Topology Optimization for Conceptual Design of Reinforced Concrete Structures

Oded Amir and Michael Bogomolny

Technical University of Denmark, Department of Mechanical Engineering Plasan LTD, Research and Development Department

WCSMO-9, Shizuoka, Japan, June 14 2011

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

1/18

Topology optimization as an architectural design tool

TheApplications in Architecture/Design facade of the Qatar convention center, designed by Arata Isozaki and Mutsuro Sasaki:

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

2/18

sign

Topology optimization as an architectural design tool

Fig. 13 Illustration of the Project UNIKABETON (2010): 50 concept of pattern gradation along the height of a building: a pattern gradation constraints; b topology optimization result with similarities to c John Hancock Center in Chicago, IL (taken from en.wikipedia. org/wiki/John_Hancock_Center)
20 50

the tension and compression lines meet at a 45 angle at the centerline-----in beam theory there is a state of pure shear stress at the centerline. the stresses at the beam edges are vertical because the problem is purely axial. It can be noted how the lines become very dense toward the edges, emphasizing how

result emphasizes the typical behavior of a tubular high rise structure which behaves similarly to a simple I-beam section. The faces of the tube orthogonal to the wind direction are acting as flanges and mainly carry the overturning moment, while the faces of the tube along the wind direction are carrying the shear force.

Stromberg et al. (2011):

70

90

110

80

c
3/18

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

Topology optimization of reinforced concrete

Potential applications: Conceptual design phase - enhance collaboration between architects and engineers. Optimize the utilization of concrete - reduce weight -> material consumption -> CO2 emissions: reinforced/plain, heavy/light. Optimize the distribution of steel reinforcement in non-standard structural elements.

The challenge
In the context of structural topology optimization: consider anisotropic nonlinear material behavior.

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

4/18

Elasto-plastic approach
The main components of the proposed procedure: SIMP-type interpolation of nonlinear elasto-plastic properties, including interpolation of yield criteria. Adjoint sensitivity analysis with no simplifying assumptions. Hybrid approach to sensitivity analysis, enabling to optimize compliance for distributed loads while using displacement control. Recent related work: Kato et al. 2009, Kato and Ramm 2010: optimization of ber reinforced concrete. Liu and Qiao 2011: topology optimization with dierent tensile and compressive properties. Liang et al. 2000, Kwak and Noh 2006, Bruggi 2009: optimized strut-and-tie layouts (linear, elastic, isotropic modeling).
Topology Optimization for Conceptual Design of Reinforced Concrete Structures 5/18

Distribution of elasto-plastic material

Steel is assumed to obey the von-Mises yield criterion; Concrete is assumed to obey the Drucker-Prager yield criterion; on Mises yield criterion which is widelyoptimization densities strength in Intermediate topology used for metals (having equal represent an articial n and compression) can be seen as a particular case of the Drucker-Prager criterion. mixture.

(a) von Mises

(b) Drucker-Prager

Figure 3: Yield surfaces in 2D principal stress

Figure 4: Demonstrative example of the interpolation between two yield surfaces, presented in 2D principal stress space. The Hybrid surface represents the behavior of an articial mixture, space. corresponding to an intermediate density in topology optimization.

s a demonstrative case we for Conceptual Design of article on the distribution of concrete Topology Optimization focus throughout this Reinforced Concrete Structures

6/18

Distribution of elasto-plastic material

Nonlinear material properties are a function of the density variable . Elastic modulus
(Sigmund and Torquato 1997)

p E (e ) = Emin + (Emax Emin )e E

Yield function: f (, , e ) = 3J2 + (e )I1 y (, e )

Pressure-dependency of the strength:

p (e ) = max (max min )e

Yield stress

(Maute et al. 1998)

:
py

0 0 0 y (, e ) = y ,min + (y ,max y ,min )e

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

+ HE (e )
7/18

Problem formulation - maximize end-compliance

Distributing steel within a solid concrete domain, point load: min c() = N p f p u N
Ne

s.t.:
e=1

ve e V 0 e 1, e = 1, ..., Ne n = 1, ..., N n = 1, ..., N

with:

Rn (vn , n ) = 0

Hn (un , un1 , vn , vn1 , ) = 0

For distributing concrete, steel and void: another element design variable xe Ne Ne and two volume fractions v x V1 , v V2 . e=1 e e e=1 e e For distributed loads, the objective N^T uN does not give the stiest f structure - requires special treatment.
Topology Optimization for Conceptual Design of Reinforced Concrete Structures 8/18

Adjoint sensitivity analysis

The augmented objective functional
N

c () = N p f p u N
n=1 N

T Rn (vn , n ) n

n=1

T Hn (un , un1 , vn , vn1 , ) n

The explicit design sensitivities

N exp c Hn = T n e e n=1

Due to path dependency, a backwards-incremental procedure is required for computing the adjoint variables (following the framework by Michaleris et al. 1994).
Topology Optimization for Conceptual Design of Reinforced Concrete Structures 9/18

Adjoint sensitivity analysis

Typical increment: solve the coupled adjoint equations to determine n
(n R) (n H) 1 (n H) (n v) (n v) (n u)
T n

= =

c T (n u) (n+1 H) (n+1 H) (n H) 1 (n H) (n u) (n v) (n v) (n u) c (n )
T n+1

c (n H) 1 (n H) (n v) (n v) (n u)

(n R) T n (n )

Compute the local adjoint vector n on a Gauss-point level

(n H) T n (n R) T n (n+1 H) = n n v) ( ( v) (n v)
T n+1

c T (n v)

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

10/18

Hybrid approach for displacement-controlled objectives

In a load-controlled setting:
f fp Optimized design
update

Design k
update

Design 1

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

11/18

Hybrid approach for displacement-controlled objectives

In a displacement-controlled setting:
f Optimized design
update

Design k
update

Design 1

up

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

11/18

Hybrid approach for displacement-controlled objectives

In a hybrid setting:
f Optimized design Design k+1 fk fk-1
update update

Design k Design k-1

up

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

11/18

Implementation
Nonlinear nite element analysis: Global nonlinear equations solved by a Newton-Raphson procedure with displacement control and automatic incrementation. Local nonlinear elasto-plastic problems solved implicitly after assuming an elastic trial stress. Simplied ow and hardening rules. Topology optimization: Regularization by a density lter Design update by MMA
(Bruns and Tortorelli 2001; Bourdin 2001)

(Svanberg 1987)

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

12/18

Example: ss-beam, 90% concrete, 10% steel

f

?
8
(a) Design domain and boundary conditions.

(b) Optimized layout after 150 design iterations with gradual renement.

Figure 9: Maximum end-compliance of a simply supported beam subject to a distributed load. Black = steel, white = concrete. Steel consists of 10% of the total volume.

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

13/18

Figure 9: Maximum end-compliance of a simply supported beam subject to a distributed load. Black = steel, white = concrete. Steel consists of 10% of the total volume.

Example: cantilevered beam, 90% concrete, 10% steel

?
2.4

?
7.2
(a) Design domain and boundary conditions.

?
2.4

(b) Optimized layout after 300 design iterations with gradual renement.

Figure 10: Maximum end-compliance of a cantilevered beam subject to a distributed load. Black = steel, white = concrete. Steel consists of 10% of the total volume.
Topology Optimization for Conceptual Design of Reinforced Concrete Structures 14/18

Example: short cantilever

1

?
{
f

(a) Design domain, boundary conditions and prescribed displacement.

(b) Optimized layout after 500 design iterations, 80% concrete, 20% steel. Black = steel, gray = concrete.

(c) Optimized layout after 200 design iterations, 30% concrete, 10% steel, 60% void. Black = steel, gray = concrete, white = void.

Figure 12: Maximum end-compliance of a short cantilever

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

15/18

Example: clamped beam, 32.5% concrete, 7.5% steel

prescribed displacement directed downwards at the middle of the bottom edge. The model of the symmetric half is discretized with a 120 80 FE mesh. Further details regarding the numerical implementation are given in Table 2.

(a) Design domain, boundary conditions and prescribed displacement.

(b) Optimized layout: 32.5% concrete, 7.5% steel, 60% void. Black = steel, gray = concrete, white = void. Figure 9: Maximum end-compliance of a thick clamped beam

Examining the Design design, it can be seen Structures Topology Optimization for Conceptual resultingof Reinforced Concretethat the optimization process leads to

16/18

Example: ss-beam, 35% concrete, 5% steel

Figure 5: Maximum end-compliance of a simply supported beam: Design domain, boundary conditions and prescribed displacement.

Topology Optimization for Conceptual Design of Reinforced Concrete Structures

The layouts obtained for the simply supported beam resemble actual design of reinforced concrete beams. Away from the supports, bending action is dominant so steel is necessary in the bottom bers where tension stresses appear. Near the supports, shear forces are dominant so concrete typically cracks in an angle of 45 , corresponding to the direction of the principal stresses in pure shear. Consequently, the steel reinforcement should be bent in order to accommodate the tensile stresses due to shear. Examining the results of the optimization procedure, it is evident that the distribution

17/18

Towards realistic modeling

Limitations of the approach: Dicult to accommodate small volume fractions of the steel phase. Strain softening of the concrete is not considered - dicult to achieve optimized strut-and-tie layouts. Current work in progress: Replace elasto-plasticity with continuum damage modeling. Consider anisotropic reinforced concrete: mixing theory, homogenization, other approaches.
2
Chapter 1. Introduction

composite material: reinforced concrete y

fibers: steel

matrix: concrete

Chapter 1. Introduction

Figure 1.1. Numerical modeling of reinforced concrete walls subjected to shear.

!

=


localization (concentration) tensor Ai which relates composite material: the homogenized macroscopic strains to the local reinforced concrete strains within eachyphase i = A i : , 2, zi = 1, x m. (16)

y z x

C1
=

Cm C2
l

Figure 2: Representative volume element (RVE).

The tensor Ai of each phase accounts for the morphology of the microstructure by considering the elasticity, the volume fraction, the aspect ratio, the orientation and the shape of each constituent. It should emy fibers: steel y phasized that Ai relates micro and macro quantities x x and depends therefore on the theory chosen for the matrix: micromechanical model. If a three-phase composite concrete is considered only two concentration tensors have to be known. The third one can be determined from the Figure 1.1. Numerical modeling of reinforced concrete walls subjected to shear. average value

matrix: concrete fibers: steel

z composite material matrix d >d matrix

Figure 1.2. Numerical modeling of reinforced concrete beams.

y 3.3 Micro-macro mapping <z A >Vx= c1 A1 + c2 A2 + cm Am = 1 1, (17) continuum mechanics, the phenomenon of strain localization caused by the appearIn ance of a crack can be represented by means of enriched kinematics describing a jump in the The local strain and stress elds within the RVE are with 1 denoting the forth-order unit tensor. strain field. 1 Due averaged over the total volume V of the RVE in order to different orientation and shape of the If Images: Rumanus and Meschke strains Linero, Oliver and Huespe 2007 inhomo- the constitutive model of the composite material considers a common strain field to evaluate the homogenized values of the 2007; geneities, which is captured by Ai , the mechanical the materials that form it, then the kinematics equipped to capture discontinuities in matrix: y among n simple Topology >V = 1 (x) dV =for cConceptual(12) responseReinforcedconcrete Design of of the related homogenized stiffness tensor materials can be applied in an equal way in the composite material, this constitutes an Concrete Structures 18/18 < Optimization i < >i