Professional Documents
Culture Documents
2003 11thICCC Paper PDF
2003 11thICCC Paper PDF
ABSTRACT
This paper presents a method for optimizing concrete mixture proportions according to the required
performance and also presents several case studies. Since various qualities are usually required of
concrete, we characterize the proportioning problems as multi-criteria optimization problems. We
dealt with the notion of Pareto optimality to derive the optimum solution and applied it to a genetic
algorithm. In this contribution, two proportioning problems are solved by the genetic algorithm: a
request of cost-performance at 60 MPa in strength as a simple problem, and a concrete mix for mass
concrete in cold weather at the coast.
1. INTRODUCTION
Proportioning problems cannot be solved by the usual methods that consist of searching for the best
solution with a single objective function, such as linear programming problems and nonlinear
programming problems. The reason for this impossibility is that various performances are required
of concrete corresponding to the environments in which the concrete is used, and it is impossible to
express the plural requests in a single objective function. It follows from the features of the
proportioning problem mentioned above that a proportioning problem is considered to be a multi-
criteria optimization problem.
Proceedings of the 11th International Congress on the Chemistry of Cement (ICCC) 11 - 16 May 2003, Durban, South Africa
‘Cement’s Contribution to the Development in the 21st Century’ ISBN Number: 0-9584085-8-0
Hosted by: The Cement and Concrete Institute of South Africa Editors: Dr G. Grieve and G. Owens
CD-ROM produced by: Document Transformation Technologies Congress Organised by: Event Dynamics
1921
In GA “genotype” is the parameter set, which means a vector in the set of optimization problem,
coded as a finite-length string with binary digits. The simple string is called “individual”. To apply
GA to a proportioning problem, the genotype is designed to represent various components and
various mixture proportions (see Fig.1). The genotype consists of two parts. One is a coded binary
string and the other is a database containing components such as cement, aggregate, admixtures,
and so on. As demonstrated in Figure 1, the string has linkage parts and volumetric ratio parts. The
linkage parts have a connection to the database, which is used when the fitness is calculated with
volumetric ratios. The volumetric ratio parts show the volumes of components in concrete as
component/water ratios. In reality, the binary string is fixed as 256 bytes whereas it is not shown in
Figure 1.
The algorithms make a “phenotype”, which demonstrates characteristic form and quality in the
designed system, from each individual according to its genotype. Each designed genotype has its
phenotype. In our system, the term phenotype means the concrete properties and performance that
are estimated from mixture proportions coded in the genotype. The kinds of concrete properties and
performance are also shown in Figure 1. Prediction formulas are formulated statistically for each
property.
“Fitness value” gives the numerical evaluation of each individual and each phenotype in the
designed system. According to the fitness values, which can be plural in Pareto optimality problem,
an individual in the “population” representing the set of individuals will be reproduced with
crossover and mutation from generation to generation. Thus the individuals in the set become fitted
under applied environment. In the developed algorithm, suitable functions are designed to meet the
way of performance requesting and those are shown in Figure 2 together with the equations. U is
the parameter representing the required performance of concrete mixture and T is the parameters
1922
that determine shape of function. These two parameters express the allowable range of requirements
for performance.
When a vector x is partially less than y and all criteria are minimizing criteria, the mathematical
expression of the vector x in the Pareto optimal set is:
Under the condition of equation (1), point x “dominates” point y. If the point is not dominated by
any other, that point is “non-inferior”. The set of these non-inferior points is what we call a Pareto
optimal set. According to this definition, if there is a point, which is not less than any other points
by all criteria, only the best point will get a good evaluation. If there is no such a point, a set of
non-inferior points, which trade off one of the set of points for another, will be evaluated as good.
To apply genetic algorithms to a multi-criteria problem, an algorithm that can derive the Pareto
optimal set is developed. Goldberg [3] had designed similar algorithms. Improvement on
Goldberg’s algorithms and application to the proportioning problem are conducted. The developed
program applied to concrete mixture problem is named MixGA. The program-flow of MixGA is
detailed as follows.
1923
5. After getting N individuals in the new population (child generation) and N individuals in
the old population (parent generation).
6. Produce a temporary generation with the child and parent generations.
7. Mutate genes by reversing the number at certain loci arbitrarily with a constant probability
of 1%. (Loci is the plural word of locus. Locus means the position of the gene.)
8. Select Pareto individuals from the temporary generation and make the next generation that
consists of N individuals. Reduction rule of population is dominated by two roles below.
(i) If the number of Pareto individuals is less than N, then preserve all the Pareto
individuals. Up to the number of all individuals becomes N, select individuals one by one
from the rest with the criteria No. 1 to No. P. The probability of selection is in proportion
to their fitness value. This method intends that the good genes in the remainder should be
carried on in the next generation.
(ii) On the other hand, if the number of Pareto individuals is N or more, select individuals
from the Pareto individuals according to criteria No. 1 to No. P. The probability of
selection is in proportion to their fitness in order.
9. Iterate steps (2) to (8) until the given number of times.
10. Conduct final selection.
The algorithm detailed above is characterized by the evolution of population with highly evaluated
genes and Pareto optimal conditions. The highly evaluated individual has a good feature in a part of
the binary string. And these parts of a binary string in individuals are inherited to next generation
explicitly. An individual in the next generation inherits several good parts of a binary string that can
be highly evaluated through phenotype. This process of evolution makes it possible to find out the
optimal concrete mixture among an almost infinite number of combinations of materials and
proportions; otherwise the process can be a random search and that will fail.
But it should be noted, however, that a Pareto optimum set is a set of non-inferior points. This
definition implies it is possible that the evolved population has individuals, which have an
outstanding performance by a certain performance criterion in exchange for bad fitness values of
the others. The Pareto set conditions do not provide a single exact solution but help to search for an
optimal set widely. Because of this undesirable aspect of Pareto optimality, the concept of “meta-
property” is used. In MixGA the average of fitness values of target properties is used as one of
properties and this kind of property of individual is called meta-property. With the meta-property
concept it is possible to inherit another possibility in gene that shows balancing in target properties
of concrete to the next generation. Besides a final process of evaluation, a final selection is
conducted using the fitness values and eliminating the individuals that do not meet the required
performance.
3. TRIAL OF MIXGA
The followings are two case studies, in which the mixture proportion of concrete is optimized under
required conditions by using MixGA.
1924
formulating is partially based on the theoretical approach and partially based on the statistical
approach. The formulas used are shown below.
Strength of mortar:
Fm = (a ( B / W ) +b) ⋅ K
(2)
where Fm denotes the strength of mortar, a, b and K denote material parameter depending on
cement, W denotes the water content per unit weight [kg/m3], and B denotes the summation of
cement and admix content per unit weight [kg/m3]. In this contribution 4 kinds of cement are
formulated; namely Ordinary Portland Cement (OPC), Low Heat Cement (LHC), Moderate Heat
Cement (MHC) and High Early Strength Cement (HESC).
where rg denotes the effect of coarse aggregate on strength, d, e and f denote material parameter
depending on aggregate, V denotes the volume of coarse aggregate per unit weight, X denotes the
maximum size of coarse aggregate and A denotes the minimum size of aggregate that is able to
affect the loss of strength.
1 X
X < 0.3 : rmix = 1 + ( g ( X 2 − h)3/ 2 ) ⋅ + ( lX 2 + m ) ×
X1 + 1 (6)
0.3
X − 0.3
X ≥ 0.3 : r0.3 × ( a1 X 2 + b1 ) ×
0.4 (7)
where rmix denotes the effect of admixture on strength, g, h, l, m, a1, b1 denote the material
parameter, X denotes the replacement ratio, X1 denotes the water to binder ratio, X2 denotes specific
surface area of slag and r0.3 denotes the value calculated by equation (6) with X of 0.3.
1925
Effect of addition of Fly-Ash (FA):
where rmix denotes the effect of admix on strength, a2 and b2 denote the material parameter, X
denotes the replacement ratio and X1 denotes the water to binder ratio.
where rmix denotes the effect of admix on strength, X denotes the replacement ratio.
Strength of Concrete:
Fc = Fm × rg × rg 2 × rair × rmix
(12)
− ( K1V + K 2 )
V
η p = ηw 1 −
c (13)
where η p denotes plastic viscosity of cement paste [Pa s], ηw denotes plastic viscosity of water,
V denotes volumetric density of cement paste, c denotes solid content of cement, K1,K2 denote
constants affected by the properties of the cement paste. –17.5 and 12.0 are used respectively.
Yield value :
τ p = aη p + b
(14)
where τ p denotes yield value of cement paste [Pa], a and b denote constant parameters. 15.505 and
1.244 are used respectively
1926
Temperature effect on the rheological parameters:
η pT = 0.00387η p 20 ⋅ T + η p 20
(15)
τ pT = 3.03τ p 20 ⋅ T + τ p 20
(16)
where η p 20 denotes plastic viscosity of cement paste at 20 oC, τ p 20 denotes yield value of cement
paste at 20 oC, T denotes temperature, η pT denotes plastic viscosity of cement paste at T oC and
τ pT denotes yield value of cement paste at T oC.
radmix ,η = fi ( X )
(17)
radmix ,τ = Gi ( X )
(18)
where radmix ,η and radmix ,τ are the coefficient of viscosity and yield value depending on the
admixtures and X is the admixture to cement ratio. In this contribution, 4 type of admixtures are
formulated; namely High Range Water Reducer (HRWR), AE agent (AE), Accerelator (AC) and
Retarder (RE)
radd ,η = f j ( X )
(19)
radd ,τ = G j ( X )
(20)
where radd ,η and radd ,τ are the coefficient of viscosity and yield value depending on the addition and
X denotes the replacement ratio.
where Z denotes solid content of combined fine aggregate and coarse aggregate, Aij is coefficient
representing the packing performance of combined aggregates of two different diameter, Xi and Xj
express volumetric ratios of the aggregates at the representative sieve sizes.
1927
Pe = 1 − Va / Z
(22)
Pe
Γ= n (23)
∑n s D
i
i i pi
where Pe denotes volume of excess paste, Va is volume of aggregate, Γ denotes relative thickness of
excess paste, ni is number of aggregate of size i, si denotes surface area of each aggregate of size
i, Dpi is diameter of aggregate of size i.
Using relative thickness of excess paste and rheological parameters of cement paste, the viscosity
and yield value of concrete can be calculated.
ηc = η p × (1 + 0.0705Γ −1.69 )
(23)
τ c = τ p × (1 + 0.0705Γ −2.22 )
(24)
where ηc denotes plastic viscosity of fresh concrete and τ c denotes yield value of fresh concrete
Cone slump:
Using the experimental data of slump predicted rheological values, the regression curve is
formulated as the function of yield value and parameter of mix proportion as below:
1
S = a3 × Ln(τ c ) × + b3
1+W / B (25)
3.1.3 Costs
In this problem, the function of cost is simply the summation of material cost.
Q∞ = a4 B + b4
(27)
γ = a5 B + b5
(28)
where Q denotes the adiabatic temperature rise at t day, a4, b4, a5 and b5 denote the material
parameters and B denotes the weight of binder per unit volume.
1928
3.3 Results
The required parameters of concrete in both cases are summarised in the same Table 1.
Case 1:
After running the MixGA with 120 individuals and 100 generations, mix proportions shown in
Tables 2 are derived regarding with case 1. As final selection, elimination of the individuals, which
do not meet required performance of strength and slump simultaneously, is conducted. Additionally
in the final selection, as an index of cost performance, a value of strength divided by cost is used.
Three mix proportions listed in Table 2 are top-to-third in strength-cost index. In left of Figure 3
shows the degree of conformity of three mix proportions listed in Table 2. Among three
proportions, mix A has the best strength-cost index. It should be noted that, as is shown in left graph
of Figure 3, mix C is much cheaper than mix A. The value of strength of mix A is higher than that
of mix C. But in MixGA both mix A and Mix C is evaluated equally and they survive selections
from generation to generation.
A 0.42 138 275 (OPC) 52 (GGBS) 936 (CR) 987 (GR) 0.000 0.000
Case 1 B 0.43 145 264 (HESC) 76 (GGBS) 967 (CR) 925 (GR) 0.232 (AC) 0.006 (AE)
C 0.41 145 301 (MHC) 52 (GGBS) 962 (CR) 921 (CR) 0.000 0.048 (AE)
D 0.47 132 280 (LHC) 0 1035 (CR) 898 (CR) 0.115 (HRWR) 0.512 (AE)
E 0.47 132 281 (LHC) 0 1029 (CR) 893 (CR) 0.005 (HRWR) 0.526 (AE)
Case 2
F 0.47 131 281 (LHC) 0 1028 (CR) 906 (CR) 0.195 (HRWR) 0.512 (AE)
G 0.48 132 273 (MHC) 0 1034 (CR) 898 (CR) 0.0215 (HRWR) 0.526 (AE)
1929
Case 2:
After running the MixGA with 140 individuals and 200 generations with meta-property concept,
mix proportions shown in Tables 2 are derived regarding with case 2. As final selection, elimination
of the individuals, which do not meet required performance of strength, slump, drying shrinkage,
chlorides diffusion coefficient, durability factor and adiabatic temperature simultaneously, is
conducted. In right of Figure 3 shows the degree of conformity of four mix proportions listed in
Table 2. As shown in Figure 3, almost all the mixtures meet the required performance. It should be
noted here that mix E has low value in slump and this mix is evaluated equally from the Pareto
optimal point of view. If only the Pareto optimization is used, it is possible that such undesirable
individual that do not have good value in all several properties is able to survive the selection. The
genetic algorithm with the notion of Pareto optimality has a good potential in initial convergence
tendency but it has a risk of undesirable acceleration in a few properties as well. With meta property
concept, which disappreciate the prominence and unbalancing it is possible to derive the sufficient
set of mixture proportions. Using the suitable fitness function and evaluation method is important to
get sufficient set.
4. CONCLUSIONS
1. A genetic algorithm system integrating the concept of Pareto optimality, which is named
MixGA, was developed for solving multicriteria optimization problems in concrete mix
proportioning.
2. As shown in the examples presented in this study, MixGA can derive the appropriate mix
proportions from the vast combinations of sorts of content and proportions of mixture to
explore. This system is maintained by suitable fitness evaluation, reasonable reproduction
and correct prediction formulas.
3. The genetic algorithm with the notion of Pareto optimality has a good potential in initial
convergence tendency but it has a risk of undesirable acceleration in a few properties as
well. Using meta property concept, it is possible to compensate for this risk of the notion of
Pareto optimality.
REFERENCES
[1] Marks, W. and Potrzebowski, J., Multicriteria optimization of structual concretemixes, Architecture and Civil
Engineering 38(4), 1992, pp.77-01
[2] Piasta, Z. and Czarneski, L., Analysis of material efficiency of resin concrete. in Brittle Matrix Composite,
Elsevier Apllied Science London and New York, 1989, pp.593-602
[3] Goldberg, A. E., Genetic Algorithms in Search Optimization & Machine Learning, Addison Wesley, 1989,
pp.192-208
[4] Kikukawa, H., Studies on viscosity equation of Portland cement paste, Journal of Material, Concrete Structure
and Pavements, V-2, 354, 1985, pp.109-118 (In Japanese)
[5] Roscoe, R., The Viscosity of Suspension of Rigid Spheres, British Jouranl of Applied Physics, Vol.3, 1952,
pp.267-269
[6] Ooi, T., Comparted bulk density of aggregate with random shape and widely ranged particle size distribution,
Journal of Structure and Construction Engineering. AIJ, No.423, 1991, pp.11-16 (In Japanese)
[7] Oh, S. G.., Noguchi, T., and Tomosawa, F., Toward Mix Design for Rheology of Self-Compacting Concrete, 1st
International RILEM Symposium on Self-Compacting Concrete, Stockholm, 1999, pp.361-372
[8] Kennedy, C.T., The Design of Concrete Mixes, Proceedings of the American Concrete Institute, Vol.36, 1940,
pp373-400
1930