International Journal of Machine Tools & Manufacture 84 (2014) 33–44

Contents lists available at ScienceDirect

International Journal of Machine Tools & Manufacture

journal homepage: www.elsevier.com/locate/ijmactool

Stiffness design of machine tool structures by a biologically

inspired topology optimization method
Baotong Li a, Jun Hong a,n, Zhifeng Liu b
a
State Key Laboratory for Manufacturing Systems Engineering, School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an 710049, PR China
b
School of Mechanical and Automotive Engineering, Hefei University of Technology, Hefei 230009, PR China

art ic l e i nf o a b s t r a c t

Article history: This paper introduces a novel approach for designing the stiffener layout inside large machine tools by
Received 29 January 2014 applying the self-optimal growth principle of plant ramifications in nature. Firstly, numerical studies are
Received in revised form carried out in order to confirm the potential of leaf venation as concept generators for creating the
25 March 2014
optimal load-bearing topology for stiffened machine tool structures. Then, a mathematical model
Accepted 26 March 2014
explaining the optimality of plant morphogenesis is presented. Based on this, an evolutionary algorithm
Available online 1 April 2014
is developed, which uses three growth strategies to determine the candidate stiffeners to grow or
Keywords: atrophy with respect to the loads applied. The proposed growth-based method could generate a distinct
Topology optimization stiffener layout, which is different to those produced by the conventional topology optimization
Machine tool structures
methods, and thus offers unique possibilities of improving the design efficiency and commonality for
Stiffener layout
machine tool development. The suggested approach is finally applied to the re-design of an actual
Adaptive growth
grinding machine column, on which the numerical analyses and experimental tests conducted exemplify
the performance enhancement, and therefore is a good choice for the stiffener layout design of machine
tool structures.
& 2014 Elsevier Ltd. All rights reserved.

1. State of the art
The increasing demand for machine tools to be more produc-
tive, ecological and cost efficient poses new challenges for
machine developers. Design activities of modern machine tools
not only strive for excellent mechanical performances but also
seek economic advantages for machine users [1–3]. On machine
tool structures, only a small fraction of the material applied is
critical to the kinematical functionality. This is because a majority
of the material is there to connect the functional interfaces
between individual components and therefore provides sufficient
stiffness to carry forces so that such kinematical functionality
could be fulfilled. For this reason, stiffness is no doubt a design
priority, and is likely to become more important for high speed
and high precision machine tools [4,5]. At this background, the
primary focus for the design of machine tool structures is to
increase stiffness whilst wherever possible keeping the compo-
nent weights constant or reducing them.
Since the arrangement of stiffener plates inside machine tool
structures could affect the overall stiffness greatly, how to deter-
mine the stiffener layout pattern is becoming a key issue. How-
ever, in traditional machine tool design, stiffener layout is usually
n
Corresponding author.
E-mail address: jhong@mail.xjtu.edu.cn (J. Hong).
selected based on manufacturing convenience, such as “X” shape,
“#” shape or other simple shape, which may not be the best way
to satisfy the criteria for required stiffness and material savings.
In addition, most of the past studies on stiffener optimization have
been limited to sizing optimization, where the shape and topology
of stiffener plates are held constant while only a set of thickness
parameters or spacing parameters is obtained [6–9]. Although
sizing optimization is very easy to implement and can guide
selection of reliable designs, it may also obscure new and better
choices. For instance, if a completely new stiffener arrangement
needs to be developed, this type of approach may not yield the
true optimal solution because of its inability to modify the
topology of stiffened structures.
To overcome the above shortcomings, it would be a wise way
for designers to seek guidance and inspiration from the natural
world [10,11]. A large number of successful biomimetic examples
applied in different branches of technology, such as robotic
systems [12], tribological products [13] and machining processes
[14], are based on the clever and elaborate morphology of
biological structures. Actually, optimization problems of stiffener
layout are not confined to engineering design, natural morpho-
genesis of living organisms also provides similar scenarios.
Specially, the ramified configurations in various biological struc-
tures, such as the leaf venation, root and axis system of plants are
candidates that may immediately come to mind. Through the
process of evolution, plants have been experimented with various

http://dx.doi.org/10.1016/j.ijmachtools.2014.03.005
0890-6955/& 2014 Elsevier Ltd. All rights reserved.
34 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
load-bearing topologies for at least billions of years before the first
human attempts in topology optimization. Although we are unable
to confirm the extent to which plant models inspired our early
ancestors, more recent examples of bionic applications for
mechanical design are well documented. In the last few decades,
the research in the Biomechanics Group Forschungszentrum
Karlsruhe concentrated upon developing a method that mimics
the growth principles of buttress roots in trees for shape optimi-
zation [15,16]. Milwich et al. developed a “technical plant stem”,
a biomimetic product inspired by mechanical properties found in
hierarchically structured plant ramifications [17]. Ding and Yama-
zaki proposed a growth technique for innovative design of heat
transfer systems by employing the optimality of biotic branch net
in the plant world [18].
Although the morphologies and growth environments are
different among various plant ramifications, they have one thing
in common, that is their topologies are formed by a continuous
growth process [19,20]. Compared with the conventional methods
in topology optimization, the growth-based approach is more
applicable to the layout design problem, because it can produce
a distinct stiffener distribution rather than a vague material
distribution, which leads to a considerable increase in design
efficiency. Based on this understanding, the author and his
colleagues performed an intensive study on adaptive growth
simulation, which employs an evolutionary algorithm for support-
ing the generation of stiffener layout [21]. However, such algo-
rithm is usable for structures with only one specified growth
region, like the simply supported plate structure, it is not very
helpful when dealing with large scale structural systems com-
posed of several substructures. For example, the stiffener layout
inside a machine tool column usually involves two parts, one is
for the front/rear panels and the other is for the side panels. If the
existing algorithm is utilized, these substructures will be treated
individually, which cannot guarantee the global optimality of the
growth results obtained. A comprehensive coordinative optimiza-
tion model is therefore, essential to correct this situation.
The work presented in this paper is undertaken to extend the
capability of adaptive growth method towards global coordinative
optimization problems and investigate contributions the growth
simulation can make to the design of large scale machine tool
structures. The reminder of this paper is organized as follows.
In Section 2, the potential of leaf venation as concept generators
for stiffener layout design of machine tool structures is confirmed.
In Section 3, a mathematical model explaining the optimality of
plant morphogenesis is presented. Then, an improved evolution-
ary algorithm is developed based on a global coordinative model
so as to alleviate the programming and computing efforts
for implementing the generation of stiffener layouts for machine
tool structures. The re-design of an actual grinding machine
Fig. 1. Similarity between leaf venation and stiffened machine tool structures.
column is conducted in Section 4. Results before and after bionic
optimization are compared in terms of numerical and experimen-
tal studies. The main findings of the research work are concluded
in Section 5.
2. Biomechanical significance of leaf veins in nature
After billions of years of selection, nature has evolved many
ingenious topologies, providing inspiration for designers to solve
engineering problems. One prominent category of such topologies
is the hierarchically structured plant ramifications like leaf vena-
tion, whose mechanical stiffening role has been widely recognized
by both biologists and engineers [22–24]. The leaf veins extend
fully in order to get enough surface stiffness to bear the dead
weight and the environmental loads. The stiffener plates inside
machine tools also require high stiffness to withstand a surface
load so as to guarantee the machining accuracy. While such a
question – ‘What can we learn for machine tool design from
natural leaf venation?’ – is the starting point of this research, the
solution is not limited to a copy or mimicry of a particular
structure, but the common configuration ideas that could be used
for a large number of products and may even generate new
technical applications for machine tool development.
To move between bionic ideas and their concrete implementa-
tion, a designer needs some methods to bridge the gap between
biology and engineering. However, it is difficult to give a precise
assessment because biological systems are more complex than
engineering systems. Assume that bio-solutions to engineering
problems exist everywhere in nature, we only need to determine
what analogous function to search for and at what level to search
for it. Basically, there are three levels at which analogical analysis
could be considered, namely structure (topology or shape), loading
and function similarities [25]. The prototype that can serve the
engineering purpose for stiffener layout design is outlined in the
flow diagram shown in Fig. 1.
(1) Structure (topology or shape) similarity is critical for analogi-
cal prototype selection, since biology is a self-optimal system
for its efficient material distribution. A plant leaf can be
viewed as a lightweight cantilever, which is permeated with
mechanically stiffening veins. Branching strategy has been
developed for leaf venation in order to guarantee the global
rigidity and local isotropy. For machine tool structures, the key
components such as bases, columns, carriages and tables are
usually box beam structures, in which the stiffener plates are
branched horizontally and vertically so as to enhance the
overall stiffness. It can be found that hierarchy and reticulum
 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
Fig. 2. The load-bearing topology of sample leaves. (a) Plant leaf with pinnate venation. (b) Plant leaf with palmate venation.
are the common features for both leaf venation and stiffener
plates.
(2) Loading similarity is also important for a successful bionic
design, because biology knows how to define its load paths so
as to make it possible for best load transmission. There are two
main forms of stress inside leaf structures, the bending stress
and the torsional stress, which are caused by the dead weight
and the environmental loads like the wind, rain and snow.
Similarly, the bending and torsional loads are also the major
considerations for designing the layout of the inner stiffener
plates, because most of the tensile and compression loads are
resisted by the outer panels of the machine tool structures.
(3) Function similarity is vital for merging biology with engineer-
ing. Assuming elastic deformation energy could be viewed as
energy flow, the morphogenesis of leaf venation is actually a
minimization route of the energy flowing consumption, which
also means the pursuit of maximum structural stiffness. It can
be expected that the key components of machine tools will get
higher stiffness if their inner stiffener plates extend following
a similar morphogenesis principle as leaf venation in nature.
The branching pattern of leaf venation functions as the optimal
mechanical reinforcement for photosynthesis and transpiration.
Whereas, how does such type of topology structure adapt to the
environmental loads? Does it have any meaning in guiding the
stiffener layout design? To answer these questions, the morpho-
logical change of leaf venation is simulated from the view point of
topology optimization. Generally, leaf venation can be divided into
two types, pinnate venation and palmate venation [26]. The
former type has only one principal vein, while the latter type
has more than one principal vein. To make the simulation typically
representative, the plant leaves with pinnate venation and pal-
mate venation are selected respectively in the following analysis.
The sample leaves are modeled by shell elements with the tool
of ANSYS APDL, and their mechanical properties are assumed to be
isotropic, homogeneous and linear elastic. Taking the petiole as the
supporting point, and all of its freedom is constrained. Leaf
venation is a complex vegetation process contributed by different
loading conditions. Maybe several loads define its structure,
the major one can be extracted for mechanical analysis. In this
35
research, the environmental loads are simulated by a uniformly
distributed pressure (wind load).
Assume that the morphological change of leaf venation is slow
enough, the elastic equilibrium can be reached at any discrete
moment. In the simulation, elastic equilibrium means the seeking
of minimum elastic energy, and the constraint function is defined
as volume fraction, which is restricted to 0.7V0, 0.5V0 and 0.3V0,
where V0 denotes the original volume of the leaf lamina. The
topology optimization is conducted based on the software Hyper-
mesh, and Fig. 2 shows the changing history of the load-bearing
topology of sample leaves. Some interesting observations can be
made as follows.
(1) The load-bearing topology is becoming increasingly clear with
the gradual elimination of redundant material from the leaf
region. The simulation results display a level of resemblance
comparable to the real leaf venation. The branching pattern of
material elements helps to keep the strain energy of the leaf
structure at a small value. The diameter of principal vein
decreases gradually along its growing direction, and each
principal vein has more lateral veins with its radius decreasing
so as to maintain enough surface stiffness to bear the
environmental loads.
(2) Leaf venation exhibits higher and active adaptability to
achieve better performance, such adaptability would be
critical for survival in a harsh environment. Branching angles
of 01 to 901 are observed between the principal vein and the
lateral vein, while most of them varied within a range from
301 to 601. Actually, when the load takes place, the leaf vein
appears where the stress seems like a communicator between
gene and topology pattern in plant morphogenesis.
According to the above analysis and observation, it can be
found that leaf venation plays a role of skeletal supporter resemble
to that of the stiffener plates inside machine tool structures.
A venation growth algorithm, which is inspired from morphogen-
esis mechanics and labels as hierarchy and reticulum, could be
used as a concept generator for the expression of stiffener layout
in machine tool structures.
36 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44

Fig. 3. Finite element model of leaf lamina in different scales.

3. Simulation strategies for adaptive growth of stiffener layout According to the Kuhn–Tucker conditions [27,28], the following
equations should be satisfied at the optimal solution:
3.1. Explanation on the optimality of plant morphogenesis 8
>
> r 0; W i ¼ W Ui ;
∂L ∂F ∂½G  W g <
In this section, new developments on the adaptive growth ¼ þλU ¼ 0; W i U 4 W i 4 W i L ð5Þ
∂W i ∂W i ∂W i > >
modeling are presented in detail, which include the concept, : Z 0; W ¼ W L;i i
calculations and evolutionary algorithm involved. The growth idea
proposed by the author [21] is adopted as the foundation concept. From Eq. (5), one can get
The prototype model is demonstrated in Fig. 3, where the leaf ∂F ∂G
Wi U þ λ UW i U ¼0 ð6Þ
lamina is simulated by a plate structure aligned with candidate ∂W i ∂W i
stiffeners of minimal cross-sectional areas. So the stiffeners would
Noticing that
have little effect on the stiffness of the entire structure in the !
initial stage of growth simulation. N
∂G=∂W i ¼ ∂ ∑ W i =∂W i ¼ 1 ð7Þ
The base plate is divided by the 4-node quadrilateral shell i¼1
elements, and the stiffeners formed by two adjacent nodes of
the shell elements are divided by the 2-node beam elements. The Eq. (6) can be rewritten as
cross-section of the stiffener is assumed to be a rectangle whose ∂F
width (ts) is constant and identical to the thickness of the base Wi U þ λ UW i ¼ 0 ð8Þ
∂W i
plate (hs ¼hb). Therefore, the weights of the stiffeners, which will
be selected as design variables in the growth simulation, are varied Thus, the weight of stiffener i is assumed to take the following
which merely depends on the heights of the stiffeners. form:
The way for a candidate stiffener to survive the growth 1
W i ¼ UDi ð9Þ
competition is to increase its mechanical performance whilst λ
wherever possible keeping the material consumption constant or where Di is the generalized weight of stiffener i, which can be
reducing it. The growth problem can be stated to minimize the expressed as
objective function subject to the stiffener weight constraint,
shown as follows: ∂F
Di ¼  W i U ð10Þ
∂W i
Growth objective : F- min ð1Þ Based on the above equations, one can get
D1 D2 D Dsum
¼ ¼⋯ i ¼⋯¼ ¼λ ði ¼ 1; …; NÞ ð11Þ
Resource constraint : GrWg ð2Þ W1 W2 Wi W sum
Eq. (11) provides mathematical insights into the growth com-
petition. Definite information is obtained on how to coordinate the
Growth variables : W i ði ¼ 1; 2; 3; …; NÞ ð3Þ growth of candidate stiffeners so as to improve the objective
function most effectively. Here, the generalized weight Di quanti-
where F is the growth objective function which usually denotes fies the potential of stiffener i to survive the competition for best.
the mechanical property, such as the strain energy, natural Based on this understanding, the optimum stiffener layout can
frequency, etc. G is the total weight allocated to all stiffeners at be considered as results of obeying such a common mechanism
one growth step and Wg is the corresponding upper limit. Wi is the that the growth rate for each stiffener is proportional to its
weight of stiffener i and N is the total number of stiffeners. generalized weight. When a new seed is extended from a sprout-
The basic idea to solve this problem is to envelop the objective ing point, the best direction for branching is selected depending
and constraint functions by a new function so that the constrained on the allocation of the generalized weight.
problem could be transformed into an unconstrained one, then From Eqs. (9) and (11), one can get the following iterative
iterates until approaching its optimum result. In this end, the formula:
Lagrange function is constructed as follows:  
W sum ðkÞ
W i ðk þ 1Þ ¼ Di U ði ¼ 1; …; NÞ ð12Þ
Dsum
L ¼ F þλ UðG  W g Þ ð4Þ
To guarantee the convergence of iterative calculation, a relaxa-
where λ is the Lagrange multiplier. tion factor α is introduced, and the final iterative formula could be
 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
updated as
 
W sum ðkÞ
W i ðk þ 1Þ ¼ αðkÞ U Di U þ ð1  αðkÞ Þ U W i ðkÞ ði ¼ 1; …; NÞ
Dsum
3.2. Global coordinative optimization model
To extend the capability of adaptive growth method towards
complex structural design applications, further efforts will be
devoted to explore a global coordinative optimization model so
as to alleviate the programming and computing efforts for imple-
menting the generation of stiffener layouts inside engineering
structures.
For a structural system composed of several substructures, the
global coordination for stiffener growth can be expressed in such a
way that the total amount of growth resources (material weight) is
Fig. 4. Global coordinative optimization model.
Fig. 5. Simulation strategy for adaptive growth of stiffener layout.
37
not allowed to exceed the given value, optimally allocating
material to its substructures, so that the performance functions
ð13Þ tend to their optimal values. This problem can be solved by a
decoupling–coordinating program, which is shown in Fig. 4. In this
mathematical model, M ¼[M1,…, Mk,…, MK]T is the design vector
of global optimization, Mk is the weight of material allocated to the
k-th substructure; Wk ¼[W1k,…, Wik,…, WIk]T is the design vector
of local optimization, Wik is the weight of stiffener i inside the k-th
substructure.
The optimization model consists of several local solvers LSk
(k ¼1,…,K) to find the optimal stiffener growth results of sub-
structures and a global solver GS to play a coordinative role among
these substructures. These two types of solvers (GS and LSk) are
tightly coupled with each other, the decoupling parameters are the
performance functions of substructures fk (k ¼1,…,K), which are
uploaded from LSk to GS, while the coordinating parameters are
specified as the allocated growth resources Mk (k¼ 1,…,K), which
are fed back from GS to LSk. Both GS and LSk (k ¼1,…,K) can be
solved by Eq. (13); the decoupling and coordinating steps are
conducted iteratively until the convergence is achieved.
3.3. Improved evolutionary algorithm
After characterizing the optimality of growth behavior, an
improved evolutionary algorithm is developed to implement the
stiffener growth and atrophy with respect to the operational loads
applied. The growth strategy consists of three basic steps, namely
initialization, competition and reconfiguration, which is illustrated
in Fig. 5.
Growth strategy step 1: initialization
For a design object, its substructures which are needed to be
reinforced are selected as growth regions. Sprouting points for
stiffeners to branch are specified within the growth region
according to the loading and constraint conditions. Stiffeners
extending from each sprouting point are activated and included
into the candidate stiffener group while other stiffeners remain
dormant. The placing direction of each candidate stiffener
represents one of the possible sprouting directions. In addition,
basic parameters such as hb (threshold for branch), hd (thresh-
old for atrophy), Wg (upper limit for weight increment at one
38 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44

growth step), Wp (initial weight of the base plate), and Wr (final
weight increment) are also defined as boundary conditions to
control the following growth process.
Growth strategy step 2: competition
According to the competition mind, an iterative calculation is
launched to allocate the growth resources (material/weight)
among candidate stiffeners. The finite element analysis is
conducted and the generalized weight of the substructures
ð  M k ð∂C=∂M k ÞÞ and the active stiffeners ð  W ik ð∂f k =∂W ik ÞÞ are
calculated respectively by Eq. (10). Then, the growth rate of
active stiffeners is controlled by the global coordinative model,
in which both the global solver GS and local solvers LSk (k¼ 1,…,
K) are driven by Eq. (13). During the simulation process, the
stiffener's height will increase gradually. As a result, the more
materials the stiffeners obtained, the more opportunities they
will get to win in the growth competition.
Growth strategy step 3: reconfiguration
The stiffener layout is constantly reconfigured by branching
and degenerating rules like those for leaf venation. If the height
of a candidate stiffener is larger than a given threshold value hb,
we consider it as a winner in the growth competition and the
placing direction of this stiffener is regarded as the sprouting
direction. As a result, the winner stiffener's two endpoints are
specified as new sprouting points, and all stiffeners around the
new sprouting points are activated to participate the growth
competition in the next iterative step. Otherwise, if the height
of a candidate stiffener is smaller than a given threshold value
hd, we consider it as a loser in the growth competition which
will be degenerated. As a result, all stiffeners connecting with
the loser stiffener's two endpoints are removed from the
candidate stiffener group. The competition and reconfiguration
steps are iterated until the weight increment reaches the given
upper limit Wr. Finally, an optimal stiffener layout could be
obtained.
4. Bionic design of grinding machine column
4.1. Design process
To examine the effectiveness of the suggested method, a high
precision grinding machine tool is selected as a design objective,
as shown in Fig. 6. The entire structure is mainly composed of five
modules, namely front base, rear base, column, working table and
carriage box. When evaluating the grinding accuracy, the movable

Fig. 6. Major components of the grinding machine tool.

Fig. 7. Boundary conditions of the column structure. (a) Transformation of the component's gravity. (b) Transformation of the grinding forces.
 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
column is a critical component whose stiffness can directly affect
the overall performance due to its frequent acceleration and
deceleration.
The stiffener plates inside the original column are uniformly
distributed, which may not be the best way for load transmission.
As mentioned above, leaf venation knows how to distribute their
materials for best efficiency, so it would be possible to create an
optimal and previous unknown proposal for stiffener design with
the proposed evolutionary algorithm.
The design objective is represented as follows:
FðWÞ ¼ ω1 f 1 ðW Þ þ ω2 f 2 ðW Þ
where F(W) is the weighted objective function; f1(W) denotes the
strain energy of the column structure, which can be used to
describe the static stiffness; f2(W) denotes the reciprocal of the
first natural frequency, which can be utilized to represent the
dynamic stiffness; ω1 and ω2 denote the corresponding weight
factors, and are selected as 0.6 and 0.4 in this design case,
respectively.
There are two kinds of boundaries for column structure: one at
the bottom interface (connected to the rear base by linear guide
sliders) and the other at the front interface (coupled with the
carriage box through linear guide sliders). For the bottom inter-
face, the nodes located at the positions of guide sliders are
assumed to be fully constrained. The inertia forces of the moving
carriage box and the operational forces applied on the grinding
wheel are transformed into equivalent forces acting on the front
interface, shown as block areas with numbers in Fig. 7.
By using the method of vector analysis [29], the equivalent
forces are calculated and their values are listed in Table 1. Fig. 8
Table 1
The values of the equivalent forces acting on the four guide sliders.
Force (N) Area_1 Area_2 Area_3
Nx 0 0 0 0
Ny 2114.85 2083.95  2058.35
Nz  2058.04  2078.50  2058.04
Fig. 8. Flowchart of the stiffener layout growth simulation.
39
shows the flowchart of the stiffener layout growth simulation,
which includes seven basic steps.
Step_1: the design object (column structure) is simplified to a
hollow box.
Step_2: the hollow box structure is discretized into shell
elements.
Step_3: the beam elements are formed by connecting the
adjacent nodes of the shell elements.
Step_4: the boundary and loading conditions are applied, and
the finite element analyses (static deformation and vibration
ð14Þ mode) are carried out.
Step_5: the front panel (substructure_1) and the side panel
(substructure_2) are defined as growth regions; sprouting
points are specified according to the loading and constraint
conditions. In this design case, the sprouting points (yellow
circle marker) are initially selected at the load action points or
edges of the panel structure so as to make it possible for
stiffeners to grow along the gradient directions from load areas
to constrained areas. As mentioned above, the sprouting points
would be evolved (add new points or delete existing points)
during the iterative calculation.
Step_6: the proposed evolutionary algorithm is adopted to
coordinate the stiffener growth between the front panel and
the side panel. The simulation steps repeat iteratively until the
total weight increment reaches the giving upper limit.
Step_7: manufacturability modifications are introduced into the
new design proposal in order to reduce weight and for easy
casting.
The material properties of the column structure is assumed to
be linear elastic, its Young's modulus E is set as 112 GPa, Poisson's
ratio μ is fixed at 0.25, and the density ρ is defined as 7200 kg/m3.
The parameters controlling the growth process are defined as
Area_4 Wr ¼ 0.25Wp, Wg ¼0.05Wp, and hb ¼104hd. The stiffener layouts
with the weight fraction varied from 0.15Wp to 0.25Wp are
constructed, as shown in Fig. 9.
 2089.25
It can be found that the strain energy of the final stiffened
 2078.50
column decreases nearly 37.13% with the first eigenvalue increases
40 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44

Fig. 9. Growth process of the stiffeners inside column structure. (a) Changing history of the strain energy. (b) Changing history of the first natural frequency.

Fig. 10. Comparison of design results with and without the coordinative algorithm.

about 9.88% compared to the original hollow box structure, so the It can be found that the total length of the stiffeners designed
proposed method is a valid topology optimizer. Design results with with the coordinative algorithm decreases nearly 1.14% compared
and without the coordinative algorithm have also been compared, to that of the stiffeners designed without the coordinative algo-
as shown in Fig. 10. rithm. At the same time, the strain energy of the column decreases
 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44 41

Fig. 11. Comparison between the proposed method and the homogenization method.

Fig. 12. Deformation patterns of the original and bionic columns.

nearly 3.07% with first eigenvalue increases about 1.36% compared Table 2
to the one designed without the coordinative algorithm, so the Comparison of the mechanical properties.
proposed method could produce better solutions in respect of both
Column Weight (kg) Max deformation (μm) First frequency (Hz)
the component weight and the structural performances. To further
examine the design commonality, a comparison between the Original design 1189.3 10.10 125.19
proposed method and the well-known homogenization method Bionic design 1174.5 8.56 133.97
is conducted, the simulation results are shown in Fig. 11. Variation (%)  1.24  15.25 þ7.01
The homogenization method is a commonly used approach in
topology optimization; it was first put forward by Bendsoe and
Kikuchi [30] in which the solution is produced by optimizing the
material density changing from 0 to 1 in each microstructure of of the external plate. However, some advantages of the proposed
the base plate. In the pseudo-density plot produced by the method can be found with respect to the homogenization method.
homogenization method, the dark red regions characterize the
areas reinforced by stiffener plates. The stiffening tendencies (1) A distinct arrangement of the stiffener layout rather than a
resulting from these two methods are similar, that is the column vague density distribution is obtained by the proposed
structure is required to reinforce the corners and the central part method, therefore, it does not require any post-processing
42 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
treatment to distinguish the locations and dimensions of the
real stiffener plates, and is easier to implement for practical
engineering design.
(2) The computational intensity is greatly reduced by the pro-
posed method, for instance, the front panel analyzed by the
homogenization method is discretized into 40  70 finite
elements, while only 4  7 finite elements are used by the
proposed method. The reason is that the homogenization
method is manipulated based on the introduction of compo-
sites with many perforated microstructures, in order to guar-
antee the resolution of topology solution, a large number of
elements are required to represent the design domain (mor-
phology or shape).
4.2. Numerical and experimental validation
Finite element simulation is carried out to evaluate both the
original and bionic columns. The deformation patterns and the
mechanical properties of the two columns are shown in Fig. 12 and
Table 2, respectively.
The maximum deformation of bionic column is reduced by
15.25% with 1.24% weight reduction, which means higher specific
stiffness is obtained. The first order natural frequency is also
improved by 7.01%, leading to a better anti-vibration performance
for column structure. To validate the correctness of numerical
simulation, static and dynamic experiments are conducted. For
the considerations of experiment cost and operation convenience,
scale-down columns with shrink ratio λl ¼ 1:5 are produced using
NC machining technology. The material of the scaled columns is
Table 3
Material properties.
Material Young's modulus E (MPa) Poisson's ratio (μ) Density ρ (kg/m3)
HT200 112 0.27 7200
PMMA 32 0.34 1180
Fig. 13. Scale-down models used in the experiments. (a) Original column. (b) Bionic column.
PMMA plastic with properties shown in Table 3. The model's
dimension is 200.0 mm (length)  121.0 mm (width)  305.5 mm
(height), as shown in Fig. 13.
All loads are calculated according to similarity theory, so that
the experiments on the scaled columns could lead to high
authenticity and reliability. The experimental setup for deforma-
tion and modal tests is shown in Fig. 14. The scaled models are
fixed on the table of a milling machine. The digital dial-indicator
with micrometer resolution is used to capture the displacement
nearby the front top points (a, b), where the maximum deforma-
tion occurs. The hammering method is employed for the modal
test. Elastic hammer of PCB-086D50 type is used for modal
excitation, acceleration sensor of PCB-356A16 type is used for
signal collection. Dynamic characteristics are analyzed through
LMS Test.lab System.
After five times of repeated tests, the experimental data are
averaged and listed in Table 4. It shows that the maximum
deformation of the bionic column is reduced by 23.60% with
1.31% weight reduction. The first order natural frequency is also
improved by 18.55%. The improvements of static and dynamic
stiffness are more significant than the results of FEM simulation.
The possible reasons might be that the column material in the
tests is PMMA plastic rather than cast iron. In any case, however,
the bionic design obtains higher specific stiffness and natural
frequency, indicating that the new stiffener layout is more efficient
than the original one, which is encouraging for further study.
5. Conclusions
Together with the inspiration of branching patterns in nature
and the engineering knowledge about machine tool structures, a
new and practical topology optimizer is developed based on a so-
called venation growth algorithm, which learns from morphogen-
esis mechanics and labels as hierarchy and reticulum. Based on the
results presented in this paper, the following conclusions can be
drawn.
 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44 43

Fig. 14. Experimental setup for deformation and modal tests. (a) Original column. (b) Bionic column.

Table 4 and experimental tests exemplify the stiffness enhancement,

Comparison of the experimental results.
and therefore the proposed growth-based method is a good
Column Weight (kg) Max deformation (μm) First frequency (Hz) choice for the design of eco-efficient machine tools.

Original design 1.53 16.1 227.34

Bionic design 1.51 12.3 269.51
Variation (%)  1.31  23.60 þ18.55

Acknowledgments

(1) Leaf venation is one type of compliance effective structure
with branching pattern. Numerical investigations find that it
has geometrical and mechanical characteristics that are of
interest for an innovative transfer into stiffener layout design
of machine tool structures.
(2) A well-founded mathematical explanation for the optimality
of plant morphogenesis is derived, which leads to a simple
criterion for formulating the growth competition among
candidate stiffeners. As an initial attempt to extend the
capability of existing growth method towards complex design
applications, this paper suggests a possible direction in model
development. A global coordinative model is introduced to
alleviate the programming and computing efforts for imple-
menting the generation of stiffener layouts inside complex
engineering structures. The unique features of this model
include its capability of coordinating the growth process
among different substructures without mathematical
sophistication.
(3) The re-design of an actual grinding machine column is con-
ducted by the proposed method. Both the numerical analyses
The work reported in this paper is supported by the National
Natural Science Foundation of China [51135004], and the Hi-Tech
Research and Development Program of China [2012AA040701].
References
[1] J.J. Zulaika, F.J. Campa, L.N. LopezdeLacalle, An integrated process-machine
approach for designing productive and lightweight milling machines, Int. J.
Mach. Tools Manuf. 51 (7–8) (2011) 591–604.
[2] L. Kroll, P. Blau, M. Wabner, U. Frieβ, J. Eulitz, M. Klarner, Lightweight
components for energy-efficient machine tools, CIRP J. Manuf. Sci. Technol. 4
(2) (2011) 148–160.
[3] A.M. Gontarz, F. Hanni, L.B. Weiss, K. Wegener, Machine tool optimization
strategies for ecologic and economic efficiency, Proc. Inst. Mech. Eng. Part B: J.
Eng. Manuf. 227 (1) (2012) 54–61.
[4] C. Brecher, P. Utsch, R. Klar, C. Wenzel, Compact design for high precision
machine tools, Int. J. Mach. Tools Manuf. 50 (4) (2010) 328–334.
[5] F. Aggogeri, A. Merlo, M. Mazzola, Multifunctional structure solutions for Ultra
High Precision (UHP) machine tools, Int. J. Mach. Tools Manuf. 50 (4) (2010)
366–373.
[6] B.C. Wu, G.S. Young, T.Y. Huang, Application of a two-level optimization
process to conceptual structural design of a machine tool, Int. J. Mach. Tools
Manuf. 40 (6) (2000) 783–794.
44 B. Li et al. / International Journal of Machine Tools & Manufacture 84 (2014) 33–44
[7] K. Jarmai, J. Farkas, Optimum cost design of welded box beams with long-
itudinal stiffeners using advanced backtrack method, Struct. Multidiscip.
Optim. 21 (1) (2001) 52–59.
[8] P.E.J. Higgins, P. Wegner, A. Viisoreanu, G. Sanford, Design and testing of the
minotaur advanced grid-stiffened fairing, Compos. Struct. 66 (1–4) (2004)
339–349.
[9] B. Gosowski, Non-uniform torsion of stiffened open thin-walled members of
steel structures, J. Constr. Steel Res. 63 (6) (2007) 849–865.
[10] M.H. Dickinson, Bionics: biological insight into mechanical design, Proc. Natl.
Acad. Sci. U.S.A. 96 (25) (1999) 14208–14209.
[11] J.F.V. Vincent, Stealing Ideas from NatureSpringer-Verlag, Vienna, 2001.
[12] F. Li, W.T. Liu, X. Fu, G. Bonsignori, U. Scarfogliero, C. Stefanini, P. Dario,
Jumping like an insect: design and dynamic optimization of a jumping
minirobot based on bio-mimetic inspiration, Mechatronics 22 (2) (2012)
167–176.
[13] M.J. Benz, A.E. Kovalev, S.N. Gorb, Anisotropic frictional properties in snakes
8339 (2012)Proc. SPIE 8339 (2012) 83390X-1–83390X-6.
[14] E. Díaz-Tena, A. Rodríguez-Ezquerro, L.N. López de Lacalle Marcaide,
L. Gurtubay Bustinduy, A. Elías Sáenz, A sustainable process for material
removal on pure copperby use of extremophile bacteria, J. Clean. Prod. (2014),
http://dx.doi.org/10.1016/j.jclepro.2014.01.061, in press.
[15] C. Mattheck, S. Burkhardt, A new method of structural shape optimization
based on biological growth, Int. J. Fatigue 12 (3) (1990) 185–190.
[16] C. Mattheck, Teacher tree: the evolution of notch shape optimization from
complex to simple, Eng. Fract. Mech. 73 (12) (2006) 1732–1742.
[17] M. Milwich, T. Speck, O. Speck, T. Stegmaier, H. Planck, Biomimetics and
technical textiles: solving engineering problems with the help of nature's
wisdom, Am. J. Bot. 93 (10) (2006) 1455–1465.
[18] X.H. Ding, K. Yamazaki, Constructal design of cooling channel in heat transfer
system by utilizing optimality of branch system in nature, ASME J. Heat Transf.
129 (3) (2007) 245–255.
[19] S. Bohn, B. Andreotti, S. Douady, J. Munzinger, Y. Couder, Constitutive property
of the local organization of leaf venation networks, Phys. Rev. E 65 (6) (2002)
1–12.
[20] L. Sack, C. Scoffoni, A.D. McKown, K. Frole, M. Rawls, J.C. Havran, H. Tran,
T. Tran, Developmentally based scaling of leaf venation architecture explains
global ecological patterns, Nat. Commun. 3 (2012) 1–10.
[21] B.T. Li, J. Hong, S.N. Yan, Z.F. Liu, Multidiscipline topology optimization of
stiffened plate/shell structures inspired by growth mechanisms of leaf veins
in nature, Math. Probl. Eng. 2013 (2013) 653895 (11 pp.), 10.1155/2013/
653895.
[22] W.Y. Liu, W.F. Hou, Y.X. Ou, Relationship between medial axis pattern of plant
leaf and mechanics self-adaptability (I): experimental investigation and
numerical simulation, J. South China Univ. Technol. 35 (3) (2007) 42–46.
[23] W.Y. Liu, J.X. Gong, W.F. Hou, Relationship between medial axis pattern of
plant leaf and mechanics self-adaptability (II): vein structures with different
vector angles and topological pattern of plant leaf, J. South China Univ.
Technol. 37 (8) (2009) 12–16.
[24] Y.X. Li, K. Xue, Mechanics in leaf venation morphogenesis and their biomi-
metic inspiration to construct a 2-dimensional reinforcement layout model,
J. Biomim., Biomater. Tissue Eng. 10 (2011) 81–93.
[25] L. Zhao, J.F. Ma, T. Wang, D.H. Xing, Lightweight design of mechanical
structures based on structural bionic methodology, J. Bionic Eng. 7 (2010)
224–231.
[26] S.J. Wang, L.Q. Ren, Y. Liu, Z.W. Han, Y. Yang, Mechanical characteristics of
typical plant leaves, J. Bionic Eng. 7 (3) (2010) 294–300.
[27] S. Chen, Some Modern Design Methods of Precise and Complex Structures-
Press of Beijing University of Aeronautics and Astronautics, Beijing, 1992.
[28] S. Chen, Analysis, Synthesis and Optimization of Engineering Structural
SystemsChina Science Culture Publishing House, Hong Kong, 2008.
[29] T.Y. Huang, J.J. Lee, On obtaining machine tool stiffness by CAE techniques,
Int. J. Mach. Tools Manuf. 41 (8) (2001) 1149–1163.
[30] M.P. Bendsoe, N. Kikuchi, Generating optimum topologies in structural design
using homogenization method, Comput. Methods Appl. Mech. Eng. 71 (2)
(1988) 197–224.