Optimization of The Rheological Properties of Self Compacting Concrete Using Neural Network and Genetic Algorithm

8th IEEE International Conference Humanoid, Nanotechnology, Information Technology
Communication and Control, Environment and Management (HNICEM)

The Institute of Electrical and Electronics Engineers Inc. (IEEE) – Philippine Section
9-12 December 2015 Water Front Hotel, Cebu, Philippines
OPTIMIZATION OF THE RHEOLOGICAL PROPERTIES OF SELF

COMPACTING CONCRETE USING NEURAL NETWORK AND
GENETIC ALGORITHM
Nolan C. CONCHA Elmer P. DADIOS
CE Department MEM Department
FEU-Institute of Tech, Manila, Philippines De La Salle University, Manila, Philippines
[email protected] [email protected]
Abstract- Self compacting concrete (SSC) is one of the A well considered concrete technology that has been
most useful innovations in concrete technology that has the constantly developed within the last two decades is self
ability to flow efficiently and maintain material compacting concrete (SCC). SCC can be defined as a new
homogeneity. However, additives particularly admixtures type of concrete with very high workability and stability that
introduced in the production of SSC to enhance some needs little to no form of external compaction [1]. These two
specific properties of fresh and hardened concrete may parameters allow the concrete to fill every corner of the
contribute undesirable effects on the workability formwork to which it is placed under its own weight without
performance. In this study, superplasticizers blended with the need of any external vibrator. Another property that is
fly ash was used in the mix and were tested for Slump always needed for SCC is its passing ability that is needed for
Flow, L-Box, and Screen Stability tests to determine its filling heavily congested sections of reinforced concrete
influence on the rheological properties of SCC. Several structures [2]. The European Federation of National Trade
mixtures were tested in order to derive a mix proportion Associations (EFNARC) dedicated to special construction
having the optimum rheological properties. Artificial chemicals and concrete system, publishes the set of standards
neural network and genetic algorithm were used to describing the requirements of SCC [3]. EFNARC lists the
determine the concrete mix proportion that will provide three workability requirements of SCC as its filling ability,
the best workability. Results showed that ANN was able to passing ability and segregation resistance and there is no
establish the relationship of rheology to the concrete single test method that can measure the abovementioned
material components and GA derived the optimum workability parameters. To enhance further the performance
proportion for best rheological performance. of concrete, admixtures are introduced immediately before or
during mixing of concrete batch. Admixtures when properly
Index Terms - self compacting concrete, admixtures, used offers beneficial effects to concrete, including improved
rheology, dispersibility, artificial neural network, genetic quality, acceleration or retardation of setting time, enhanced
algorithm frost and sulfate resistance, control of strength development,
improved workability, and enhanced finishability [4].
Admixtures vary widely in chemical composition, and many
I. INTRODUCTION perform more than one function. They are added to concrete in
very small amounts and should meet standard specification
Nowadays, researchers whisk to technological requirements. The chemistry of concrete admixtures is
advancement, exploring scientific outbreaks that will complex requiring in-depth knowledge and experience. A
ultimately benefit humanity. One of the immensely focused general understanding of the options available for concrete
studies is on concrete modification for high workability and admixtures is necessary for acquiring the right product for the
performance. To ensure the viability of these innovations, job, based on climatic conditions and job requirements [5].
comprehensive studies have been conducted. The basic However such chemical intervention in the resulting mixture
composition of concrete has been continually enhanced and may introduce a considerable reaction on the basic
several chemical composites were introduced to achieve a composition of concrete, which will eventually contribute to
more desirable concrete mix. some adverse effects on its physical and chemical properties
[6]. The presence of admixtures on self-compacting concrete
can inadvertently influence its rheological properties [7].
978-1-5090-0360-0/15/$31.00 ©2015 IEEE
Authorized licensed use limited to: University of Nantes. Downloaded on March 30,2023 at 11:25:11 UTC from IEEE Xplore. Restrictions apply.
Rheological characteristics of concrete may be affected II. FRAMEWORK OF THE STUDY

significantly by these chemicals and modify its flow behavior
during mixing, hence a sound understanding on the A data set of 60 different mixes of concrete materials will
contribution of admixtures on the intrinsic flow properties of be prepared and tested for slump flow, L-Box, and screen
concrete is a vital requirement to develop the overall stability tests. As shown in figure 1, the results of these tests
performance of concrete. will be used as training, validation, and test data to develop a
model using neural network describing the relationship of
Artificial Neural Network (ANN) is a family of statistical concrete materials to the rheological behavior of the resulting
learning algorithms. There are many different types of ANN concrete mix.
and the simplest and most commonly used is the feedforward,
multilayered, supervised neural network with the error
backpropagation algorithm, also known as the back Weights
propagation (BP) network [8]. ANN shown in figure 1, needs
to learn or be trained from an existing training set containing Inputs
pairs of input–output elements that in turn form a system of Neurons
interconnected neurons which can calculate values from Output
inputs. Genetic algorithm on the other hand is an iterative
procedure based on the theory of evolution which states that
Output Layer Bias
the survival of an organism is affected by rule "the strongest
species that survives" which usually operates on a population
of constant size [9]. A solution generated by genetic algorithm Input Layer Hidden Layer
is called a chromosome, while collection of chromosome is
referred as a population. A chromosome is composed of genes Figure 1: ANN Architecture
and its value can be either numerical, binary, symbols or
characters depending on the nature of the problem considered The model obtained from ANN is then utilized to develop a
[10]. These chromosomes will be assessed using a fitness fitness function that will be used in Genetic Algorithm and
function to measure the suitability of a solution generated by derive a proportion of concrete having the best rheology. The
GA with problem. Chromosomes in population will mate results from the GA will be validated by preparing actual mix
through process called crossover producing new offspring in proportions and will be compared with other mixes having
which the genes composition are the combination of their different proportions. A schematic diagram of the flow of
parent. In a generation, a few chromosomes will also undergo methods applied is shown in figure 2.
mutation in their genes. The number of chromosomes which
will undergo crossover and mutation is controlled by
crossover rate and mutation rate values. Chromosome in the
population that will maintain for the next generation will be
Testing of Samples
selected based on Darwinian evolution rule, the chromosome
which has higher fitness value will have greater probability of
being selected again in the next generation. After several
generations, the chromosome fitness will converge to a certain
value which is the best solution for the problem [11].
Rheological Modeling
Using Neural Network
In this study, several mixes were prepared and tested for
rheological tests. The results were then utilized to develop a
set of rheological models using artificial neural networks. The
model generated was optimized with the aid of genetic Model Optimization
algorithm to determine the most efficient proportions of Using Genetic Algorithm
material components that will ultimately provide the best
rheology. In the succeeding sections a thorough discussion of
the processes and algorithms involved in this research are
presented in the first few sections. The results of the
experiments and the discussion of results as well as the Experimental
conclusions and recommendations are presented in detail in Validation
the last two sections.
Figure 2: Flow Chart of the Study
III. EXPERIMENTAL PROGRAM L-box Passing Ability Test
Each locally available material composition of Self The vertical section of the L-box was filled with about 14
Compacting Concrete (SCC) were tested for structural liters of concrete and let it rest for 1 minute. During this time,
properties in accordance with the minimum requirements set a visual inspection to check the stability of concrete was
by the American Society for Testing of Materials (ASTM). performed then gate was opened to let the concrete flow into
The Portland cement undergone sieving to eliminate the horizontal section as shown in figure 4.
impurities and tested for fineness and specific gravity in
accordance with the standard set by ASTM. The specific
gravity and absorption of coarse and fine aggregates used in
the experimental program were measured. Drying and
selection of aggregates size of less than 20 mm diameter to
provide the desired passing ability were performed. Concrete
design proportions with the addition of superplasticiser were
developed to meet the minimum requirements for Self
Compacting Concrete. EFNARC Trouble shooting guide to
achieve SCC was followed and adjustments of concrete
materials was performed accordingly. Several SCC
proportions were designed based on the suggested proportions Figure 4: L-Box Test
provided by the American Concrete Institute (ACI) and that
served as initial proportions to determine the influence of When the flow of concrete has stopped moving, the height
flyash admixture once compounded with superplasticizer. Test at the end of the horizontal section (h1) and that of the
specimens with different admixture content were prepared and remaining in the vertical section were measured (h2) and
tested for workability and viscosity requirements which are expressed as a proportion.
general criteria for Self Compacting Concrete. To satisfy the h1
basic requirements for SCC, superplasticizers were mixed to L − Box Ratio = (2)
the concrete and the rheological performance of SCC blended
h2
binary with Fly ash will be tested for Slump Flow, L-Box, and
Screen Stability tests. An average ratio in the range of 0.80 to 1.0 indicates the
better passing ability of the concrete.
Determination of Slump Flow
Screen Stability Test
The slump flow test for SCC is described by EFNARC. In
this test, the truncated cone mold is placed on a smooth and Ten liters of concrete sample was used and allowed to
level plate, filled with about 6 liters of concrete, and lifted stand for 15 minutes in a bucket to allow any internal
upwards. The concrete is allowed to flow freely in all segregation to occur. The bucket was covered with a lid to
directions forming a flat circular section as shown in figure 3. prevent evaporation. The top 2-liter of the concrete in the
bucket was poured into a pouring container, and the filled
container was weighed as shown in figure 5.
Figure 3: Slump Test

Figure 5: Screen Stability Test
d1 + d2
Slump Flow = (1) Wb
2 Ratio = x100% (3)
Wa
To apply ANN and GA to this study, 60 different design Table 2: Allowable Values
mixes, samples of which are shown in table 1, were prepared
and tested for slump flow, L-box, and Screen stability tests.
ALLOWABLE VALUES
Table 1: SCC Design Mixes Test Min Max
Slump Flow, mm 550 850
Sample Materials (kg/m3) L-Box Ratio 0.8 1
ID Flyash Cement Sand Gravel SP Screen Stability,% 0 15
1 45 344 1038 650 20.0
2 35 350 938 750 20.0 The individual weights of each chromosome in the three
rheological tests will be determined using the following
3 60 355 938 750 17.0 equations. The measurements obtained in each test from the
4 65 339 1038 650 17.5 samples should fall within the specified range in order to
5 45 350 938 750 17.5 satisfy the requirements for filling ability, passing ability, and
resistance to segregation ability of SCC. For values beyond
6 65 330 838 850 17.0 the limits, the weight will be zero.
7 60 344 938 750 15.5
8 50 354 1038 650 20.0 ⎧0 if SF<550
⎪
9 60 344 838 850 20.0 ⎪⎛ 150 − 700 − SF ⎞ (4)
WSF = ⎨⎜ ⎟ if 550 ≤ SF ≤ 850
10 50 354 1038 650 17.0 ⎪⎝ 150 ⎠
⎪0 if SF>850
⎩
The results obtained were used as training, validation, and
test data in ANN to generate a constitutive relationship ⎧0 if LB < 0.80
between rheological parameters and concrete material ⎪ (5)
WLB = ⎨ 0.2 − (1 − LB )
components. This equation in turn was used to determine the ⎪ if LB ≥ 0.80
fitness of each chromosome in GA which eventually derived ⎩ 0.20
the optimal concrete mix proportion as shown in figure 6.
⎧0 if RS>15
⎪ (6)
WSS = ⎨ (15 − SS )
⎪ if RS ≤ 15
⎩ 15
The objective function of this study is to design a concrete

proportion having the best rheological properties. The
rheology function is represented by equation (7) described by
the product of the three rheological weights that can impart a
maximum score of 1. The higher the value, the more desirable
will be the concrete mix.
Figure 6: Overall Research Process R = W SF x W LB x W SS (7)
The evaluation of the fitness value as a function of the IV. RESULTS AND DISCUSSION
slump flow diameter, L- Box ratio, and Screen stability test
was then determined. As shown in table 2 are the allowable Fly ash admixture is used to substitute a portion of cement
values in each test as per the EFNARC minimum acceptance in the concrete production without significantly affecting the
requirements for self-compacting concrete. To facilitate GA in hardened properties and typical amounts used ranges from 15-
this study, Mathlab software was used to perform the 25 percent. This additive has an important role on the
algorithm. workability of self compacting concrete and its rheological
performance will be tested on this study once compounded
with superplasticizers. The rheology function generated by
ANN provided satisfactory results. Each of the rheology
models demonstrated significantly high correlation results in
the validation, training, and test data as shown in figure 7.
The concrete material distributions obtained from GA as

shown in table 4 yielded the highest rheology. A validation of
the result was performed by preparing a mixture sample
labelled “5” of the material components distribution obtained
from GA and 3 other mixtures having different material
proportions. These mixtures were tested in slump flow, L-box,
and Screen Stability tests in accordance with the EFNARC
minimum requirements of acceptance for self-compacting
concrete.
Table 3: Optimum Proportion
CONCRETE MATERIALS (kg/m3)

Flyash Cement Sand Gravel SP
28.857 434.385 779.193 848.28 24.80
Results showed in table 4 that the optimized concrete

proportion provided the best rheology. Both the flow diameter
Figure 7: ANN Correlation Coefficients and L-box ratio provided desirable results of 705 mm and 0.95
respectively while the screen stability ratio of the sample
With the increasing amount of fly ash and superplasticizer, measured 11 %.
the slump flow and L-box values in the model displays a
substantial increase in the flow diameter of concrete showing a Table 4: Validation of Results
more flowable mixture. As the proportion of fly ash increases
with respect to cement, the mix becomes non-cohesive leading
in an increase of flow value. The screen stability test model Mix Slump Screen
L-Box Fitness
established a direct relationship with the amount of fly ash and ID Flow Stability
superplasticizer in the concrete. The computed values in the
1 627.50 0.95 14 83.45
design mixes indicate a higher chance to segregation. It can be
realized that fly ash reduced the viscosity of the concrete 2 611.00 0.83 11 72.20
registering values near the lower limit of acceptance. After 3 618.00 0.88 11 79.82
running the GA for over 100 generations as shown in figure 8,
the best design mixture was determined having a fitness value 4 605.00 0.84 10 68.85
of 0.99338. 5 705.00 0.95 11 89.95
As a general observation, the design samples exhibited

impressive and enhanced flowing ability with an increasing
amount of fly ash. This is mainly contributed by the spherical
and smooth texture of the fly ash as shown in figure 9 offering
significant roller bearing effect between the concrete
aggregates.
Figure 8: Genetic Algorithm Generations

Figure 9: Fly Ash under Scanning Electron Micrograph [12]
The excessive free water present in the mixes with high fly [6] Feng, X. and Clark, B. (2011) “Evaluation of Physical
ash content imparted considerable aggregate spacing resulting and Chemical Properties of Fly Ash Products for use
to a less particle collision and build up. The dispersibility in Portland Cement Concrete”, World Coal Ash
effect of the sup`erplasticizer also contributed to the Conference, Denver, USA.
remarkable mechanism of the workability of concrete. [7] Ozawa, K., Sakata, N., and Okamura, H. (1995)
Excessive free water production and loss in interparticle “Evaluation of Self- Compactibility of Fresh Concrete
interaction however shared less stable concrete components. Using the Funnel Test”, Concret Library of JSCE, Vol
The viscosities of the mixtures were partly reduced allowing No. 25, pp.59-75.
the samples to experience difficulty in maintaining material [8] Adeli H, Hung SL. (1995) “Machine learning––Neural
homogeneity. networks, genetic algorithms, and fuzzy systems”.New
York: Wiley
IV. CONCLUSIONS AND RECOMMENDATIONS [9] Najem R. and Yousif S. (2013) “Optimun Cost Design of
Reinforced Concrete Columns Using Genetic
ANN is a promising tool in deriving the mathematical Algorithms”, Al-Rafadain Engineering Journal
model of multivariable concrete mixture. The influence of [10] Park W. (2013) “Genetic-Algorithm-Based Mix
each material component on the rheological performance of Proportion Design Method for Recycled Aggregate
concrete was completely established. The Genetic Algorithm Concrete”, Transaction of the Canadian Society for
approach also provided the optimal design of the material Mechanov]cal Engineering, Vol 37, No 3.
distribution achieving the most desirable rheology of the [11] Wagner S., Affenzeller M., Winkler S., & Beham A.,
concrete mixture with large constraints requirements. The (2009) “Genetic Algorithms and Genetic
blending of fly ash and superplasticizer significantly improved Programming”, Taylor and Francis Group, FL, USA.
the flowability performance of the design mixes. The two [12] University of Kentucky Center for Applied Energy
admixtures complements in improving the horizontal free flow Research, November 2015.
of the concrete. The blending however imparts reduction in
the viscosity of the material exposing the concrete to
segregation. Moreover the higher the dosage of the two
admixtures, the more fluid will be the resulting SCC and the
greater will be its susceptibility to segregation.
As a result of this experimental research, the use of a
rheometer to obtain a more direct and precise yield stress and
plastic viscosity measurements can be performed. In order to
enhance the viscosity of self-compacting concrete, viscosity
modifying admixtures should be introduced in the mix. It is
also recommended to extend further this study to determine
and optimize the correlation between rheology and hardened
strength of self-compacting concrete.
REFERENCES
[1] Chia K. (2006) “Workability and Stability of

Lightweight Aggregate Concrete from Rheology
Perspective”, National University of Singapore.
[2] Ozawa, K, Okamura H. (1995) “Mix Design for Self
Compacting Concrete”, Concrete Library of JSCE.
[3] EFNARC (European Federation of National Trade
Associations). (2002) “Specification and Guidelines for
Self-Compacting Concrete”, EFNARC Association,
UK.
[4] Nawa, T., Izumi, and Edamatsu, Y. (1998) “State of the
Art Report on Materials and Design of Self
Compacting Concrete”, Proceeding of International
Workshop on SCC, Japan, pp 160-190
[5] Baskoca A., Ozkul M.H., & Artima S. (1998) “Effect of
Chemical Admixtures on Workability and Strength
Properties of Prolonged Agitated Concrete”.

Optimization of The Rheological Properties of Self Compacting Concrete Using Neural Network and Genetic Algorithm

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Optimization of The Rheological Properties of Self Compacting Concrete Using Neural Network and Genetic Algorithm

Uploaded by

Copyright:

Available Formats

8th IEEE International Conference Humanoid, Nanotechnology, Information Technology

Communication and Control, Environment and Management (HNICEM)

OPTIMIZATION OF THE RHEOLOGICAL PROPERTIES OF SELF

Rheological characteristics of concrete may be affected II. FRAMEWORK OF THE STUDY

III. EXPERIMENTAL PROGRAM L-box Passing Ability Test

Figure 3: Slump Test

The objective function of this study is to design a concrete

Figure 6: Overall Research Process R = W SF x W LB x W SS (7)

The concrete material distributions obtained from GA as

Table 3: Optimum Proportion

CONCRETE MATERIALS (kg/m3)

Results showed in table 4 that the optimized concrete

As a general observation, the design samples exhibited

Figure 8: Genetic Algorithm Generations

[1] Chia K. (2006) “Workability and Stability of

You might also like