COMPUTATIONAL METHODS IN ENGINEERING AND SCIENCE EPMESC X, Aug. 21-23, 2006, Sanya, Hainan, China ©2006 Tsinghua University Press & Springer Multi-Objective Optimization for Shape Design of Arch Dams Linsong Sun 1*, Weihua Zhang 1, Nenggang Xie 2 1 2 College of Hydraulic Science & Engineering, Yangzhou University, Yangzhou, 225009 China College of Mechanical Engineering, Anhui University of Technology, Maanshan, 243002 China Email: sunlinsong@sohu.com Abstract A multi-objective optimization model is established for the shape design of arch dams. In this model, the geometric parameters describing dam shape is taken as design variables; four objectives are considered of dam volume, maximal tensile stress, maximal pressure stress and relative depth of the zone with tensile stress that is larger than 1.0MPa; and the constraints include geometric constraints, dam volume constraint and mechanical property constrains, i.e., stress constraints, stability constraints, etc. Traditionally, multi-objective optimization schemes transform multiple objective functions into a single objective function in some manner, weighted sum method and utopia point method for examples, and the resulting problem is solved as a single objective optimization problem. In this paper the cooperative game model analogy to the multi-objective optimization is proposed with each player correspond to one of the objective functions. After defining the utility of each player, the Nash arbitrary scheme is used to solve the cooperative game. The optimization of BAIHETAN arch dam, which is located in Sichuan province of China, is calculated as an engineering example. The results are compared to those obtained by weighted sum method and utopia point method and indicate that the cooperative game method is superior to traditional methods. The optimal design saves 9.77×104m3 of dam volume, compared to initial design, along with the decrease of maximal tensile stress with 38.22%, the decrease of maximal pressure stress with 28.83%, and the decrease of relative depth of large tensile stress zone with 10.15%. Key words: multi-objective optimization; shape design; cooperative game; arch dam INTRODUCTION Shape optimization of arch dams starts from 1970s. The early studies focus on economic aspect that is to minimize dam volume [1]. In recent decades the safety aspect has been taken into account more and more in shape optimization of arch dams. Sun et al [2] discussed the optimal design of arch dam with maximum tensile stress as objective to be minimized. Li [3] discussed the equivalence of safe model and economic model in arch dam optimization, and a bin-objective optimization model is established to minimize both dam volume and maximum stress. Xie et al [4] proposed a multi-objective optimization model for arch dam design with dam volume, maximum tensile stress and maximum strain energy to be minimized. Wang et al [5] introduced a flexible modeling method for multi-objective optimization of arch dams. The dam volume is taken as economic index. Stress level, probability of failure and area of high stress zone are taken as safety indices. The different types of optimal model can be formed according to decision-maker’s favor. Multi- objective optimization methods appeared for the first time in economics, and a growing interest arose in engineering. Many multi-objective structural optimization examples, solved with different methods, may be found in a survey paper of Marler et al [6] Sun et al [7] solved the bin-objective optimization problem for shape design of arch dams with fuzzy theory. The non-inferior solutions are obtained by weighted sum method first, and the optimal solution is selected according to grade of fuzzy closeness. In the work of Wang et al [5], the optimal solution was obtained with utopia point method. Xie ⎯ 1009 ⎯ et al [4] proposed a fuzzy evaluation function, which is the combination of grade of membership of objectives, for the multi-objective optimization of arch dams under the actions of static and dynamic load. In this paper, the multi-objective optimization of arch dams is discussed with four objectives, i.e., dam volume, maximum tensile stress, maximum pressure stress and relative depth of large stress zone. The optimization of BAHETAN arch dam located in Sichuan province of China is presented as an example. The cooperative game method is used to treat the presence of multiple objectives. The results obtained are presented and compared with those obtained by other multi-objective optimization techniques as weighted sum method and utopia point method. GEOMETRIC MODEL OF PARABOLIC DOUBLE-CURVATURE ARCH DAM’S SHAPE The shape of arch dams is usually described by center vertical cantilever and some horizontal arch rings in reference elevations. ϕL ϕR ϕm Figure 1: The cantilever of arch dam Figure 2: Horizontal arch ring 1. The geometric description of center cantilever The center cantilever, which is shown in Figure1, can be described by its upstream curve ycu and width Tc. Suppose they are cubic polynomial function of z, i.e. y cu ( z ) = a0 + a1 z + a 2 z 2 + a3 z 3 (1) Tc ( z ) = b0 + b1 z + b2 z 2 + b3 z 3 (2) Where a0 ~ a3 and b0 ~ b3 can be expressed as follows by ycu ( z ) and Tc (z ) at the four reference elevations with z coordinates of z1, z2, z3 and z4. ⎧a0 ⎫ ⎡1 ⎪a ⎪ ⎢ ⎪ 1 ⎪ ⎢1 ⎨ ⎬=⎢ ⎪a 2 ⎪ ⎢1 ⎪⎩a3 ⎪⎭ ⎢⎣1 z1 z12 z2 z3 z z z4 z 2 2 2 3 2 4 ⎧b0 ⎫ ⎡1 ⎪b ⎪ ⎢ ⎪ 1 ⎪ ⎢1 ⎨ ⎬=⎢ ⎪b2 ⎪ ⎢1 ⎪⎩b3 ⎪⎭ ⎢⎣1 z1 z12 z2 z3 z 22 z 32 z4 z 42 z13 ⎤ ⎥ z 23 ⎥ z 33 ⎥ ⎥ z 43 ⎥⎦ −1 z13 ⎤ ⎥ z 23 ⎥ z 33 ⎥ ⎥ z 43 ⎥⎦ −1 ⎧ ycu ( z1 ) ⎫ ⎪ y ( z )⎪ ⎪ cu 2 ⎪ ⎬ ⎨ ⎪ ycu ( z 3 ) ⎪ ⎪⎩ ycu ( z 4 )⎪⎭ (3) ⎧Tc ( z1 ) ⎫ ⎪T ( z )⎪ ⎪ c 2 ⎪ ⎬ ⎨ ⎪Tc ( z 3 ) ⎪ ⎪⎩Tc ( z 4 )⎪⎭ (4) From Eqs. (1) and (2), The downstream curve of cantilever is expressed as following: ⎯ 1010 ⎯ y cd ( z ) = y cu ( z ) + Tc ( z ) (5) So, the center cantilever can be defined by character parameters of y cu ( z i ) and Tc ( z i ) at the four reference elevations. It should be note that, for the convenient of construction, the overhangs on upstream and downstream, i.e. KU and K D respectively, must be constrained, and they can be expressed as ′ ( H ) = a1 + 2a 2 H + 3a 3 H 2 K U = y cu (6) ′ (0) = y cu ′ (0) + Tc′(0) = a1 + b1 K D = y cd (7) 2. The geometric description of horizontal arch ring The arch ring shown in Figure 2 can be described by axial curve and ring thickness. Take left half for example, the axis of parabolic arch ring can be expressed by parameter variable ϕm as ⎧ xm = RCL tan ϕ m ⎪ x m2 ⎨ y y = + c ⎪ m 2 RCL ⎩ (8) 1 Where yc = ycu + Tc is the y coordinate of ring axis at arch center, RCL is the radius of axis at arch 2 center. The thickness of arch ring is supposed to change by ϕm as Tm = Tc + [TL − Tc ] ⋅ 1 − cos ϕ m 1 − cos ϕ L (9) where Tc and TL are the thickness at arch center and arch end respectively. ϕ L is called semi-center angle and can be expressed as ϕ L = tan −1[ XL ] RCL (10) in which, X L is the length of left chord of arch axis and can be treated as a constant at certain elevation. After the determining of arch axis and thickness, the upstream curve and downstream curve can be calculated by the following formulas xmu = xm + 0.5Tm sin ϕ m ⎫ ⎬ y mu = y m − 0.5Tm cos ϕ m ⎭ (11) xmd = xm − 0.5Tm sin ϕ m ⎫ ⎬ y md = y m + 0.5Tm cos ϕ m ⎭ (12) So, if the center cantilever has been defined, the shape of left half of certain arch ring can be determined by RCL and TL , and similarly, the shape of right half can be determined by RCR and TR corresponding to the radius of right half axis at arch center and the thickness at right arch end respectively. As to the whole dam, it is supposed that RCL , TL , RCR and TR are cubic polynomial functions of z, i.e. RCL ( z ) = c0 + c1 z + c2 z 2 + c3 z 3 (13) TL ( z ) = d 0 + d1 z + d 2 z 2 + d 3 z 3 (14) RCR ( z ) = e0 + e1 z + e2 z 2 + e3 z 3 (15) ⎯ 1011 ⎯ TR ( z ) = f 0 + f1 z + f 2 z 2 + f 3 z 3 (16) in which, the coefficients c0 ~ c3 , d 0 ~ d 3 , e0 ~ e3 and f 0 ~ f 3 can, similarly to center cantilever, be expressed by corresponding parameters RCL ( z i ) , TL ( z i ) , RCR ( z i ) and TR ( z i ) at the four reference elevations with z coordinates of z1, z2, z3 and z4. MULTI-OBJECTIVE OPTIMIZATION MODEL FOR ARCH DAM DESIGN 1. General description of multi-objective optimization problem A constrained multi-objective optimization problem can be mathematically formulated as follows: min F ( X ) = [ f 1 ( X ) f 2 ( X ) L f m ( X )]T s.t. X iL ≤ X i ≤ X iU i = 1,2,L , n g j ( X ) ≤ 0 j = 1,2, L p hk ( X ) = 0 k = 1,2, L q ⎫ ⎪ ⎪ ⎬ ⎪ ⎪⎭ (17) where F is the optimal objective vector of the scalar objective functions, in number of m, X is the design variables vector, X iL and X iU are the lower and upper bounds for the ith design variable, in number of n, gj(X) are the inequality constraints, in number of p, hk(X) are the equality constraints, in number of q. 2. Design variables in arch dam optimization From above text, it is known that the shape of parabolic double curvature arch dam can be described by 24 character parameters of y cu (z ) , Tc (z ) , RCL ( z ) , TL ( z ) , RCR ( z ) and TR ( z ) at the four reference elevations. So, in the shape optimization of arch dams, it’s very natural to take these parameters as design variables. 3. Objective functions in arch dam optimization The aims of arch dam optimization contain two aspects of economy and safety. The dam volume can be taken as the economical objective function, which is simple relatively. The objective function of safety is complicated. The maximum tensile principal stress is taken as safety objective function in [1]. But it is not enough to use maximum stress as safety index alone. A dam with smaller region of large stress is safer than that with larger region of large stress, even if they have the same maximum stress. So, it's necessary to take the region of high stress as another safety objective function. To convenient for the comparison between different design schemes, the relative depth of large tensile stress zone, which is the radio of the depth of large tensile stress zone on dam base to corresponding dam thickness, is taken as a safety objective function. So, the scalar objective functions in multi-objective optimization of arch dams contain the follows: f1 ( X) = V f 2 ( X) = σ t max f3 ( X) = σ c max f 4 ( X) = d max ⎫ ⎪ ⎪ ⎬ ⎪ ⎪⎭ (18) where V is the dam volume, σ tmax is the maximal tensile principal stress, σ cmax is the absolute maximal pressure principal stress, and dmax is the relative depth of large tensile stress zone on dam base. The scalar objective functions in Eq.(18) are in different dimension, and their value are discrepant very much. It's necessary to make them normalization as f1 ( X) = V [V ] σ f3 ( X) = c max [σ c ] σ t max ⎫ [σ t ]⎪ ⎬ d ⎪ f 4 ( X) = max [d ] ⎭ f 2 ( X) = (19) where [·] are the upper bounds of corresponding variable, which are determinate according to design specification or the analysis of initial design scheme. ⎯ 1012 ⎯ 4. Constraints in arch dam optimization The constraints, which should be subject to in arch dam optimization, include geometric constraints, dam volume constraint and mechanical property constrains, i.e., stress constraints, stability constraints, etc. Mathematically, the constraints are formulated as K U ≤ [K U ] K D ≤ [K D ] V ≤ [V ] σ t max ≤ [σ t ] σ c max ≤ [σ c ] d max ≤ [d ] ϕ ≤ [ϕ ] ⎫ ⎪ ⎪ ⎪ ⎪ ⎬ ⎪ ⎪ ⎪ ⎪ ⎭ (20) in which, ϕ is semi-center angle of arch ring. MULTI-OBJECTIVE OPTIMIZATION TECHNIQUES Multi-objective optimization schemes usually transform multiple objective functions into a single objective function in some manner and the resulting problem is solved as a single objective optimization problem. 1. Weighted sum method In this method, the objective functions are made scalars with the weighted sum, and the new problem is posed as follows: m f ( X ) = ∑ wl f l ( X ) min l =1 X iL ≤ X i ≤ X iU i = 1,2, L, n g j ( X ) ≤ 0 j = 1,2, L p s.t. hk ( X ) = 0 k = 1,2,L q ⎫ ⎪ ⎪⎪ ⎬ ⎪ ⎪ ⎪⎭ (21) where wl is a positive constant indicating the weight (and hence importance ) assigned to fl. By giving a relatively large value to wl it is possible to favor fl over other objective functions. Note that the condition m ∑w l =1 l = 1 should be hold in Eq. (21). 2. Utopia point method With the utopia point method, the following optimization problems are solved one at a time: min s.t. fl (X ) X iL ≤ X i ≤ X iU i = 1,2, L , n g j ( X ) ≤ 0 j = 1,2, L p hk ( X ) = 0 k = 1,2, L q ⎫ ⎪ ⎪ ⎬ ⎪ ⎪⎭ (l = 1, 2, L , m) (22) Let f l * = f l * ( X * ) be the optimum of lth objective function and X* is corresponding vector of design variables. In objective function space, the point of ( f1* , f 2* ,…, f m* ) is called utopia point and the distance between the solution point and utopia point is taken as an evaluation function. Then, the multi-objective optimization problem (17) is transformed to the following scalar optimization problem: m ∑( f (X ) − f min f (X ) = s.t. X ≤ X i ≤ X iU l =1 L i g j (X ) ≤ 0 l * l ) i = 1,2,L , n j = 1,2, L p hk ( X ) = 0 k = 1,2,L q ⎫ ⎪ ⎪⎪ ⎬ ⎪ ⎪ ⎪⎭ (23) ⎯ 1013 ⎯ The solution of Eq.(23) is Pareto optimum of Eq.(17). 3. Cooperative game theoretic model and Nash arbitration scheme The multi-objective optimal design problem is analogy to a game in which each player corresponds to one of the objective functions. Let fl,w be the worst value of lth objective function that the corresponding player will accept, i.e. an upper limit on the lth objective. The utility of lth player is defined as u l ( X ) = f l,w − f l ( X ) . The aim of each player in a game is to maximize its utility. In this situation of conflicting interests, a cooperative approach, based on the concept of Pareto-optimality, is adopted. Nash [8] suggests an arbitration scheme to solve the cooperative game by maximizing the product of the players’ utilities, i.e. maximizing the following function: m m l =1 l =1 C ( X ) = ∏ u l ( X ) = ∏ [ f l,w − f l ( X )] (24) In the case of multi-objective optimization of arch dams with normalized objective functions in Eq.(19), the worst value of each objective function is 1. So, the transformed problem by Nash arbitration scheme is as follow: m max C ( X ) = ∏ [1 − f l ( X )] l =1 s.t. X ≤ X i ≤ X iU i = 1,2, L, n g j ( X ) ≤ 0 j = 1,2, L p L i hk ( X ) = 0 k = 1,2,L q ⎫ ⎪ ⎪⎪ ⎬ ⎪ ⎪ ⎪⎭ (25) ENGINEERING EXAMPLE BAIHETAN arch dam, which located in Sichuan province of China, is a parabolic double curvature arch dam with the height of 277.0m, the dam crest elevation of 827.0m, and the normal high water level of 820.0m. In this study, four objectives in Eq. (2) are adopted in optimization, where dmax is the relative depth of the zone with tensile stress larger than 1.0MPa. The finite element method is taken for the structural analysis, and the FE mesh of dam body is arranged as 8-layer elements in height and 6-layer elements in thickness. The load case considered is “hydrostatic pressure of normal high water + temperature-down load + self-weight of dam body before grouting”. The material constants are shown in Table 1, in which, rock A distributes form elevation of 780.0m to 827.0m in left bank; rock B distributes from elevation of 600.0m to780.0m in left bank and from elevation of 700.0m to 827.0m in right bank; rock C distributes from elevation of 600.0m to 700.0m in right bank and rock D distributes under elevation of 600.0m. Table 1 Material constants of dam concrete and foundation rocks Materials Young’s module (GPa) Poisson’s ratio Unit weight (kN/m3) Thermal coefficient (/°C) Concrete Rock A Rock B Rock C Rock D 21.0 12.0 14.0 16.0 20.0 0.17 0.25 0.22 0.22 0.20 24.0 / / / / 1.0×10-5 / / / / The initial design of dam shape provided by design department is analyzed first, and based on the results listed in the last row of Table 2, the main constraints in optimization process are taken as follows: ⎯ 1014 ⎯ v max ≤ 700.0 × 10 4 m 3 σ tmax ≤ 9.0MPa σ cmax ≤ 17.5MPa d max ≤ 0.4 ′ ≤ 0.5MPa σ tmax K U ≤ 0.30 K D ≤ 0.25 ⎫ ⎪ ⎪ ⎪ ⎪ ⎬ ⎪ ⎪ ⎪ ⎪ ⎭ (26) ′ where, σ tmax is the maximal tensile stress by self-weight of dam body before grouting, KU and KD are overhung of upstream surface and downstream surface, respectively. The solutions obtained with different multi-objective optimization techniques are reported in Table 2, in which, the weights for four objectives are the same of 0.25 in weighted sum method, and the utopia point is * (V * , σ t*max , σ c*max , d max ) = (611.27 × 10 4 m 3 , 4.29MPa , 11.93MPa , 0.3225) which is obtained by four single objective optimization. Table 2 Multi-objective optimal solutions by different methods Weighted sum Utopia point Game theory Initial design V/×104m3 σtmax/MPa σcmax/MPa dmax 699.40 643.72 670.17 679.94 4.41 6.13 5.30 8.58 12.05 12.31 12.12 17.03 0.3242 0.3767 0.3523 0.3921 The results show that, although a normalization of four objective functions is adopted in the form of Eq.(19), to deal with numerical homogeneous quantities, weighted sum method gives no optimization to dam volume, while utopia point method and game theory method both optimize all of the four objectives. If the global design quality defined by Spallino et al [9] is used as following, Q= V σ t max σ c max d max + + + * V * σ t*max σ c*max d max (27) Figure 3: Compare of center sections The value of Q for utopia point method and game theory method are 4.683 and 4.440, respectively. That is to say the optimal solution by game theory method is better than that by utopia point method. In additional, the game theory method adopted here doesn’t need to solve single objective optimization problems, in order to obtain the utopia point, first. So, the computational cost of game theory method is less than that of ⎯ 1015 ⎯ utopia point method. The center cantilevers of optimal designs and initial design are compared in Figure 3. It is shown that the position of bulgy point on upstream bound of center cantilever is elevated in optimal designs. This change is in favor of improving the dam’s state of stress. Compared to initial design, the optimal design by game theory decreases 38.22% of maximum tensile stress, 28.83% of maximum pressure stress and 10.15% of relative depth of zone with tensile stress larger than 1.0MPa. Table 3 shows the dam parameters of optimal design by game theory method and Figure 4~6 show the contours of stresses. Table 3 Dam parameters of optimal design by game theory Elevation (m) 827.00 780.00 740.00 690.00 640.00 600.00 570.00 550.00 Parameters of center cantilever Thickness yc (m) 0.000 -22.018 -34.276 -41.911 -41.789 -36.673 -30.192 -24.713 15.638 35.136 44.415 50.622 55.453 61.653 69.327 76.476 Thickness of arch end (m) Semi-center angle (°) Radius of arch axis at center (m) Left Right Left Right Left Right 17.068 39.416 48.933 51.844 66.092 76.996 76.089 84.103 17.450 35.241 51.247 58.035 65.551 72.399 77.491 84.282 48.538 47.652 50.377 48.742 44.545 41.589 34.468 17.530 48.170 44.636 41.045 42.233 44.094 38.162 31.213 13.444 322.419 298.259 245.776 226.129 213.754 185.601 178.707 163.446 234.237 250.309 273.350 247.236 205.344 207.682 203.521 216.391 3 3 5 2 3 1 8 6 7 5 1 -2.00 2 -1.00 3 0.00 4 1.00 5 2.00 6 3.00 7 4.00 8 5.00 min -2.72 max 5.29 4 Figure 4: Contours of principal stress σ1 on upstream surface of optimal design by game theory 4 7 6 2 5 4 3 1 1 -12.00 2 -10.50 3 -9.00 4 -7.50 5 -6.00 6 -4.50 7 -3.00 min -12.13 max -2.78 2 2 Figure 5: Contours of principal stress σ3 on downstream surface of optimal design by game theory ⎯ 1016 ⎯ 1 4 1 1 1 2 4 3 6 7 1 -1.00 2 0.00 3 1.00 4 2.00 5 3.00 6 4.00 7 5.00 min -1.43 max 5.29 5 Figure 6: Contours of principal stress σ1on base of optimal design by game theory CONCLUSIONS This study proposed a multi-objective optimization model of arch dam design with four objectives of dam volume, maximal tensile stress, maximal pressure stress and relative depth of large stress zone. Three multi-objective optimization techniques are used to solve the problem. The optimization of BAIHETAN arch dam is calculated as an engineering example. The results are compared and indicate that game theory method is superior to weighted sum method and utopia point method in the field of multi-objective optimization of arch dams. Acknowledgements The support of National Nature Science Foundation of PR China (NSFC) under grant No. 90410011 and 50409017 is gratefully acknowledged. REFERENCES 1. Zhu Bofang. Shape optimization of arch dams. International Water Power and Dam Construction, 1987; 39(3): 43-48 2. Sun Linsong, Wang Dexin and Pei Kaiguo. Stress-oriented shape optimization of arch dams. Journal of Hohai University (Natural Sciences edition), 2000; 28(1): 57-60 (in Chinese). 3. Li Yisheng. Non-inferior solution set of bin-objective optimization and bin-objective optimization of arch dams. Water Power, 1998; (10): 10-14 (in Chinese). 4. Xie Nenggang, Sun Linsong, Wang Dexin. Multi-objective optimization of high arch dam shape under the actions of static and dynamic load. Journal of Hydraulic Engineering, 2001; 32(10): 8-11 (in Chinese). 5. Wang Shuyu, Liu Guohua, Du Wanggai, Ma Yichao. Study and application of multi-optimization for arch dam design. Journal of Hydraulic Engineering, 2001; 32(10): 48-53 (in Chinese). 6. Marlers RT, Arora JS. Survey of multi-objective optimization methods for engineering. Struct. Multidisc Optim, 2004; 25: 369-395. 7. Sun Wenjun, Sun Linsong, Wang Dexin, Li Chunguang. Bin-objective shape optimization of arch dams. Journal of Hohai University (Natural Sciences edition), 2000; 28(3): 39-43 (in Chinese). 8. Nash J. The bargaining problem. Econometrica, 1950; 18: 155-162. 9. Spallino R, Rizzo S. Multi-objective discrete optimization of laminated structures. Mechanics Research Communications, 2002; 29: 17-25. ⎯ 1017 ⎯
© Copyright 2025 Paperzz