here, there was a 4.14 × 103
increase in the score value from the previous 2% reinforcement percentage to give a 1.3020 × 105
score. The score range value here was distributed between two score ranges with this highest number

6
of individual at a particular score being 26. Figure 7 shows the optimisation statistical outputs for 3%
reinforcement. The score for the variable population varies and reduces progressively, and the highest
number of individuals with the same score being 29 and the least being 2. The score for the variable
population also increased by the same value of 4.14 × 103 increase recorded for the change in
percentage between 2% and 2.5% percentage reinforcement.
Figure 8 shows the optimisation statistical outputs for 3.5% reinforcement. There was an increase in
the score to a value of 1.4263 × 105 . The score range for the variable population were majorly
dominated by two score values taking about 20 and 26 number of individuals. Figure 9 shows the
optimisation statistical outputs for 4% reinforcement. There was an increase in the score to a value
of 1.4263 × 105 . The score range for the variable population converges to two scores with 26 and 20
number of individuals converging to similar scores.

Figure 5. Plots of optimisation statistics for 2% reinforcement

Figure 6. Plots of optimisation statistics for 2.5% reinforcement

7
Figure 7. Plots of optimisation statistics for 3% reinforcement

Figure 8. Plots of optimisation statistics for 3.5% reinforcement

8
 Figure 9. Plots of optimisation statistics for 4% reinforcement

3.2 Column strip optimisation

Figure 10 shows the plots of various optimisation outputs such as the best fit of the objective function,
the range of score of variable population and the optimal depth for the slab for the column strip design
with 0% reinforcement ratio. It is evident from the plots that the score of the variables is zero even at
the optimal depth of 140mm. No single member of the variable population shows any sign of
optimisation because the plain concrete cannot survive under the loading in tension. Figure 11 shows
the same optimisation statistical outputs for 1% reinforcement. It can be seen clearly that the score for
the variable populations has jumped from zero to 1.2908 × 105 . The various score range distribution
is a fairly reducing with the highest number of individual having the same score value being 27. The
optimisation statistical outputs for 1.5% is shown in Figure 12. At this reinforcement ratio, the score
for the variable populations moved from the 1.2908 × 105 obtained from the 1% reinforcement
percentage to 1.3361 × 105 resulting in a 4.53 × 103 difference. Here, the range of the scores is
concentrated between two scores with the number of individuals for each being 28 and 19.
The same optimisation statistical outputs for 2% reinforcement is as shown in Figure 13 below.
Here, the score range distribution shows that the score for the variable population was centred around
two score with the highest number of individual with the same score being 32. The score range value
was also recorded to be 1.2606 × 105 . Figure 14 shows the statistical outputs for the 2.5%
reinforcement. Here, there was a 4.54 × 103 increase in the score value from the previous 2%
reinforcement percentage to give a 1.4269 × 105 score. The score range value here was distributed
between two score ranges with this highest number of individual at a particular score being 24 and 20
being the next. Figure 15 shows the optimisation statistical outputs for 3% reinforcement. The score
for the variable population varies and reduces progressively, and the highest number of individuals
with the same score being 25 and the least being 1. This is observed to be similar to that obtained for
the same percentage in the middle strip. The score for the variable population also increased by the
same value of 4.54 × 103 increase recorded for the change in percentage between 2% and 2.5%
percentage reinforcement.

9
Figure 10. Plots of optimisation statistics for 0% reinforcement (column strip)

Figure 11. Plots of optimisation statistics for 1% reinforcement (column strip)

10
Figure 12. Plots of optimisation statistics for 1.5% reinforcement (column strip)

Figure 13. Plots of optimisation statistics for 2% reinforcement (column strip)

11
 Figure 14. Plots of optimisation statistics for 2.5% reinforcement (column strip)

Figure 15. Plots of optimisation statistics for 3% reinforcement (column strip)

Figure 16 shows the optimisation statistical outputs for 3.5% reinforcement. There was an increase
in the score to a value of 1.5177 × 105 . The score range for the variable population was majorly
dominated by a single score value with 38 number of individuals sharing the very same score. Figure
17 shows the optimisation statistical outputs for 4% reinforcement. There was an increase in the score
to a value of 1.5631 × 105 . The score range for the variable population varies, with the highest
number of individuals sharing the same score being 31.

12
 Figure 16. Plots of optimisation statistics for 3.5% reinforcement (column strip)

Figure 17. Plots of optimisation statistics for 4% reinforcement (column strip)

4. Conclusions
From the middle strip optimisation statistical outputs, it can be deduced that the optimum thickness of
the slab is 130mm as generated by the software. From the various reinforcement ratio, the 2% ratio
gives rise to almost all the variable population scoring the same value. This indicates that the for the
cost optimum design of the reinforced concrete flat slab, Engineers should endeavour to keep the
thickness of the middle strip around 130mm and the reinforcement ratio should be kept below 2%.
The column strip optimisation as expected gives a slightly higher optimum thickness of 140mm.
This can be linked to the higher percentage of moment shared with the middle strip. Furthermore, the
reinforcement ratio at 3.5% gives the least variation with about 80% of the variable population
converging to a single value. Hence, it can be further recommended that Engineers try to keep the
thickness of the column strip around 140mm and reinforcement ratio kept below 3.5%.
The current work is not exhaustive enough because the objective functions for both the middle and
column strips are considered individually. It will be interesting to explore the interaction of these

13
objective functions as a multi-objective constrained optimisation whereby the Pareto plot could reveal
the point of intersection of the two objective functions.

References
[1] Oyenuga V O 2008 Simplified reinforced concrete design (Lagos: Asros).
[2] Mosley W H and Bungey J H 1987 Reinforced concrete design (London: Macmillan).
[3] Coello C C, Hernandez F S and Farrera F A 1997 Optimal design of reinforced concrete beams
using genetic algorithms Expert Systems with Applications 12(1) 101-108.
[4] Sivanandam S N and Deepa S N 2008 Introduction to Genetic Algorithms Verlag Berlin
Heidelberg. New York.
[5] Galeb A C and Atiyah Z F 2011 Optimum design of reinforced concrete waffle slabs.
International Journal of Civil and Structural Engineering 1(4) 862–880.
[6] Matthias M, Sven S, Svetlana V, Tine T and Rajan F C 2015 Multi-objective weight and cost
optimization of hybrid composite-concrete beams Composite Structures 134 369-377.
[7] Ghandi E, Shokrollahi N and Nasrolahi M 2017 Optimum cost design of reinforced concrete
slabs using cuckoo search optimization algorithm International Journal of Optimization in Civil
Engineering 7(4) 539-564.
[8] Sudarshana H R and Ramesh B 2009 Optimized column design using genetic algorithm based
neural network International Journal of Engineering and Manufacturing Science 13 503-511.
[9] BSI (British Standards Institution) 1997 Structural use of concrete: code of practice for design
and construction. BS 8110: Part 1. London, UK.

14