6th World Congresses of Structural and Multidisciplinary Optimization

Rio de Janeiro, 30 May - 03 June 2005, Brazil

OPTIMIZATION OF CROSS SECTION OF REINFORCED CONCRETE BEAM USING EXPERIMENTAL DESIGN

ANA LCIA H. C. EL DEBS1, ANSELMO CHAVES NETO2, ISABELLA ANDRECZEVSKI CHAVES3, ROBERTO MAURO FLIX SQUARCIO4, SACHIKO ARAKI LIRA5
(1) Universidade de So Paulo Escola de Engenharia de So Carlos Av. Trabalhador Socarlense, 400 - Cep: 13566-590 - So Carlos (SP) Brazil [email protected] (2) PPGMNE/DEST - UFPR Centro Politcnico Caixa Postal 19081 CEP 81531-990 Curitiba (PR) - Brazil [email protected] (3) Universidade de So Paulo Escola de Engenharia de So Carlos Av.Trabalhador Socarlense, 400 - Cep: 13566-590 - So Carlos (SP) Brazil [email protected] (4) PPGMNE/CESEC UFPR Centro Politcnico Caixa Postal 19081 CEP 81531-990 Curitiba (PR) - Brazil [email protected] (5) PPGMNE/IPARDES Rua Mximo Joo Kopp, 274 - Bloco 2 - Santa Cndida CEP 82630-900 - Curitiba (PR) - Brazil [email protected] 1. ABSTRACT This paper presents the application of experimental design in optimization of project cost through minimization of dimensions of cross section of reinforced concrete beam. The methods used here are recommended for external variables optimization. In the case of cross section of beam, the external variables are the geometric dimensions of the structure and the properties of the material. Three methods of experimental design were used: One-side Gradient Design I, One-side Gradient Design II and Central Gradient Design. After the application of these methods the minimum cost of cross section of beam was gotten. The comparison between the methods was made through mean square deviations. These methods make possible the cost optimization of cross section of beam by the simple way, not using sophisticated computational programs. The algorithm used is also simple, not spending great computational effort. Hence, the experimental design can be applied with great success in the optimization of reinforced concrete structures, being a simple and versatile approach, being able also to be used for different structures, considering all the practical requirements. 2. KEY WORDS: Experimental Design, Optimization, Reinforced Concrete, Beam. 3. INTRODUCTION This work presents the application of experimental design in optimization of project cost through minimization of dimensions of cross section of reinforced concrete beam. The methods used here, presented by KRAKOVISKI [1], are recommended for external variables optimization that are related with general characteristics of the structure, the cross section of the concrete and the materials. In the external variables optimization the objective function as well as constraints are composed only by external variables. In the case of cross section of beam, the external variables are the geometric dimensions of the structure and the properties of the material. The objective function, which is the cost function of the cross section of beam, is given the according with VIANNA [2] by:
F( x ) = (b h ) Cc + (A s s ) Cs + ( b + 2h ) C f

(1)

where: b is the width of cross section of beam; h is the height of cross section of beam; AS is the steel area; S is the specific mass of the steel; CS is the cost of the steel, for unit of mass; CC is the cost of the concrete, for unit of volume; CF is the cost of the mold, for unit of area.

The objective function is composed only by external variable that is the sum of unitary costs of the steel, mold and concrete. The unitary cost of the concrete varies in accordance with resistance (fck). The following resistances of concrete were used: 15 MPa, 20 MPa, 25 MPa and 30 MPa. The unitary costs of the concrete, steel and mold were obtained in Construo e Mercado Custos, Suprimentos, Planejamento e Controle de Obras Magazine1. The unitary costs are presented in Table 1. Materials Concrete 15 MPa 20 MPa 25 MPa 30 MPa Steel Mold Table 1. Costs of Materials Unit Cm3 Kg Cm2 Cost (R$) 136,53 148,50 162,77 176,12 4,03 40,71

The external constraints are gotten from inequations about height, width and resistance of concrete, that are: 10 cm b 20 cm; 40 cm h 70 cm and 15 MPa fck 30 MPa. It was taken the constant bending moment Md = 10000 kN.cm and strength of steel CA-50A in the course of optimization. Three methods of experimental design presented by KRAKOVISKI [1] were used: One-side Gradient Design I, One-side Gradient Design II and Central Gradient Design. These methods use normalized external variables represented by x i , i = 1, 2, , n. A linear approach between the objective function and the normalized external variables is obtained by:
Y = b 0 + bi x i
i =1 n

(2)

where b0 and bi, i =1, 2, , n, are the coefficients. The coefficients are gotten in accordance with design proposed. The normalized variable is calculated by following formula:
xi = Xi Xo a

(3)

where: Xi is the dimension of the external variable to be normalized; X0 is the dimension of the reference external variable; a is the interval of external variable. The external variables are taken into account by rounding of their continuous values to nearest discrete ones. If Xi takes on the values of xi0 + ai, xi0, xi0 a, then x i is equal +1, 0, -1, respectively. In the following sections the applications of the methods and the comparison of designs will be presented. 4. EXPERIMENTAL DESIGNS The basic algorithm used to solve external optimization problem with experimental design consists of following operations: (a) Select a starting point x(k,0), k = 1; (b) At the k th step of algorithm perform the following operations: (1) Choose a design, execute the tests, if they have not been executed previously, at the points x(k,0), x(k,u) where x(k,u) is point determined by the design. (2) Using the determined values of the objective function in operation 1, obtain a linear approximation between objective function Y and normalized external variable x i .
Y = b0 + bi x i
i =1 n

where b0 and bi, i =1, 2, , n, are the coefficients. (3) Move along the antigradient of objective function Y which has to reduce the objective function and all constraints have to be satisfied. Store the point with the minimum value of the objective function through x(k,p). Do k = k + 1; p = 0 and return to operation (1). (c) Finish calculations if at some step the value of objective function Y cannot be reduced.
1

Editora Pini, March 2005.

4.1 One-side Gradient Design I In this method the coefficients b0 and bi, i =1, 2, , n, are gotten by the following formula:
b 0 = Y0
b i = Yi Y0 , i = 1, 2, ..., n

(4) (5)

A start point must be chosen (x(k,0)). The procedure (a) is defined by the design which determines the direction of the gradient. The procedure (b) determines the movements along the gradient. To get the normalized variable of procedure (b), bi is calculated as above and got the vector

as follow formula:

b Z = max bi

(6) (7) (8)

l = (b1; b 2 ; ...; b n ) / b Z = (l1, l2 , ..., ln )

x
(1, u )

= Rnd ((l1 , l 2 , ..., l n ) j , j = 1, 2, ...,

For j = 1 in the Eq.(9), it is possible to get step u = 4 of procedure (b) and so on. The point x(1,0) = (18, 58, 20) was chosen as a start point. The dimensions of external variables are: b = 18, h = 58 and fck = 20. The results, after applying the method, are presented in Table 2. The method required one step and 2 procedures to reach the minimum point. The minimum cost of cross section is R$ 94,35 with the following dimensions: width equal 10 cm, height equal 56 cm and resistance of the concrete equal 20 MPa. Table 2. The result of One-side Gradient Design I Method Normalized Variables External Variables Dimensions u b h fck x1 x2 x3 0 1 2 3 4 5 6 7 0 1 0 0 -1 -2 -3 -4 0 0 1 0 0 -1 -1 -1 0 0 0 1 0 0 0 0 18 20 18 18 16 14 12 10 58 58 60 58 58 56 56 56 20 20 20 25 20 20 20 20

k 0 1

x(k,0) -

Proc. (a) (b)

Total Cost (R$) 100,77 102,31 101,25 100,85 99,23 97,37 95,86 94,35

4.2 One-side Gradient Design II

The coefficient b0 is gotten using the Eq. (4) and bi, i =1, 2, , n, by following formula:
bi = Y0 Yi , i = 1, 2, ..., n

(9)

The following steps are the same those using in the One-side Gradient Design I method. It was chose the point x(1,0) = (18, 58, 20) as a start point, which is the same point used before. The result of One-side Gradient Design II method is presented in Table 3. The method required one step and 2 procedures to obtain the minimum point. The minimum cost of cross section is R$ 94,37 with the following dimensions: width equal 10 cm, height equal 56 cm and resistance of concrete equal 30 MPa. Table 3. The result of One-side Gradient Design II Method Normalized Variables External Variables Dimensions u b h fck x1 x2 x3 0 1 2 3 4 5 6 7 0 -1 0 0 -1 -2 -3 -4 0 0 -1 0 0 0 -1 -1 0 0 0 -1 0 1 1 2 18 16 18 18 16 14 12 10 58 58 56 58 58 58 56 56 20 20 20 15 20 25 25 30

K 0 1

x(k,0) -

Proc. (a) (b)

Total Cost (R$) 100,77 99,23 100,40 101,41 99,23 97,75 95,90 94,37

4.3 Central Gradient Design

The coefficient b0 is gotten using the Eq.(4) and bi, i =1, 2, , n, by following formula:

bi = (Y2 i Y2 i 1 ) / 2

, i = 1, 2, ..., n

(10)

It was chose the same point as the start point. The results, after using Central Gradient Design, are presented in Table 4. After one step and 2 procedures, the minimum cost of cross section equal to R$ 94,38 was gotten with the following dimensions: width equal 10 cm, height equal 56 cm and resistance of concrete equal 25 MPa. Table 4. The result of Central Gradient Design Method Normalized Variables External Variables Dimensions u b h fck x1 x2 x3 0 1 2 3 4 5 6 7 8 9 10 0 1 -1 0 0 0 0 -1 -2 -3 -4 0 0 0 1 -1 0 0 0 0 -1 -1 0 0 0 0 0 1 1 0 0 0 1 18 16 20 18 18 18 18 16 14 12 10 58 58 58 56 60 58 58 58 58 56 56 20 20 20 20 20 15 25 20 20 20 25

K 0 1

x(k,0) -

Proc. (a)

(b)

Total Cost (R$) 100,77 99,23 102,31 100,40 101,25 101,41 100,85 99,23 97,69 95,86 94,38

5. COMPARISON OF EXPERIMENTAL DESIGNS

The followings linear functions were adjusted for each one of the experimental designs:
Y1 = 100,77 + 1,54 x 1 + 0,48x 2 + 0,08x 3 (One-side Gradient Design I) Y2 = 100,77 + 1,54 x 1 + 0,37 x 2 0,64 x 3 (One-side Gradient Design II) Y3 = 100,77 + 0,43x 1 + 1,815x 2 0,28x 3 (Central Gradient Design)

(11) (12) (13)

The comparison was made through mean square deviation calculated by the following formula presented by KRAKOVISKI [1]:
2 ij j=1 27

i =

27

,
2

i = 1,2,3
, i = 1, 2, 3

(14)

where

2 ij = Yij Yj

Yj

(15)

All 27 nearest points to the starting point are considered in Table 5. The objective function values Yj are obtained directly from the objective function and the approximation values from Eq.(11), Eq.(12) and Eq.(13), represented by Y1j, Y2j and Y3j, respectively. Table 5. Comparison of Designs Objective Function (R$) Fck Y1j Y2j Yj 25 102,90 102,87 102,04 20 102,82 102,79 102,68 15 103,55 102,71 103,32 25 102,39 102,39 101,67 20 102,31 102,31 102,31 15 103,02 102,23 102,95 25 101,99 101,91 101,30 20 101,91 101,83 101,94 15 102,60 101,75 102,58 25 101,33 101,33 100,50 20 101,25 101,25 101,14 15 101,91 101,17 101,78 25 100,85 100,85 100,13 20 100,77 100,77 100,77 Desviation 2j2 69,85 1,85 4,93 49,45 0,00 0,46 45,77 0,09 0,04 67,09 1,18 1,63 50,97 0,00

x1 1 1 1 1 1 1 1 1 1 0 0 0 0 0

x2 1 1 1 0 0 0 -1 -1 -1 1 1 1 0 0

x3 1 0 -1 1 0 -1 1 0 -1 1 0 -1 1 0

b 20 20 20 20 20 20 20 20 20 18 18 18 18 18

h 60 60 60 58 58 58 56 56 56 60 60 60 58 58

Y3j 102,46 102,74 103,02 102,03 102,31 102,59 101,60 101,88 102,16 100,92 101,20 101,48 100,49 100,77

1j2 0,08 0,09 65,80 0,00 0,00 58,80 0,62 0,62 68,63 0,00 0,00 52,73 0,00 0,00

3j2 18,28 0,61 26,20 12,36 0,00 17,42 14,62 0,09 18,39 16,37 0,24 17,80 12,74 0,00

0 0 0 0 -1 -1 -1 -1 -1 -1 -1 -1 -1

0 -1 -1 -1 1 1 1 0 0 0 -1 -1 -1

-1 1 0 -1 1 0 -1 1 0 -1 1 0 -1

18 18 18 18 16 16 16 16 16 16 16 16 16

58 56 56 56 60 60 60 58 58 58 56 56 56

15 25 20 15 25 20 15 25 20 15 25 20 15

101,41 100,47 100,40 101,02 99,75 99,68 100,27 99,30 99,23 99,80 98,95 98,89 99,44

100,69 100,37 100,29 100,21 99,79 99,71 99,63 99,31 99,23 99,15 98,83 98,75 98,67

101,41 99,76 100,40 101,04 98,96 99,60 100,24 98,59 99,23 99,87 98,22 98,86 99,50

101,05 100,06 100,34 100,62 99,38 99,66 99,94 98,95 99,23 99,51 98,52 98,80 99,08

50,41 0,99 1,20 64,29 0,16 0,09 40,74 0,01 0,00 42,42 1,47 2,00 59,96

0,00 49,94 0,00 0,04 62,72 0,64 0,09 51,12 0,00 0,49 54,43 0,09 0,36

12,60 16,65 0,36 15,68 13,76 0,04 10,83 12,42 0,00 8,44 18,88 0,83 13,11

The mean square deviations obtained are: 1 = 4,35; 2 = 4,36; 3 = 3,21.

6. CONCLUSION

For the choice of the best design, it must consider two criteria: (i) the number of testes necessaries to get linear approach and (ii) the precision of the approach. One-side Gradient Design I and One-side Gradient Design II involve a minimum number of tests. To determine "n + 1" coefficients are necessaries "n + 1" tests. For the Central Gradient Design the number of tests are substantially higher and equal to "2n + 1". After the application of these methods the minimum cost of cross section of beam was gotten. The comparison of the methods was made through mean square deviation. Thus, the Central Gradient Design gives the highest accuracy of approximation that is 3 = 3,21. These methods make possible the cost optimization of cross section of beam by the simple way, not using sophisticated computational programs. The algorithm used is also simple, not spending great computational effort. Hence, the experimental design can be applied with great success in the structures optimization of reinforced concrete, being a simple and versatile approach, being able also to be used for different structures, considering all the practical requirements.
7. REFERENCES

1. 2.

Krakovski, M. B. Optimization of RC Structures using Design of Experiments. Computers & Structures, Vol. 63, n 1, 1997. Vianna, L C. Otimizao de Sees Transversais de Concreto Armado: Aplicao a Prticos. Dissertao de Mestrado em Engenharia de Estruturas. Escola de Engenharia de So Carlos da Universidade de So Carlos. So Carlos, 2003.