Design of Structural Braced Frames Using Group Optimization

Lauren Stromberg1, Alessandro Beghini2, William F. Baker2, and Glaucio H. Paulino1

1

Newmark Laboratory, Department of Civil and Environmental Engineering, University of

Illinois at Urbana-Champaign, Urbana, IL, 61801; PH (217) 333-3817; email:
paulino@illinois.edu
2
Skidmore, Owings & Merrill, LLP, 224 S. Michigan Avenue, Chicago, IL 60604
ABSTRACT
Though topology optimization has been applied to many fields, ranging from mechanical
to aerospace engineering, more work must be done to tailor it to the needs of the structural
engineer, especially in regards to the design of high-rise buildings. Thus, this work aims to
improve its application to structural engineering by describing an integrated topology
optimization approach involving continuum and discrete finite elements to design the lateral
systems in structural braced frames for high-rise buildings. The approach is implemented using
concurrent continuum finite elements and discrete beam/truss elements to simplify and improve
the overall design process by creating optimal geometries for a given volume of material. For
example, after an engineer develops a structural frame consisting of beams and columns sized for
gravity loads, topology optimization on the continuum (e.g. quadrilateral) elements is used to
create a conceptual design for the braces of the lateral system resulting in highly efficient
structures. Several practical examples are demonstrated to show the importance and relevance of
this work to the structural design industry.
INTRODUCTION
In the construction industry, many works of modern architecture are lacking balance
between engineering principals and architectural aesthetics. Often, the engineering can be seen as
a means to make the architecture stand, where in this case the architecture is similar to a work of
art. Often these structures may use much more material than required or cause difficulties in
constructability and realization of the design. Other cases include buildings which are built solely
to serve a specific function at low cost, with little or no regard for aesthetics. Examples of such
structures include big box stores, warehouses, convention centers, and strip malls. This lack of
balance provides motivation to introduce topology optimization to connect engineering and
architecture in the design of structures. Topology optimization provides unique optimal solutions
for a given design space and set of applied forces and boundary conditions. Such optimal
solutions often provide aesthetic value in the realization of patterns in the final design as well.
Though topology optimization is typically know for its use in mechanical and
aeronautical engineering, and it is becoming more common for civil engineering applications in
recent years. Examples of such applications include the multi-story building design or long span
bridge design applications presented in Stromberg et al. (2011a), Allahdadian and Boroomand
(2010), Neves et al. (1995), or Huang and Xie (2008). Though the applications within the civil
engineering field encompass a wide range of topics, the focus of this work is towards the highrise building industry to provide engineers with a tool that can be used to identify the optimal
topology of the lateral bracing systems in addition to minimize material usage and corresponding

cost. Thus, this work aims to develop a methodology that enables engineers to design the lateral
system from the conceptual optimal bracing angles to the final sizing of the members.
By using topology optimization to express the engineering together with the architecture,
buildings can be developed with unique bracing systems. Such braced frame and moment frame
structural systems are commonly deployed in the lateral design of high-rise buildings, such as the
John Hancock Center (Chicago, IL), Broadgate Tower (London, UK) and Bank of China Tower
(Hong Kong). However, the design of such systems is traditionally based on diagonal braces
arranged according to 45 through 60 degree angles, though there have been few engineering
studies in the past to identify the optimal bracing angle and the parameters affecting such angles
(Huang et al., 2010). This work, on the other hand, aims to optimize the geometry of the frame
members in terms of structural performance (maximum stiffness) while minimizing the material
usage. Other measures of structural performance might also include tip deflection, frequency,
critical buckling load, etc.

(a)

(b)

(c)

Figure 1. Comparison of existing topology optimization techniques with new methods

proposed in this work: (a) continuum topology optimization with no frame elements, (b)
with column elements (in red), (c) with beam and column frame elements (in red)

In Stromberg et. al (2011a), an early attempt was made to identify optimal bracing angles
using continuum elements. However, some limitations were evident in the optimization results as
seen in Figure 1 (left). For example, the results show high concentrations of material towards the
extreme edges of the domain, as expected from the web-flange behavior discussed in Stromberg
et. al (2011a). These dense regions of material make it difficult to locate the bracing work points
(i.e. the locations where the diagonals intersect the columns). Furthermore, such high

concentrations may lead to incorrect flexural stiffness in the analysis of the structure. Moreover,
with a constraint imposed on the fraction of material available for the design, if the majority of
the material is optimal at the extreme edges of the domain, relatively small amounts of
material are leftover to form the diagonal members and the structure has an incomplete
diagonalization, which is not practical for realistic building design. An additional constraint on
the material distribution between the columns and diagonals might be imposed to circumvent the
issue, but the addition of beam-column elements to the design domain eliminates the problem
altogether.
For instance, with the additional column elements to the design problem as shown in
Figure 1 (center), the issues mentioned are no longer of concern for structural design. The
discrete column elements, much narrower in width, now give practical bending stiffness to the
structure. Furthermore, the diagonalization is complete along the structures height and the
bracing members are clearly identified. We highlight next how this methodology might be
incorporated into the structural design process.
GROUP OPTIMIZATION CONCEPTS
In this work, group optimization refers to the combination of different groups of elements
(e.g. Q4, bar, beam, etc.) in designing an optimal structure. Using group optimization, a clear,
complete diagonalization results for maximum stiffness design with minimal volume. This
integration of beam and quadrilateral elements can be modeled using one of the two connections
types described next, which can be incorporated into the classical topology optimization
formulation by introducing a few modifications as described below.
Combining Q4 and Beam Elements. The beams and quadrilateral elements are connected in
this work either at the extreme ends of the beam only, or continuously along the beam line, as
illustrated by Figure 2. In the first case, the degrees of freedom shared between elements (shown
in red) are only the two translations at the corners of the quadrilateral and edges of the beam
mesh. For the latter, the two translations at coincident nodes are shared (in red) along both the
beam and quadrilateral meshes.

Figure 2. Connection types for combining continuum and discrete elements: (left) attached
at extreme ends only, (right) attached continuously along the beam line
Problem Statement. In this work, the objective of the optimization problem is to maximize the
overall stiffness of a building, or minimize the compliance. The outer skin or shell of the
building is taken as the design domain (see Figure 3) so that the structural system would be
expressed in the exterior together with the architecture. Thus, the optimal layout problem in
terms of minimum compliance can be stated using the density, , and the displacements, , as
follows:

The overall compliance of the structure is denoted by in the equations above, while
represents the global stiffness matrix which depends on the material densities,
and are the
vectors of nodal displacements and forces, respectively,
is the volume fraction constraint
which represents the maximum volume of material permitted for the design of the structure, and
is the material density for each design variable. A void is signified by a null material density,
0, and
represents solid material. For regions of gray material or intermediate
densities, the commonly used Solid Isotropic Material with Penalization (SIMP) model is
employed (Zhou and Rozvany (1991), Rozvany et al. (1992), Bendsoe (1989), Bendsoe and
Sigmund (1999)):

This power-law relationship between stiffness and element density uses the Youngs Modulus of
solid material
and penalization power
to force the material to tend towards 0 or 1 (void
or solid respectively) where the element density assumes a value somewhere in this range. The
optimization process presented in this work for braced frames also includes continuation on the
penalization power from 1 to 4 in steps of 0.5 until convergence.

Figure 3. Design domain (outer skin or shell) for topology optimization of a building

Design Process. The optimization techniques described in Baker (1992) help streamline the
design decisions at various stages of a project from the conceptual characterization of a braced
frame layout to the final sizing of the members. Combining the Principle of Virtual Work with
the Lagrange Multiplier Method, the deflection, , is expressed as

where
and are the internal forces due to the applied and unit loads respectively,
is the
length, is the Youngs Modulus, and
is the cross-sectional area for the
member, and
is the total volume of the structure. Solving, the final cross-sectional areas required to achieve a
target deflection,
, are given as
[

Once the overall shape or outer skin of the building is known, the optimal bracing layout could
be established assuming that frame columns are arranged around the outer perimeter at a regular
spacing to ensure that the tributary areas for the columns are similar. At each floor level, a
horizontal beam (spandrel) would span between two subsequent columns. Beams and columns
would be modeled using beam elements while the space bounded by two columns and two
beams would be meshed using quadrilateral elements. After the finite element mesh is completed
the following steps can be applied in sequence in the design flow process:
1. size vertical line elements (columns) according to gravity load combinations (accounting
for dead, superimposed dead and live loads) according to Baker (1992)
2. run topology optimization on the continuum elements for lateral load combinations
(accounting for wind and seismic loads)
3. identify the optimal bracing layout based on results and create frame model
4. optimize the member sizes using the virtual work methodology
These steps describe a complete process from the conceptual design and geometry up to the final
sizing of the frame members. Throughout the design phase, these steps may be implemented in
isolation or the entire process may be repeated to resize members.

OPTIMAL BRACED FRAMES

In this section, we explore several important analytical aspects of optimal braced frames in
regards to the frame geometry.
Fully Stressed Design. The energy-based design method presented in Baker (1992) and
described in the previous section implies that any frame with optimal cross-sectional members
subject to a point load at the top is under a state of constant stress (Fully Stressed Design) as
demonstrated next.
By taking the derivative of the equation for deflection with respect to the areas

and solving

for the Lagrangian multiplier, we obtain:

When the structure is loaded with a single point load at the top of the frame, the force can be
expressed by
where is a proportionality constant, thus the following equation holds:
( )
In the above expression the Lagrangian multiplier is a constant, therefore the stress in the
member,
, is also a constant. This conclusion applies to each of the
members of the
frame, therefore the stress level is constant in every member. It follows that the stress is then
constant throughout the frame.
For structures where multiple loads are applied, the compliance is given using the
external and internal work,
and
and the applied forces and displacements and at
the
node as follows:

Using the Lagrange multiplier method as was done previously,

Minimizing the compliance and solving for the Lagrange multiplier, we obtain the following
expression:
( )
This expression shows that the minimization of compliance leads to a state of constant stress. We
note that for the compliance minimization problem, a state of constant strain energy density
represents the condition of optimality (Bendsoe and Sigmund, 2002). Since the Von Mises
stresses and the strain energy density are proportional (Hill, 1950; Lubliner, 1990), the effective
stresses in optimal structures are constant.
The constant stress condition is verified in the continuum approach for the structure in
Figure 4 (bottom right) which was derived using a Q4 element mesh. As shown here, the VonMises stresses are nearly constant within each optimized member.

Figure 4. Plot of Von Mises Stresses for topology optimization problem of symmetric
cantilever problem

Optimal Frame Geometry. The optimal geometry of braced frames can be described by
considering the geometry shown in Figure 5, with the overall width of the structure given as
,
the total height as and the unknown location of the base to the optimal bracing point as .
By applying a unit load and taking advantage of the symmetry of the problem, the internal forces
of each member are given as

Figure 5. Single module frame geometry for discrete optimization problem with unknown
bracing height,
Assuming each member to have a constant stress,
displacement can be written as

, (as previously demonstrated), the tip

Thus, the tip displacement, or compliance in this problem, is minimal when

(
( (
(

)
)

)
)

So finally, the optimal brace work point height gives the minimum deflection when

The rationale behind this solution can be explained by considering the topology optimization of a
simple structural frame with a point load applied to the top left corner and symmetry constraints
(see Figure 6). This problem (a simplification of a cantilever beam representing a high-rise
problem) run with topology optimization does not lead to the 45 degree diagonalization, as one
might expect, but rather to the high-waisted cross-brace with a working point at 75% of the
structures height. The cantilever (high-rise) problem must also account for the overturning
moment,
, which is not present in the pure shear problem, causing a vertical shift in the
intersection of the braces. (Note: the results presented here are dependent on the assumption of
constant stresses in the members.)

Figure 6. Comparison of optimal geometries for a cantilever problem (left) and a pure
shear problem (right)
Similarly, the optimal geometry analysis can be extended for the case of multiple
modules as is shown in Stromberg et. al (2011b), where the optimal bracing work point location
of each module still remains at 75% of the module height.

NUMERICAL RESULTS
Numerical results are given here based on a code written by the authors in C++ using the
methodology described in this work for a realistic building system. The structural frame is
modeled in Figure 7 by assuming the beam elements are not engaged vertically by the
quadrilateral mesh, while the column elements are attached continuously along their length.
Furthermore, the connections from the beams (used to represent the floor levels) to the columns
are pinned for the prototypical high-rise building. This structure is loaded with uniform point
loads at each module. The volume of material for the topology optimization problem is 30% (i.e.
30% of the design domain will be filled with solid material).
The resulting braced frame geometry shows working points near 75% of each modules
height, as was derived analytically in the previous section. The discrepancy between the
numerical results shown here and the analytical derivation is due to the nonuniformity of stress
in the optimized solution. This behavior was explained by Bendsoe and Sigmund (2002) by
noting that the strain energy density is constant for the intermediate densities but not for all
possible values (e.g. the extreme regions of solid and void, or 1 and 0). In Figure 4, this behavior
is shown by observing the Von Mises stresses are constant for the intermediate densities.
Another interesting feature in the optimized structure of Figure 7 is that the size of the
bracing members increases from the top to the base of the building as expected, due to the
increase in shear forces. Note that the discrete columns are sized a priori to increase in size along
the height, in accordance with gravity loading.

(a)

(b)

Figure 7. Braced frame design using topology optimization: (a) problem statement, (b)
optimization results using continuum (black) and discrete elements sized a priori (red)
CONCLUSIONS
The methodology described in this work provides structural engineers with an effective and
efficient means to design the lateral bracing systems for high-rise buildings by using topology
optimization with a combination of continuum (Q4) and discrete (beam) elements. The main
contributions of this work are as follow:

A method to describe optimal braced frames was proposed.

The constant state of stress in the members of an optimized frame was verified.
The relevance of this work has been demonstrated in the structural engineering industry.
The optimal geometry of these frames was described analytically and numerically.

REFERENCES
Allahdadian, S., Boroomand, B. (2010) Design and Retrofitting of Structures under Transient
Dynamic Loads by a Topology Optimization Scheme. In: Proceedings of the 3rd
International Conference on Seismic Retrofitting. October; p. 19.
Baker, W. F. Energy-Based Design of Lateral Systems. (1992) Structural Engineering
International, 2:99-102.

Bendsoe, M.P. Optimal shape design as a material distribution problem. Structural Optimization
1989;1(4):193-202.
Bendsoe, M.P., Sigmund, O.. Topology Optimization: Theory, Methods and Applications.
Springer; 2002.
Bendsoe, M.P., Sigmund, O. Material interpolation schemes in topology optimization. Archive
of Applied Mechanics 1999;69(9-10):635-654.
Hill, R.. The Mathematical Theory of Plasticity. New York: Oxford University Press Inc.; 1950.
Huang, X., Xie, Y.M. (2008) Topology optimization of nonlinear structures under displacement
loading. Engineering Structures; 30(7):20572068.
Huang, C., Han, X., Wang, C., Ji, J., and Li, W. (2010) Parametric analysis and simplified
calculating method for diagonal grid structural system. Jianzhu Jiegou Xuebao/Journal of
Building Structures, 31(1):70-77.
Lubliner, J. (1990). Plasticity theory. New York: Macmillan Publishing Company.
Neves, M.M., Rodrigues, H., Guedes, J.M. (1995) Generalized topology design of structures
with a buckling load criterion. Structural and Multidisciplinary Optimization 10(2):71
78.
Rozvany, G.I.N., Zhou, M., Birker, T. Generalized shape optimization without homogenization.
Structural and Multidisciplinary Optimization 2001;22(2):116-124.
Stromberg, L.L., Beghini, A., Baker, W.F., Paulino, G.H. (2011a) Application of layout and
topology optimization using pattern gradation for the conceptual design of buildings.
Structural and Multidisciplinary Optimization; 43(2):165180.
Stromberg, L.L., Beghini, A., Baker, W.F., Paulino, G.H. (2011b) Topology Optimization for
Braced Frames: Combining Continuum and Discrete Elements. Engineering Structures,
(under review).
Zhou, M., Rozvany, G.I.N. The COC algorithm, Part II: Topological, geometrical and
generalized shape optimization. Computer Methods in Applied Mechanics and
Engineering 1991; 89(1-3):309-336.