Professional Documents
Culture Documents
Multi-Objective Optimization of An On-Road Bicycle Frame by Uniform Design and Compromise Programming
Multi-Objective Optimization of An On-Road Bicycle Frame by Uniform Design and Compromise Programming
Abstract
With the rapid changes in manufacturing technology of bicycle, the safety and performance of a bicycle are important
and remarkable research subjects. In this study, an innovative and integrated optimization procedure for multi-objective
optimization of an on-road bicycle frame is presented. The multiple objectives are to reduce the bicycle frame’s perma-
nent deformations and to decrease the bicycle frame’s mass. First, uniform design of experiments is applied to create a
set of sampling points in the design space of control factors. Second, three-dimensional solid models of bicycle frames
are constructed and permanent deformations of bicycle frames under dropping-mass and dropping-frame impact test
simulations are measured by ANSYS and ANSYS/LS-DYNA. Third, Kriging interpolation is used to transform the dis-
crete relations between input control factors and output measures to continuous surrogate models. Fourth, compro-
mise programming and mixture uniform design of experiments are used to integrate the multiple-objective functions
into one compromise objective function. Finally, generalized reduced gradient algorithm is employed to solve the optimi-
zation problem. After executing the innovative optimization procedure, an optimized on-road bicycle frame is obtained.
Comparing with the original design, the frame’s permanent deformations and mass are reduced. Therefore, both consoli-
dation and lightweight of on-road bicycle frame are achieved.
Keywords
On-road bicycle frame, uniform design, Kriging interpolation, mixture uniform design, compromise programming, gener-
alized reduced gradient algorithm
Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering
and URB IV type of bicycle in which the front fork has frames in all experiments are measured by SolidWorks
an extended front projection. Nakagawa et al.2 studied system. Third, Kriging interpolation method is used to
the stabilization of the bicycle using driving forces and create surrogate models of permanent deformations
designed a controller using linear-quadratic control the- and mass. Fourth, compromise programming and mix-
ory. Schwab et al.3 introduced the dynamics of the ture uniform design of experiments are applied to inte-
bicycle model which consists of four rigid bodies. The grate multiple-objective functions into one compromise
linearized equations of motion of the 3 degrees of free- objective function. Fifth, generalized reduced gradient
dom bicycle model, considering the roll angle of the (GRG) algorithm is employed to optimize the dimen-
rear frame, the steering angle, and the rotation angle of sions of bicycle frame. From the results shown in this
rear wheel with respect to the rear frame, are derived, article, the innovative and integrated optimization pro-
and the dynamic stability analysis of the bicycle model cedure is very useful for simultaneously reducing the
is also investigated. permanent deformations and the mass of on-road
Some topics focused on the finite element analysis bicycle frame.
and the simulation for a bicycle frame model had been
presented in the literatures.4–6 The deformations and
stresses of a bicycle model have also been studied by Dynamic finite element analysis
the more complete bicycle frame model.7–11 Using two Components of bicycle frame model and two
monologue frame designs and tubular metal frames,
dropping impact tests
Larry et al.7 investigated and compared the stiffness of
in-plane and out-of-plane frames by both the experi- The on-road bicycle frame system, including the head
ments and finite element analysis. Alex and Thomas8 tube, top tube, down tube, seat tube, seat stays, chain
introduced bicycle stress level as a tool for providing stays, and front fork, is given as shown in Figure 1.
the full range of criteria needed to determine the bicycle Some basic dimensions of the bicycle frame model are
compatibility of roadways. According to fracture given in Figure 2. Considering the riding safety of a
mechanics, and using an estimate of the bending stress bicycle, the qualified bicycle frames have to pass the
from a 1-m drop, Shelton et al.9 presented an analysis drop-mass and drop-frame impact tests under EN
of a mountain bike front shock failure. Redfield10 14781 bicycle testing standard.13 The drop-mass impact
developed baseline predictions for the performance of test method is shown in Figure 3. A solid-steel bar
mountain bikes during extreme maneuvers using the fitted in place of a fork is assembled into the head tube.
bond graph model. The static analysis results by A striker is put vertically above the solid-steel bar with
ANSYS Workbench software for a sport bicycle are the distance of 212 mm. This heavy striker freely falls
investigated by Wang et al.11 By means of the genetic downhill and then impacts the solid-steel bar. The per-
algorithm, Huang and Wang12 studied the optimal manent deformation of the fork along the longitudinal
designs of the two-dimensional model of the bicycle direction of the frame can be measured and obtained as
frame under the lightest weight considerations. the striker comes to rest on the solid-steel bar.
Although the optimization designs for the bicycle Figure 4 shows schematically the drop-frame impact
frame have been illustrated from Huang and Wang,12 test. A mass of 70 kg is placed and fixed on the top of
the optimal design for the three-dimensional (3D) mod- seat tube. Then, the bicycle frame system and mass are
els of the bicycle frame has never been investigated. In placed vertically above the rear axle. Next, the bicycle
addition, the drop-frame impact testing simulation frame system is rotated about the rear axle, freely drop-
about the on-road bicycle frame is never presented by ping downhill and impacts the steel anvil. Finally, the
the explicit dynamics finite element method. The opti-
mization design for testing a bicycle frame model by
uniform design and Kriging interpolation has never
been presented.
In this study, an innovative and integrated optimiza-
tion procedure is presented to reduce the weight of
bicycle frame as well as the permanent deformations of
frame. The procedure has five steps. First, a set of
experiments are created in the design space based on
the uniform design method. Second, explicit dynamic
finite element analysis packages, ANSYS and ANSYS/
LS-DYNA, are applied to analyze permanent deforma-
tions of the bicycle frames in all experiments under two
dropping impact test simulations. Masses of bicycle Figure 1. Structures of bicycle frame.
Cheng et al. 3
Figure 7. Distribution of deformation of bicycle frame along the longitudinal direction of frame for (a) drop-mass and (b) drop-
frame impact test simulations.
for the drop-frame impact test, the deformation of the deformations of the front fork increase first and then
front fork tardily increases with increasing the thickness decrease as TT-5t increases in Figure 10(b). However,
of the first section of top tube. the deformations of the front fork under the drop-mass
Figure 10 shows the effects of the length (TT-5L) impact test decrease with increasing the thickness of the
and thickness (TT-5t) of the fifth section of top tube on fifth section of top tube.
the deformation of the front fork for the drop-mass Figure 11 shows the effects of the length (DT-1L)
and drop-frame impact tests. As can be seen, the defor- and thickness (DT-1t) of the first section of down tube
mation of the front fork evaluated using the drop-mass on deformation of the front fork for the drop-mass and
impact test always exceeds those obtained from the drop-frame impact tests. In Figure 11(a) and (b), the
drop-frame impact test simulation. In Figure 10(a), the deformations of front fork under two drop impact tests
deformation of the front fork under the drop-mass decrease as the length and thickness of the first section
impact test simulation increases as TT-5L increases. of down tube increase. Furthermore, in Figure 11(b),
For the drop-frame impact test simulation, the the effect of the thickness of the first section of down
6 Advances in Mechanical Engineering
Figure 10. Effects of (a) length and (b) thickness of fifth section
of top tube on deformation of front fork for drop-mass and
drop-frame impact tests.
DY1: deformation of front fork for drop-mass impact test; DY2:
deformation of front fork for drop-frame impact test.
Figure 11. Effects of (a) length and (b) thickness of first Figure 12. Effects of (a) length and (b) thickness of fifth section
section of down tube on deformation of front fork for drop- of down tube on deformation of front fork for drop-mass and
mass and drop-frame impact tests. drop-frame impact tests.
DY1: deformation of front fork for drop-mass impact test; DY2: DY1: deformation of front fork for drop-mass impact test; DY2:
deformation of front fork for drop-frame impact test. deformation of front fork for drop-frame impact test.
Control factor Notation Lower bound (mm) Basic value (mm) Upper bound (mm)
Experiment no. 1 2 3 4 5 6 7 8 9 10 11
1 1 3 4 5 6 7 8 9 11 15 16
2 2 6 8 10 12 14 16 18 3 11 13
3 3 9 12 15 18 2 5 8 14 7 10
4 4 12 16 1 5 9 13 17 6 3 7
5 5 15 1 6 11 16 2 7 17 18 4
6 6 18 5 11 17 4 10 16 9 14 1
7 7 2 9 16 4 11 18 6 1 10 17
8 8 5 13 2 10 18 7 15 12 6 14
9 9 8 17 7 16 6 15 5 4 2 11
10 10 11 2 12 3 13 4 14 15 17 8
11 11 14 6 17 9 1 12 4 7 13 5
12 12 17 10 3 15 8 1 13 18 9 2
13 13 1 14 8 2 15 9 3 10 5 18
14 14 4 18 13 8 3 17 12 2 1 15
15 15 7 3 18 14 10 6 2 13 16 12
16 16 10 7 4 1 17 14 11 5 12 9
17 17 13 11 9 7 5 3 1 16 8 6
18 18 16 15 14 13 12 11 10 8 4 3
values of factors, the uniform experiments are con- 3D model of the on-road bicycle frame in the 16th
structed as shown in Table 3. In each experiment, the experiment can be regarded as the improved version of
geometric modeling tool SolidWorks is used to create original design. After completing the uniform experi-
the 3D solid model of the specific design of bicycle ments, the permanent deformation in drop-mass impact
frame and then the dynamic deformations of the bicycle test has an improved rate of 4.39%, the permanent
frame undergoing drop-mass and drop-frame impact deformation in drop-frame impact test has an improved
test simulations are calculated using ANSYS and rate of 4.65%, and the mass of bicycle frame has an
ANSYS/LS-DYNA software. Based on the dynamic improved rate of 1.58%. As the main objective of the
finite element procedures given in section ‘‘Explicit design is to reduce the weight, the improved rate of
dynamics finite element analysis for the dropping mass should be increased further.
impact test simulation,’’ the permanent deformations
for all experiments can be obtained as shown in
Table 4. Kriging interpolation surrogate modeling method
In drop-mass impact test simulation, the permanent A surrogate model is an engineering method employed
deformation of the original designed bicycle frame is as an outcome of interest cannot be easily and directly
6.61 mm. In drop-frame impact test simulation, the per- measured. It is created by some statistic methods, such
manent deformation of the original designed bicycle as regression method, artificial neural networks, or
frame is 12.47 mm. The mass of the original designed Kriging interpolation. Compared with neural network
bicycle frame is 1146.49 g. By examining the improved method, Kriging interpolation does not involve train-
rates in all experiments, only experiments 2, 8, and 16 ing time, but the neural network model needs training
have all positive improved rates. Moreover, the 16th time to determine weights between neurons. Moreover,
experiment has the most improved rates. Therefore, the Kriging interpolation has been applied to study the
Cheng et al. 9
Experiment TT-1L (mm) TT-1t (mm) TT-5L (mm) TT-5t (mm) DT-1L (mm) DT-1t (mm) DT-5L (mm) DT-5t (mm)
no.
Experiment no. DY1 (mm) DY2 (mm) BM (g) Improved rate Improved rate Improved rate
of DY1 (%) of DY2 (%) of BM (%)
optimization problems in engineering.21–25 Therefore, where ^y(x) is the Kriging surrogate model of the
in this study, Kriging interpolation is applied to create unknown function of interest, f (x) is the known
a continuous surrogate model from the discrete experi- approximation (usually polynomial) function, and Z(x)
mental points. Kriging surrogate models combine a is the realization of a stochastic process with mean
global model and localized departures, given as24 zero, variance s2 , and nonzero covariance. f (x) is
regarded as a polynomial regression function, provid-
^yðxÞ ¼ f ðxÞ þ ZðxÞ ð1Þ ing a ‘‘global’’ model of the design space. As f (x)
10 Advances in Mechanical Engineering
approximates globally the design space, Z(x) creates deformations and mass can be obtained. These surro-
‘‘localized’’ deviations so that the Kriging model inter- gate models are regarded as the objective functions.
polates the sampled data points. The objective function of mass of bicycle frame is
Considering a zero-order regression and Gaussian denoted by ^yBM . The permanent deformation of bicycle
correlation function in the Kriging surrogate model, the frame in drop-mass impact test simulation is denoted
surrogate model ^y(x) of the unknown response function by ^yDY 1 . The permanent deformation of bicycle frame
y(x) can be re-built from the equation as21 in drop-frame impact test simulation is denoted by
^yDY 2 . Three single-objective optimization problems can
_ _
^y(x) = b + rT (x)R1 (Y Fb) ð2Þ be represented as follows
of ^yBM in solving the single-objective problems min ^yDY 1 Table 6. Corresponding values of objective functions.
and min ^yDY 2 ; ^ymax
DY 1 is the maximum value of ^
yDY 1 in sol- Single-objective Objective functions
ving the single-objective problems min ^yBM and optimization problem
min ^yDY 2 ; ^ymax ^yDY1 (mm) ^yDY2 (mm) ^yBM (g)
DY 2 is the maximum value of ^ yDY 2 in
solving the single-objective problems min ^yBM and min ^yDY1 5.81 15.27 1125.81
min ^yDY 1 . min ^yDY2 6.98 11.07 1123.60
min ^yBM 6.70 12.98 1113.04
Integrate multiple-objective functions into a compromise objec-
tive function. Compromise programming method31–34 is
a useful and powerful tool to integrate multiple-
objective functions into a compromise objective func- Table 7. Mixture uniform design of weights for weighting
tion. According to compromise programming method, factors.
the compromise objective function F is determined by
^yDY 1 , ^ymax ymin
DY 1 , ^ yDY 2 , ^ymax
DY 1 , ^ ymin
DY 2 , ^ yBM , ^ymax
DY 2 , ^ ymin
BM , and ^BM as
Experiment no. W1 W2 W3
follows
1 0.787 0.087 0.126
2 0.631 0.285 0.084
^y 1 ^ymin ^y 2 ^ymin 3 0.523 0.065 0.412
F = W1 DY DY 1
+ W2 DY DY 2
4 0.436 0.282 0.282
^ymax
DY 1 ^
y min
DY 1 ^ymax
DY 2 ^
y min
DY 2
5 0.360 0.552 0.088
^yBM ^ymin
BM 6 0.293 0.161 0.546
+ W3 max 7 0.232 0.454 0.314
^yBM ^ymin
BM 8 0.174 0.788 0.038
^yDY 1 5:81 ^y 11:07 9 0.121 0.280 0.599
= W1 + W2 DY 2
6:98 5:81 15:27 11:07 10 0.071 0.634 0.295
^yBM 1113:04 11 0.013 0.024 0.963
+ W3 ð8Þ
1125:81 1113:04
Here, W1 , W2 , and W3 are the weights of ^yDY 1 , ^yDY 2 , mixture uniform design of experiment method.35–37
and ^yBM , respectively. In application of compromise The mixture uniform design of weights for weighting
programming method, it is difficult to determine the factors is given as shown in Table 7. Since each row in
weight for each single-objective function. In this study, Table 7 represents a set of definite weights, each row in
the weights of single-objective functions are considered Table 7 determines a definite compromise objective
to be variable weighting factors. The values of weight- function. In other words, there are 11 single-objective
ing factors are determined by the application of optimization problems in Table 7. By application of
12 Advances in Mechanical Engineering
Table 8. Corresponding values and improved rates of DY1, DY2, and BM.
Experiment DY1 (mm) DY2 (mm) BM (g) Improved Improved rate Improved rate
no. rate of DY1 (%) of DY2 (%) of BM (%)
Solution TT-1L (mm) TT-1t (mm) TT-5L (mm) TT-5t (mm) DT-1L (mm) DT-1t (mm) DT-5L (mm) DT-5t (mm)
no.
GRG algorithm to solve the 11 optimization problems, Check prediction errors between surrogate models and analysis
the corresponding values of DY1, DY2, and BM are models. By solving the optimization problem in equa-
obtained as shown in Table 8. tion (9) with GRG algorithm, the optimal solutions of
control factors are obtained as solution 1 shown in
Select an appropriate combination of weights for weighting Table 9. Since the optimal solutions of control factors
factors. Since it is impossible to test all possible combi- are determined by surrogate models rather than analy-
nations of weights for weighting factors due to the lim- sis models, the prediction errors of DY1, DY2, and BM
itation of time, the concession strategy applied here is at the optimal solutions of control factors must be
to select an appropriate combination of weights from checked. Here, the value predicted by surrogate model
Table 7. The original designed bicycle frame has a DY1 is called the predicted value and the value analyzed by
of 6.61 mm, a DY2 of 12.47 mm, and a BM of analysis model is called the analyzed value. The bicycle
1146.49 g. The improved rates of DY1, DY2, and BM frame whose control factors use the optimal solutions
are determined and given as shown in Table 8. of control factors is called the optimized bicycle frame.
Experiments 1, 3, and 11 are deleted since the improved To obtain the analyzed values of DY1, DY2, and BM,
rate of DY2 is zero. Experiments 2, 5, and 8 are also the analysis procedure is executed one more time. The
deleted since the improved rate of BM is less than 2%. 3D model of the optimized bicycle frame is created in
From experiments 4, 6, 7, 9, and 10, we select the com- SolidWorks. The mass of the optimized bicycle frame is
bination of weights in the fourth experiment since the measured using SolidWorks. The permanent deforma-
improved rate of DY1 is required to be higher than tions of the optimized bicycle frame are analyzed by
that of DY2 by the bicycle frame manufacturing com- ANSYS/LS-DYNA. At optimal solution 1, the pre-
pany. After selecting the appropriate combination of dicted and analyzed values of DY1, DY2, and BM are
weights for weight factors W1 , W2 , and W3 , the optimi- shown in Table 10. The predicted error of DY1 is high
zation problem with a compromise objective function up to 3.78%. The predicted error of DY2 is high up to
can be represented as follows 4.92%. Because the predicted errors of DY1 and DY2
are higher than 2%, the surrogate models must be
_ ^yDY 1 5:81 ^y 11:07 updated. The optimal solutions of control factors are
min F = 0:436 + 0:282 DY 2 added into the 18 experimental points shown in Table
6:98 5:81 15:27 11:07 ð9Þ
^yBM 1113:04 3; the corresponding analyzed values of DY1, DY2,
+ 0:282 and BM are added into Table 4, and the surrogate
1125:81 1113:04
Cheng et al. 13
Solution No. Measure Predicted value Analyzed value Predicted error (%)
models of DY1, DY2, and BM are updated. The Table 11. Values and improved rates of measures in various
updated surrogate models of DY1, DY2, and BM are phases.
represented by ~yDY 1 , ~yDY 2 , and ~yBM , respectively.
Phase Measure Value Improved
Similarly, by solving three single optimization problems rate (%)
min ~yDY 1 , min ~yDY 1 , and min ~yBM , the minimum values
of ~yDY 1 , ~yDY 2 , and ~yBM are obtained as 6.26, 11.21, and Original design DY1 (mm) 6.61 –
1113.7, respectively. The maximum values of ~yDY 1 , ~yDY 2 , DY2 (mm) 12.47 –
BM (g) 1146.49 –
and ~yBM are obtained as 6.37, 13.33, and 1142, respec- After uniform DY1 (mm) 6.32 4.39
tively. The optimization problem consists of an updated experiments DY2 (mm) 11.89 4.65
compromise objective function can obtained as follows BM (g) 1128.33 1.58
After multi-objective DY1 (mm) 6.23 5.75
~ = 0:436 ~yDY 1 6:26 + 0:282 ~yDY 2 11:21
optimization DY2 (mm) 12.13 2.73
min F BM (g) 1111.35 3.07
6:37 6:26 13:33 11:21
~yBM 1113:7
+ 0:282 ð10Þ
1142 1113:7
increased to 12.13 mm. The mass BM is reduced to
The optimization problem in equation (10) is also 1111.35 g. The improved rate of mass BM can be
solved by GRG algorithm. The final optimal solution increased up to 3.07%. From the point view of light-
for the control factors for this optimization problem in weight, the design of bicycle frame after multi-objective
equation (10) are as the solution 2 shown in Table 9. optimization is superior to the design of bicycle frame
To obtain the analyzed values of DY1, DY2, and BM after uniform experiments.
for this optimized bicycle frame, the analysis procedure
is executed one more time. The 3D model of the opti-
mized bicycle frame is created in SolidWorks. The mass
Conclusion
of the optimized bicycle frame is measured using The explicit dynamics finite element simulation analysis
SolidWorks. The permanent deformations of the opti- of the on-road bicycle frame model subjected to drop-
mized bicycle frame are analyzed by ANSYS/LS- mass and drop-frame impact tests is presented in this
DYNA. At optimal solution 2, the predicted and ana- study. The factor characteristic analysis is performed
lyzed values of DY1, DY2, and BM are also shown in for the bicycle frame model of basic dimensions. The
Table 10. The predicted errors of DY1, DY2, and BM analysis result shows that the permanent deformations
are 1.93%, 0.25%, and 0.32%, respectively. Since all of the on-road bicycle frame model under the drop-
the predicted errors are smaller than 2%, the optimiza- frame impact test simulation are always smaller than
tion procedure is completed. those under the drop-mass impact test simulation. In
Table 11 shows the values and the improved rates of the optimization strategy, the uniform design of experi-
permanent deformations and mass under various ments and Kriging interpolation are used to create the
phases. In the original design phase, the permanent experimental data and three surrogate models, includ-
deformation DY1 is 6.61 mm, the permanent deforma- ing the mass and permanent deformations of the on-
tion DY2 is 12.47 mm, and the mass BM is 1146.49 g. road bicycle frame model, respectively. Minimizing the
In the phase after executing uniform experiments, the mass and permanent deformation of bicycle frame is
permanent deformation DY1 is reduced to 6.32 mm, achieved using mixture uniform design of experiment,
the permanent deformation DY2 is reduced to compromise programming method, and GRG algo-
11.89 mm, and the mass BM is reduced to 1128.33 g. In rithm. After applying the innovative optimization pro-
the phase after executing multi-objective optimization, cedure presented in this article, the total improved rates
the permanent deformation DY1 is reduced further to of the permanent deformations are 5.75% and 2.73%
6.23 mm. The permanent deformation DY2 is slightly for the drop-mass and drop-frame impact test
14 Advances in Mechanical Engineering
simulations, respectively. Furthermore, the improved 11. Wang JH, Zhao JN, Zhao YS, et al. Simulation about
rate of the mass of bicycle frame model is 3.07%. sports bicycle frame based on the experiments. Appl Mech
Therefore, the lightweight design of the bicycle frame is Mater 2010; 37–38: 1142–1147.
attended. This article has shown that the integration of 12. Huang YM and Wang KJ. Optimization of bicycle
uniform design of experiments, Kriging interpolation, frames using genetic algorithms. In: Proceedings of the
ASME international design engineering technical conferences
compromise programming, mixture uniform design of
and computers and information in engineering conference, Las
experiments, and GRG algorithm is an innovative and Vegas, NV, 4–7 September 2007, vol. 6, pp.325–334.
useful optimization strategy for solving multi-objective 13. European Standard EN 14781:2005. Racing bicycles—
optimization problems. safety requirements and test methods.
14. Fang KT and Wang Y. Number-theoretic methods in sta-
Declaration of conflicting interests tistics. London: Chapman & Hall, 1994.
15. Li JF, Liao H, Normand B, et al. Uniform design method
The author(s) declared no potential conflicts of interest with for optimization of process parameters of plasma sprayed
respect to the research, authorship, and/or publication of this TiN coatings. Surf Coat Tech 2003; 176: 1–13.
article. 16. Cheng YC and Wu PH. Optimisation for suspension sys-
tem of a railway vehicle with a new non-linear creep
Funding model developed by uniform design. Int J Heavy Veh
The author(s) disclosed receipt of the following financial sup- Syst 2015; 22: 157–191.
port for the research, authorship, and/or publication of this 17. Meng J, Li QM and Liu J. The application of uniform
article: The authors gratefully acknowledge the financial sup- design on minesweeping effectiveness forecasting model
port provided to this study by the Ministry of Science and based on neural network. In: Proceedings of the 2010 inter-
Technology of Taiwan under grant numbers 103-2221-E-327- national conference on computing, control and industrial
018 and 104-2622-E-327-019. engineering, Wuhan, China, 5–6 June 2010, pp.214–217.
Washington, DC: IEEE.
References 18. Shu X, Gu C, Xiao J, et al. Centrifugal compressor
1. Jones DEH. The stability of the bicycle. Physics Today, blade optimization based on uniform design and genetic
April 1970, pp.34–40. algorithms. Front Energ Power Eng China 2008; 2: 455–456.
2. Nakagawa C, Suda Y, Nakano K, et al. Stabilization of 19. Peng L, Wang YZ, Dai GM, et al. A novel differential
a bicycle with two-wheel steering and two-wheel driving evolution with uniform design for continuous global opti-
by driving forces at low speed. J Mech Sci Technol 2009; mization. J Comput 2012; 7: 3–10.
23: 980–986. 20. Yang YZ, Jiang M, Xu J, et al. Uniform design of opti-
3. Schwab AL, Meijaard JP and Papadopoulos JM. Bench- mizing formulation of friction materials with composite
mark results on linearized equations of motion of an mineral fiber (CMF) and their friction and wear behavior.
uncontrolled bicycle. J Mech Sci Technol 2005; 19: Appl Compos Mater 2012; 19: 161–170.
292–304. 21. Lophaven SN, Nielsen HB and Søndergaard J. DACE—a
4. Hull ML and Bolourchi F. Contributions of rider MATLAB Kriging toolbox–Version 2.0. Technical report
induced loads to bicycle frame stress. American Society of Informatics and Mathematical Modelling, Technical
of Mechanical Engineers, Design Engineering Division University of Denmark, Lyngby, Denmark, August 2002.
1987; 11: 61–72. 22. Gu YT, Wang QX and Lam KY. A meshless local Kri-
5. Wong MG and Hull ML. Analysis of road induced loads ging method for large deformation analyses. Comput
in bicycle frames. J Mech Transm: T ASME 1983; 105: Method Appl M 2007; 196: 1673–1684.
138–145. 23. McLeana P, Leger P and Tinawi R. Post-processing of
6. Bolourchi F. Bicycle frame stresses by means of finite ele- finite element stress fields using dual Kriging based meth-
ment analysis. American Society of Mechanical Engineers, ods for structural analysis of concrete dams. Finite Elem
Design Engineering Division 1986; 1: 25–28. Anal Des 2006; 42: 532–546.
7. Larry BL, James AN and Patrick LL. Utilization of 24. Simpson TW and Mistree F. Kriging models for global
FEA in the design of composite bicycle frames. Compo- approximation in simulation-based multidisciplinary
site 1995; 26: 72–74. design optimization. AIAA J 2001; 39: 2233–2241.
8. Alex S and Thomas W. Bicycle stress level as a tool to 25. Yin J, Ng SH and Ng KM. A study on the effects of para-
evaluate urban and suburban bicycle compatibility. meter estimation on Kriging model’s prediction error in
Transp Res Rec 1994; 1438: 17–24. stochastic simulations. In: Proceedings of the 2009 winter
9. Shelton H, Sullivan JO and Gall K. Analysis of the fati- simulation conference, Austin, TX, 13–16 December 2009,
gue failure of a mountain bike front shock. Eng Fail Anal pp.674–685.
2004; 11: 375–386. 26. Hwang CL, Williams JL and Fan LT. Introduction to the
10. Redfield R. Extreme mountain biking dynamics: develop- generalized reduced gradient method. Manhattan, KS:
ment of a bond graph model. In: Proceedings of the Institute for Systems Design and Optimization, Kansas
IASTED international conference on modelling and simu- State University, 1972.
lation (MS 2003), Palm Springs, CA, 24–26 February 27. Sharma R and Glemmestad B. On generalized reduced
2003, pp.450–455. gradient method with multi-start and self-optimizing
Cheng et al. 15
control structure for gas lift allocation optimization. J Rc correlation function for Kriging
Process Contr 2013; 23: 1129–1140. interpolation
28. Kalaiselvan K and Murugan N. Optimizations of friction Rc (x, xi ) correlation value of x and xi ,
stir welding process parameters for the welding of Al- i = 1, 2, . . . , n
B4C composite plates using generalized reduced gradient U18 (1811 ) uniform table
method. Proced Eng 2012; 38: 49–55. Vf initial falling speed of bicycle frame
29. Gabriele GA and Beltracchi TJ. Resolving degeneracy in Vs initial speed of striker
the generalized reduced gradient method. J Mech x vector formed by unknown input
Transm: T ASME 1987; 109: 263–267.
variables
30. Ulas A and Cengiz F. Calculation of detonation proper-
ties of gaseous explosives using generalized reduced gra-
xi experimental points, i = 1, 2, . . . , n
dient nonlinear optimization. Propell Explos Pyrot 2011; y(x) unknown response function to be
36: 314–319. interpolated
31. Kanellopoulos A, Gerdessen JC and Claassen GDH. ^y(x) Kriging surrogate model of y(x)
Compromise programming: non-interactive calibration of ^yBM Kriging surrogate model of mass of
utility-based metrics. Eur J Oper Res 2015; 244: 519–524. bicycle frame
32. Shiau JT and Wu FC. Compromise programming metho- ~yBM updated Kriging surrogate model of mass
dology for determining instream flow under multiobjec- of bicycle frame
tive water allocation criteria. J Am Water Resour As ^yDY 1 Kriging surrogate model of permanent
2006; 1179–1191. deformation of bicycle frame in drop-
33. Razmil B, Aryanezhad MB, Tavakkoli-Moghaddam R,
mass impact test simulation
et al. Using compromise programming to solve a new
~yDY 1 updated Kriging surrogate model of
multi-objective model for industrial clusters. JBASR
2012; 2: 8085–8090. permanent deformation of bicycle frame
34. Carpinelli G, Caramia P, Mottola F, et al. Exponential in drop-mass impact test simulation
weighted method and a compromise programming ^yDY 2 Kriging surrogate model of permanent
method for multi-objective operation of plug-in vehicle deformation of bicycle frame in drop-
aggregators in microgrids. Int J Elec Power 2014; 56: frame impact test simulation
374–384. ~yDY 2 updated Kriging surrogate model of
35. Ning JH, Fang KT and Zhou YD. Uniform design for permanent deformation of bicycle frame
experiments with mixtures. Commun Stat: Theor M 2011; in drop-frame impact test simulation
40: 1734–1742. ^ymax ymin
BM , ^ BM maximum and minimum values of ^yBM
36. Ning JH, Zhou YD and Fang KT. Discrepancy for uni- max min
~yBM , ~yBM maximum and minimum values of ~yBM
form design of experiments with mixtures. J Stat Plan
^ymax
DY 1 , maximum and minimum values of ^yDY 1
Infer 2011; 141: 1487–1496.
37. Prescott P. Nearly uniform designs for mixture experi- ^ymin
DY 1
ments. Commun Stat: Theor M 2008; 37: 2095–2115. ~ymax
DY 1 , ~ymin
DY 1 maximum and minimum values of ~yDY 1
^yDY 2 , ^ymin
max
DY 2 maximum and minimum values of ^yDY 2
~ymax
DY 2 , ~ymin
DY 2 maximum and minimum values of ~yDY 2
Appendix 1 Y known response vector for Kriging
Notation interpolation
Z(x) realization of a stochastic process with
BM mass of bicycle frame mean zero, variance s2 , and nonzero
DY1 permanent deformation of bicycle frame covariance.
in drop-mass impact test simulation _
DY2 permanent deformation of bicycle frame b generalized least squares estimate
in drop-frame impact test simulation um unknown coefficient of correlation
f (x) known regression function function for Kriging interpolation,
F known column vector of length n that is m = 1, 2, . . . , p
_
filled with ones mi mean of values in all experiments for
F compromise objective function factor i, i = TT 1L, TT 1t,
~
F updated compromise objective function TT 5L, TT 5t, DT 1L,
h drop height of striker DT 1t, DT 5L, DT 5t
n number of experimental points si standard deviation of values in all
p number of input variables experiments for factor i,
r correlation vector for Kriging i = TT 1L, TT 1t,
interpolation TT 5L, TT 5t, DT
R correlation matrix for Kriging 1L, DT 1t, DT 5L, DT 5t
interpolation