Cleveland State University

EngagedScholarship@CSU

Civil and Environmental Engineering Faculty Civil and Environmental Engineering

Publications

2012

Structural Topology Optimization: Moving Beyond Linear Elastic

Design Objectives
James K. Guest
Johns Hopkins University, [email protected]

Reza Lotfi
Johns Hopkins University

Andrew T. Gaynor
Johns Hopkins University

Mehdi Jalalpour
Cleveland State University, [email protected]

Follow this and additional works at: https://engagedscholarship.csuohio.edu/encee_facpub

Part of the Civil Engineering Commons, and the Structural Engineering Commons
How does access to this work benefit you? Let us know!

Original Citation
Guest, J.K., Lotfi, R., Gaynor, A., Jalalpour, M., "Structural topology optimization - moving beyond linear
elastic design objectives", Structures Congress 2012; March 29-31, 2012, Chicago, Illinois, p. 245-256.

This Conference Proceeding is brought to you for free and open access by the Civil and Environmental Engineering
at EngagedScholarship@CSU. It has been accepted for inclusion in Civil and Environmental Engineering Faculty
Publications by an authorized administrator of EngagedScholarship@CSU. For more information, please contact
[email protected].
 Structural Topology Optimization:
Moving Beyond Linear Elastic Design Objectives

James K. Guest, Reza Lotfi, Andrew T. Gaynor, Mehdi Jalalpour

Department of Civil Engineering, The Johns Hopkins University, 3400 North Charles
Street, Baltimore, MD 21218; PH (410) 516-3923; FAX (410) 516-7473;
[email protected]

ABSTRACT
Topology optimization is a systematic, free-form approach to the design of structures.
It simultaneously optimizes material quantities and system connectivity, enabling the
discovery of new, high-performance structural concepts. While powerful, this design
freedom has a tendency to produce solutions that are unrealizable or impractical from
a structural engineering perspective. Examples include overly complex topologies
that are expensive to construct and ultra-slender subsystems that may be overly
susceptible to imperfections. This paper summarizes recent tools developed by the
authors capable of mitigating these shortcomings through consideration of (1)
constructability, (2) nonlinear mechanics, and (3) uncertainties.

INTRODUCTION
The potential for topology optimization as a powerful, free-form design tool
has long been demonstrated in literature. The freedom to add (or remove) material
resources to (or from) any point in the design domain means material quantities and
system connectivity are simultaneously optimized – a powerful concept. While this
design freedom enables the discovery of high performance solutions, it often leads to
structural designs that are impractical from engineering and/or construction points of
view, or that are overly sensitive to uncertainties such as imperfections or random
loads. These flaws tend to be amplified by the fact that most topology optimization
works consider design objectives and specifications associated with linear elastic
structural behavior. This paper presents recent efforts by the authors for mitigating
these shortcomings. Specifically, methodological and algorithmic tool development
to optimize structural topology for nonlinear structural behavior, constructability
objectives, and for robustness in the presence of uncertainties.
The first step in topology optimization is discretization of the design
domain. This meshing must ultimately be driven by the governing mechanics, but
generally uses either solid (continuum) finite elements or discrete members such as
truss or frame elements. In continuum topology optimization, the goal is to determine
the optimal phase composition in each elemental domain. In traditional solid-void
design, each element carries an indicator variable ρe, known as the volume fraction,
that identifies the element as either solid material (ρe=1) or a void (ρe=0). The
connectivity of the solid elements defines the topology of the structure. Such an
approach may be appropriate, for example, in the design of concrete or steel
components. Discrete element topology optimization, on the other hand, typically
follows a ground structure approach, where the domain is discretized with a dense,
redundant structural system and members deemed inefficient are removed from this
ground structure.
While straightforward in concept, topology optimization problems are
fundamentally and numerically challenging to solve. This has led the vast majority of
research to focus on linear elastic and deterministic mechanics. The most commonly
solved topology optimization formulation is minimum compliance, where the goal is
to minimize internal strain energy (maximize stiffness) in a structure of fixed mass
for given load case and boundary conditions. A design example of a simply
supported beam loaded at midspan is shown in Figure 1. A simple design containing
six uniformly spaced holes is shown in Figure 1b. This design features a 50% volume
fraction and undergoes a vertical deflection d at midspan. Figure 1c displays a
topology-optimized solution using the same volume of material but offering an
increased stiffness (reduced deflection) of approximately 42%. The design follows
the principal stress trajectories with compression and tension members lying
orthogonal. Alternatively, Figure 1d displays a solution having the same stiffness as
the simple beam with regular holes (deflection d) but at a reduced weight of
approximately 50%.
While these improved performance metrics are compelling, the optimized
topologies clearly illustrate the deficiencies that have inhibited the application of this
tool to structural engineering. The topology in Figure 1c is certainly more
challenging to fabricate than the solution in Figure 1b, while the topology in Figure
1d may offer the same elastic stiffness as the simple design but will more readily
enter the nonlinear regime due to slender compressive load paths and will be more
susceptible to flaws.

?
? Ω
     
 
(a) Design Domain (b) Simple Design:
Weight = W, max deflection = d

     
(c) Optimized Topology: (d) Optimized Topology:
Weight=W, max deflection ~ 0.58 d Weight~0.50 W, max deflection = d

Figure 1. Simply supported beam example. (a) Design domain and continuum mesh,
(b) simple 50% volume fraction design with uniformly spaced holes, (c) topology
optimized design offering same weight as (b) but offering 42% improved stiffness,
and (d) topology optimized design offering same linear elastic stiffness as (b) but at
half the weight.
OPTIMIZATION FORMULATION: LINEAR ELASTIC MECHANICS

We begin with a brief review of the linear elastic minimum compliance

formulation, where the goal is to maximize stiffness of a fixed-mass system by
minimizing internal strain energy or equivalently external work. The independent
design variable for the topology optimization problem is denoted as ρe and represents
element cross-sectional areas in truss topology optimization and element volume
fractions, in continuum domains. The response nodal displacements d are the state
(dependent) design variables, and are related to the design through the equilibrium
conditions. The design problem is given in general as

1 T 1
min d K(ρ e ) d = f T d
e
ρ ,d 2 2
e
subject to : K(ρ ) d = f
(1)
∑ρe ve ≤ V
e∈Ω

0 ≤ ρ e ≤ ρmax
e
∀ e ∈Ω

where f are the nodal applied loads, ve is element volume for unit ρe (element length
e
for truss structures), V is the available volume of material, and ρ max is the upper
e €
bound on ρ . The global stiffness matrix K is assembled in the usual manner from
element stiffness matrices Ke, which are related to the design variables as follows:

! €
K e ( ! e ) = ( ! e ) K e0 (2)

where Ke0 is the element stiffness matrix for unit ρe and the exponent parameter η ≥ 1
is an optional penalty term that may be used to drive solutions to the design variable
bounds (Bendsøe 1989). The sensitivity analysis is performed using the adjoint
method (see e.g., Bendsoe and Sigmund (2003)) and Method of Moving Asymptotes
(MMA) (Svanberg 1987) is used in this work as the gradient-based optimizer.

CONSTRUCTABILITY

Maintaining constructability of topology-optimized solutions is critical to

ensuring designs are cost effective and useable. The primary challenge here is
quantifying constructability and, in the case of continuum design, computationally
detecting structural members, or features, which are defined by the union of elements
of like phase. Significant progress has been made in this arena through geometric
restriction of the design space. For example, by controlling the minimum length scale
(diameter) of designed members (Guest et al. 2004, Guest 2009a), maximum length
scale of members (Guest 2009b), and pattern repetition (Stromberg et al. 2011). By
tightening these restrictions, the designer is able to reduce design complexity and
enhance constructability.
 This idea has been demonstrated in building design concepts (Stromberg et al.
2010) and, at the 19th Analysis and Computation Specialty Conference, the
optimization of concrete reinforcement patterns via strut-and-tie modeling (Guest and
Moen 2010). We focus here on the latter, where topology optimization can be used to
idealize the flow of forces in concrete members and automate generation of strut and
tie models, where resulting tensile zones indicate regions for steel reinforcement or
prestressing placement (Schlaich, 1987). For example, we consider the concrete
hammerhead pier design problem presented in Guest and Moen (2010) where it was
demonstrated that allowing smaller diameters of structural members leads to stiffer
structures, but structures with more complex reinforcing steel geometries. This is
shown in Figure 2, where the number of steel ties decreases from nine to five when
the minimum allowable diameter increases from r to 6r. This trend increases the steel
material cost but reduces the number of reinforcing members to be placed, thereby
lowering labor cost. Although difficult to quantify, the trends are evident.

(a) (b) (c)

Figure 2. Hammerhead pier example solved continuum topology optimization and

minimum prescribed length scales of (a) r , (b) 4r , and (c) 6r. Smaller length scales
offer higher performance but typically increase design complexity
(Guest and Moen 2010).

A limitation of this idea, however, is that the shape of the reinforcement

remains generally unconstrained (e.g., Liang et al. 2000, Bruggi 2009). In other
words, reinforcement bars may be curved as seen in Figure 2. Such a design may be
acceptable in the case of prestressing strands, but may be problematic or expensive
for use with steel rebar. To circumvent this issue, we propose here a hybrid approach,
which combines the truss and continuum topology optimization approaches. The idea
is that steel reinforcement, which needs to have simple layouts and straight members
in order to be cost efficient, is better idealized using truss elements. The concrete in
the strut and tie model is an idealization of the flow of compressive forces, and thus
need not be explicitly ‘constructed’, allowing geometric restrictions to be relaxed.
Continuum elements are therefore used to model the concrete phase.
A bilinear constitutive model is adopted where truss elements offer high
tensile stiffness but low compressive stiffness, while continuum elements offer high
compressive stiffness but low tensile stiffness, a valid assumption if the concrete in
tension is assumed cracked. The concrete is idealized using the model suggested by
Darwin and Pecknold (1977). To achieve this effect, the design domain is discretized
with a lattice mesh of nodes. The continuum mesh uses every node, while the truss
mesh is more sparse with members connected at every few nodes in order to reduce
complexity of the final steel configuration. This is seen below in Figure 3 where there
are 12 continuum elements in each direction, but the truss elements are connected
every 4 nodes (green continuum elements, blue truss elements). Force transfer
between the meshes occurs at the shared nodes, and a nonslip condition for the steel
reinforcement is assumed.

Figure 3. Interaction between continuum and truss domains

The following are some preliminary results from the hybrid optimization. The
first example is a simply supported beam with a point load at the center, where
symmetry has been used to reduce the computational cost (Figure 4). As can be seen,
the tension steel is located in the bottom portion of the beam and is located along the
principal stress trajectories. This is also seen in Figure 1, with the noticeable
difference here being that the tensile members are sparse and straight, as would be
desired to reduce labor costs.

Figure 4. Simply supported beam design domain (left) and optimized strut and tie
topology (right).

Topology optimization is especially useful for more complex geometries,

where the flow of forces is difficult to visualize, let alone optimize. Figure 5 shows a
modified dapped end beam with a hole in the middle that presents such a
complication. The optimal solution is shown in Figure 5, where again the concrete
compression zones are represented by the continuum structure and the tensile
reinforcement pattern is restricted to be sparse and composed of straight members.
While these examples highlight the capability of topology optimization to create
strut-and-tie models from scratch, this tool can also be used to ‘reinforce’ a
traditional, uniform rebar pattern. One would simply include the desired
reinforcement pattern as a subset of the truss mesh and keep the associated cross-
sectional areas fixed (or bounded) to maintain their existence throughout the
optimization.
 Figure 5. Modified dapped end beam with hole design domain (left) and optimized
strut and tie topology.

NONLINEAR BEHAVIOR

Extending topology optimization to design under nonlinear mechanics is

another challenging but important endeavor, particularly given the increasing use of
performance-based design methodologies. As a first step, several researchers have
proposed using linearized eigenstability metrics for optimizing structural stiffness
(see e.g., Rahmatalla and Swan, 2003). Formal incorporation of geometric and/or
material nonlinearities requires iterative analysis and a constraint on the residual R of
equilibrium equations. Final compliance, cumulative compliance, and complimentary
elastic work are have been suggested as objective functions by Buhl et al (2000). The
drawback of final compliance as the objective function is that the structure may
collapse under the load before reaches the end point. Cumulative compliance as the
objective function is defined as

min f = %% " T
d#d$
$ (3)

where it should be evident that the nonlinear optimization is now a path dependent
problem.
!
We begin by considering geometric nonlinearities. The residual is defined as

T
R(U) = P ! " B ! dV (4)

Strains are related to the displacements as follows

d! = B(U)!d(U) (5)

where we use advanced B matrix to get higher order strains for Green-Lagrangian
strains
1
! ij = (ui, j + u j,i + uk,i uk, j )
2 (6)
 
 This system of nonlinear equations is incorporated into the finite element and
sensitivity analyses and is solved using displacement control iterative method. To
demonstrate the effect of geometric nonlinearity on optimal designs, consider the
simple cantilever structure shown in Figure 6. We optimize for cumulative
compliance under various load magnitudes. It is clearly seen that considering
geometric nonlinearities leads to solutions that are load-dependent and, potentially,
asymmetric. The latter can be understood by examining the structural member at the
tip of the beam in Figures 7b and 7c. As the beam deflects, this member will become
vertical and aligned with the applied load, creating a highly efficient tension element.
 

 
Figure 6. Cantilever beam design domain.

(a) Load = P

(b) Load = 2P

(c) Load = 2.5 P

Figure 7. Cantilever beam topology optimized solutions for various load
magnitudes under assumptions of (left) linear mechanics and (right) geometric
nonlinearity.

The consideration of material nonlinearities is more complex, as it requires

assumptions and sensitivity analysis of post-yield behavior. We use here an
elastoplastic model and, following the work of Maute et al. (1998), relate design
variable ρe to material properties and constitutive tensors following the SIMP logic
(equation (2)) as
D = !i" D0 !!!!!!!!!! H = !i" H 0 !!!!!!!! # y = !i"# y0 (7)

where D is the elastic constitutive matrix, H is the plastic hardening, " y is the yield
stress. For the material nonlinearity, Von-mises Yield surface is used.

!
 F = 3J 2 " k (8)
k = "y + H (9)
1
where J 2 = Sij S ji , H is the hardening and "y is the yield stress.
2 !
To demonstrate the effect
! of including material nonlinearities on final designs,
consider the fixed-fixed beam design domain shown in Figure 8. The goal here is to
! minimize !cumulative strain energy! under a prescribed midspan deflection of 0.2 m.
According to symmetry properties of the example, only one half of the structure is
solved.

Figure 8. Fixed-fixed beam design domain.

The optimized topology under linear elastic and elastoplastic material models
are shown in Figure 9, superimposed on the load-deflection response for each
strcutre. The topology resulting under the elastic-plastic material model diversifies
the load path resulting in a lower stiffness in the elastic regime but higher stiffness in
the nonlinear regime due to the structural redundancies, as the plot demonstrates.

Figure 9. Load-deflection curves for topology optimized beams under conditions of

elastic material model and elastic-plastic material model.
DESIGN UNDER UNCERTAINTY

Uncertainty in applied loads is often a governing issue in the design of civil

engineering structures. Formal incorporation of such randomness into the design
optimization framework has recently become a strong focus of the topology
optimization community (e.g., Kharmanda et al. 2004, Nguyen et al. in press).
Receiving less attention has been consideration of uncertainties in structural stiffness.
Such uncertainties arise due to construction and fabrication errors, such as geometric
imperfections, construction misfit, and member misalignment, or environmental
degradation and damage. The effect of such errors or damage is amplified in
structures that are designed using deterministic structural optimization, as such
techniques usually produce designs consisting of slender members and structural
systems.
The challenge of incorporating random structural stiffness is that the structural
response (displacements) are a function of the inverse of a random matrix. This
makes the use of simulation-based strategies, such as Monte Carlo simulation,
computationally inefficient as each realization results in a different global stiffness
matrix (and thus linear system) that must be solved. A key focus of the authors has
been in developing intrusive methods that couple the uncertainty quantification with
the design sensitivity analysis. These have included the use of perturbation (Guest
and Igusa, 2008; Asadpoure et al., 2011) and Polynomial Chaos expansion
(Tootkaboni et al., 2012) to optimize structures that are robust in the presence of
uncertainty. The robust formulation replaces a deflection metric, for example, with a
probabilistic formulation composed of the expected (mean) value of the deflection
plus the standard deviation of deflection as follows:

!"# ! ! +  !  !"# ! (10)

where E[] and std[] are the expected value and standard deviation operators,
respectively, and k is a designer-selected positive scalar that indicates the importance
of controlling response variability (with larger k meaning stricter control on response
variability).
Of particular interest to the design of structures is the recent coupling of
stability analysis with the perturbation-based optimization under uncertain stiffness
methodology. Jalalpour et al. (2011) proposed an algorithmic modification that
mathematically accounts for the fact that compression amplifies the effect of
geometric imperfections while tension dampens it. This led to optimized designs that
were dependent on the uncertainty source distribution, the applied load direction, and
the applied load magnitude. The work considered mean stiffness only (k = 0), which
we extend here to consider robust design objectives with k > 0.
To illustrate the effectiveness of the formulation, we optimize the cantilever
truss structure shown in Figure 10 while considering 5% randomness in all nodal
locations to represent the geometric imperfections. The solutions optimized for mean
stiffness only (k=0) and mean plus three standard deviations of stiffness (k=3)
considering stability are shown in Figure 10. It is immediately evident that the tensile
zones are relatively light and sparse, while the compression zones feature diversified
and braced load paths to mitigate the imperfection-induced onset of buckling.
Therefore, including the provisions for resisting buckling leads to designs that have
bracings where it is needed the most. We note this diversification and bracing
systems is amplified when the designer increases the importance of structure
robustness by increasing the variable k.

Figure 10. Cantilever design problem (left) and robust topologies optimized for mean
stiffness (center) and mean plus three standard deviations of stiffness (k=3) (right) in
the presence of geometric imperfections. Considering structural stability leads to
diversification and bracing of the compressive load paths.

CONCLUDING REMARKS

While topology optimization has a demonstrated history of generating new,

high performance design ideas, it has generally seen limited, albeit increasing use, by
the structural engineering community. The primary reason for this slow adaptation is
the reliance of many works on the assumptions (1) that material cost (weight) drives
design, and that structural behavior is governed by (2) linear elastic and (3)
deterministic mechanics.
This paper summarized recent work by the authors aimed at alleviating these
restrictive assumptions, improving the applicability of topology optimization to
structural engineering design problems. Specifically, computationally efficient
algorithmic tools for (1) influencing the constructability of systems and
manufacturability of components, (2) optimizing design under geometric
nonlinearities and nonlinear material models, and (3) optimizing robustness of
designs under the possibility of construction or fabrication errors, or damage,
including the potential stability-induced amplification of such uncertainties.
ACKNOWLEDGEMENTS

This work was supported by the National Science Foundation under Grant No.
CMMI-0928613 with Dr. Christina Bloebaum serving as program officer. The
authors also thank Cris Moen (Virginia Tech) for his input on the strut and tie work
and Tak Igusa (JHU) for input on the uncertainty work.

REFERENCES

Asadpoure, A., Tootkaboni, M., Guest, J.K. (2011). “Robust topology optimization of
structures with uncertainties in stiffness – Application to truss structures.”
Computers & Structures 89(11-12): 1131-1141.
Bendsøe, M.P. (1989). "Optimal shape design as a material distribution problem."
Structural Optimization, 1, 193-202
Bendsøe, M.P., Sigmund, O. (2003). Topology Optimization: Theory, Methods, and
Applications, Springer, 2003.
Bruggi, M. (2009). “Generating strut-and-tie patterns for reinforced concrete
structures using topology optimization.” Computers and Structures, 87(23-
24), 1483-1495.
Buhl, T., Pederson, C.B.W., Sigmond, O. (2000). “Stiffness design of geometrically
nonlinear structures using topology optimization.” Structural and
Multidisciplinary Optimization, 19(2), 93-104
Darwin, D., Pecknold, D. A. (1977). “Nonlinear biaxial stress-strain law for
concrete.” ASCE J Eng Mech Div, 103(2), 229-241.
Guest, J.K., Prevost, J.H., Belytschko, T. (2004). "Achieving minimum length scale
in topology optimization using nodal design variables and projection
functions." International Journal for Numerical Methods in Engineering,
61(2), 238-254.
Guest, J.K., Igusa, T. (2008). “Structural optimization under uncertain loads and
nodal locations.” Computer Methods in Applied Mechanics and Engineering,
198(1), 116–124.
Guest, J.K. (2009a). “Imposing maximum length scale in topology optimization.”
Structural and Multidisciplinary Optimization, 37(5): 463–473.
Guest, J.K., (2009b). “Topology optimization with multiple phase projection.”
Computer Methods in Applied Mechanics and Engineering, 199(1–4), 123–
135.
Guest, J.K., Moen, C.D. (2010). “Reinforced concrete design with topology
optimization.” Proceedings of the 19th Analysis and Computation Specialty
Conference, 445-454.
Jalalpour, M., Igusa, T., Guest, J.K, (2011). “Optimal design of trusses with
geometric imperfections: Accounting for global instability.” International
Journal of Solids and Structures, 48(21), 3011-3019.
Kharmanda, G., Olhoff, N., Mohamed, A., Lemaire, M. (2004). “Reliability-based
topology optimization.” Structural and Multidisciplinary Optimization, 26,
295-307.
Liang, Q.Q., Xie, Y.M., Steven, G.P. (2000). "Topology optimization of strut-and-tie
models in reinforced concrete structures using an evolutionary procedure."
ACI Structural Journal, 97(2), 322-330.
Maute, K., Schwarz, S., Ramm, E., (1998). “Adaptive topology optimization of
elastoplastic structures” Structural and Multidisciplinary Optimization, 15(2),
81-91.
Nguyen, T.H., Song, J., Paulino, G.H. (in press). “Single-loop System Reliability-
Based Topology Optimization Considering Statistical Dependence between
Limit-States.” Structural and Multidisciplinary Optimization.
Rahmatalla S, Swan C.C. (2003). “Form finding of sparse structures with continuum
topology optimization.” Journal of Structural Engineering 129:1707-1716.
Schlaich, J., Schaefer, K., Jennewein, M., (1987). "Toward a consistent design of
structural concrete." PCI Journal, 32(3), 74-150.
Stromberg, L.L., Beghini, A., Baker, W.F., Paulino, G.H. (2011) “Application of
layout and topology optimization using pattern gradation for the conceptual
design of buildings. Structural and Multidisciplinary Optimization, 43(2),
165-180.
Svanberg, K., (1987). “The method of moving asymptotes – a new method for
structural optimization.” International Journal for Numerical Methods in
Engineering, 24, 359-373.
Tootkaboni, M., Asadpoure, A., Guest, J.K. (2012). “Topology Optimization of
Continuum Structures under Uncertainty – A Polynomial Chaos Approach.”
Computer Methods in Applied Mechanics and Engineering, 201-204, 263-
275.