Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

International Journal of Pavement Engineering

ISSN: 1029-8436 (Print) 1477-268X (Online) Journal homepage: https://www.tandfonline.com/loi/gpav20

Simplified and relatively precise back-calculation


method for the pavement excitation of the
monorail

Yongzhi Jiang, Pingbo Wu, Jing Zeng, Sheng Qu, Xing Wang & Shuai Wang

To cite this article: Yongzhi Jiang, Pingbo Wu, Jing Zeng, Sheng Qu, Xing Wang & Shuai Wang
(2019): Simplified and relatively precise back-calculation method for the pavement excitation of the
monorail, International Journal of Pavement Engineering, DOI: 10.1080/10298436.2019.1623401

To link to this article: https://doi.org/10.1080/10298436.2019.1623401

Published online: 17 Jun 2019.

Submit your article to this journal

Article views: 25

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=gpav20
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING
https://doi.org/10.1080/10298436.2019.1623401

Simplified and relatively precise back-calculation method for the pavement excitation
of the monorail
a
Yongzhi Jiang , Pingbo Wua, Jing Zenga, Sheng Qua, Xing Wangb and Shuai Wanga
a
State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu, People’s Republic of China; bInstitute of Chemistry, National
Institute of Measurement and Testing Technology, Chengdu, People’s Republic of China

ABSTRACT ARTICLE HISTORY


In order to establish a more accurate model to simulate actual working conditions, a back-calculation Received 28 November 2018
method for road excitation of monorail is introduced and dynamic transcendental equations are set up Accepted 20 May 2019
in this paper. By neglecting the subtle influence of the pitching angle on the vertical acceleration of
KEYWORDS
the bogies, this dynamic model (i.e. the dynamic transcendental equations) is greatly simplified. Back-calculation method;
Besides, the calculation result shall be closer to the measured data. Comparison between the back- transcendental equations;
calculation results of the two vehicles in a monorail proves that this method can be applied to inverse Fourier transform; UM
experiments accurately. Through validating the model by comparing the experiment data and the simulation model;
simulated results, it can be noticed that the simulated result of the acceleration under the action of the experimental data; band pass
back-calculated road irregularities is consistent with the experimental data, which indicates the back- filtering
calculation process is precise enough. Band pass filtering is applied to the measured acceleration data
to get the signals in the frequency range of 0.5–10 Hz as per the standard, which proves that the
model is valid only in low frequencies but good enough for researches of vehicle stability.

1. Introduction To analyse the interaction of the vehicle–bridge coupling


system, the vehicle–bridge coupling dynamics is carried out.
1.1. Background
The bridge model was built by grid element (Nassif 2003),
In order to save time and cost, a model is proposed to simulate beam element (Choi and Song 2000, Moghimi and Ronagh
the experimental process. However, no matter how complex 2008, Wang et al. 2016, Zhang et al. 2016), shell element
the model is, the simulation results cannot be completely con- and Bernoulli–Euler beam element (Law and Zhu 2005,
sistent with the experiment data. Besides, the more complicated Lombaert and Conte 2012). Bridge FE models of straddle
the simulation model is, the more parameters are needed to monorail (Lee et al. 2005, 2006, Kim and Kawatani 2006,
measure. A small error of the measured parameters will result Kim et al. 2013, Naeimi et al. 2015, Guo et al. 2018) and bridge
in a great difference between the simulation results and the FE models of suspended monorail (Bao 2016, Cai 2019) are
experimental data. coupled with the dynamic model of the vehicle for simulation,
In order to avoid the complex modelling process of the respectively.
vehicle–bridge coupling system, many researchers have selected Because the monorail track beams hang high above the
a single component as the research object. Finite element model ground, it is extremely hard and dangerous to measure road
will be a good choice to analyse the structure of an object. Du irregularities of the monorail. Because there are no relevant
et al. (2017), instead of building a model of the whole vehicle, regulations for monorail pavement, Lee et al. (2005) and Cai
built a three-dimensional finite element model of the wheel– et al. (2019) have measured road irregularities of the straddle
rail contact system; and thus, greatly reduced the complexity and suspended monorail. However, due to lack of relevant
of the modelling process and obtained a relatively precise simu- regulations for monorail pavement, German low disturb spec-
lation result of the tire abrasion. There are some other examples trum (Zhang and Xia 2013), ISO8608 Standard road spectrum
like the FE model of I type beam and bogie frame set up in some (Wen et al. 2017), U.S. rail spectrum (Zhou 2005) and other
researches (Abdulhameed et al. 2015, Li 2016). standard track spectrums that bring vehicle vibration closer
In order to further analyse the interaction of each com- to experimental values were chosen to simulate the monorail
ponent in the vehicle system, regardless of the structure of road irregularities. In order to better simulate road irregulari-
the bridge, the method of multi-rigid body dynamics is intro- ties, Zhou (2005) chose various track power spectrums to
duced. 3D multi-rigid body model is set up by some dynamic serve as the road excitation. By comparing the experimental
software like RecurDyn (Kim et al. 2008), Simulink (Jiang and simulated acceleration data, he selected the best track
et al. 2015, 2017), Simpack (Pu et al. 2015, Xiao et al. 2015), power spectrum. However, if road irregularities (that is, the
DADS (Tsunashima 2003) and mathematic calculation pro- main cause of the vibration of the vehicle) cannot be accurately
gramme (Muller 1978, Wang and Zhu 2018). simulated, the simulation results of the dynamic model can

CONTACT Yongzhi Jiang 644263958@qq.com; Pingbo Wu wupingbo@163.com


© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 Y. JIANG ET AL.

hardly be consistent with the experimental data. In order to 3D model in the software. This part is to make sure these
prove that the abnormal vibration is caused by the wheel eccen- simplifications will not cause big errors.
tricity, Lv et al. (2017) neglected road irregularities in his simu- 3. Experimental verification of the back-calculation method.
lation model. He verified his speculation by comparing the By comparing road irregularities calculated from the data
dominant frequency of the experimental acceleration and of two vehicles, it is proved that the method can be used
simulated acceleration. However, this calculation process is in actual measurement.
not accurate enough. So, he carried out a second experiment. 4. Comparative analysis of the experimental and simulation
Lack of road irregularity data made the study very difficult. results. It is to prove that the simulation model can accu-
With the genetic algorithm, Jiang (2018) scaled the amplitude rately simulate actual working conditions. By inputting
and wavelength of road irregularities to make the experimental the back-calculation road irregularities calculated by the
data consistent with the simulation data. However, the simu- measured acceleration data into the software, the simulated
lation accuracy is low. acceleration signal is obtained. If the measured acceleration
There are many methods to simulate track irregularities. is consistent with the simulated signal, it proves the model
Based on a spatial and statistical decomposition, taking the can simulate actual working condition accurately.
spatial and statistical variability and dependency of the track
geometry into account, Perrin et al. (2013) developed a sto-
chastic model. Xu et al. (2017) provided a probabilistic 2. Inverse Fourier transform
model for random track irregularities. In his research, random
field modelling and highly efficient sample selection are devel- Considering the effect of the non-linear factors such as the
oped. By using an MCM-KLE unification method, Xu et al. wheel–rail contact factors, the accuracy of the numerical simu-
(2018) developed a model to evaluate the dynamic perform- lation for track irregularities, an important part of the dynamic
ance of train-track-bridge systems in random temporal– research, directly affects the accuracy of the whole study. The
spatial fields. Other methods like Improved Grey Model and most commonly used numerical simulation methods of road
PSO-SVM (Ma et al. 2018), Markov-based model (Bai et al. irregularities are the white noise filtering method, secondary
2015), gray interval prediction model (Wang and Wang filtering method, trigonometric series method and the inverse
2015), Data Mining and Linear Model (Jia et al. 2013) are Fourier transform method (Chen and Zhai 1999).
also adopted to simulate track irregularities for the railway When the inverse Fourier transform method (Chen and
vehicle. However, there are still not many researches of the Zhai 1999) based on the principle of fast numerical algorithm
monorail. for the power spectrum is used in the calculation process, com-
Similar to the way that was adopted by Wei et al. (2014) to pared with the triangular series method, the simulation speed is
calculate the wheel–rail forces through the acceleration signals, greatly improved. It cannot be widely applied because certain
this paper is devoted to finding a back-calculation method for filters have to be designed in order to be able to use the second-
road irregularities. Even if the vehicle is under the same road ary filtering method and proper filters have to be designed for
excitation, due to differences between the real vehicle structure, different power spectral density functions. Trigonometric series
multi-rigid body model and mathematical equations, accelera- method and white noise filtering method regard the road irre-
tion results of the simulation and experiment will vary from gularities as a stationary Gaussian stochastic process, which is
each other. If there are too many acceleration parameters in obviously not completely consistent with actual road proper-
the back-calculation formula, errors of each acceleration signal ties. Lin (1982) proved that road irregularities are not always
will accumulate into the back-calculation result and make a the stationary stochastic process. In view of the weaknesses of
large error. This paper is aimed to find a simple but relatively the above methods, inverse Fourier transform has the advan-
precise back-calculation formula for the road excitation of the tages of less errors, faster operation speed and better generality.
monorail. Therefore, the inverse Fourier transform is chosen to simulate
road irregularities in this paper.
According to the Blackman–Turkey method for estimating
the power spectrum, road irregularities are set as {xs },
1.2 Objectives
s = 0, 1, · · · , (N − 1). Time range is set as T = ND. D is the
The main objectives of this paper are as follows: time interval. The time delay of the correlation function is
also the discrete value, namely t = rD. The following equation
1. Introduction of the inverse Fourier transform. The aim of is obtained:
this part is to prove that the actual power spectral density ⎧
⎪ T
of road irregularities obtained by inverse Fourier transform ⎪
⎪ 1

is basically consistent with the theoretical value, that is to ⎨ Rxx (t) = T x(t)x(t + t)dt

say, this calculation process will not result in critical errors. 0 . (1)

⎪  −1
2. Theoretical verification of the back-calculation method. To ⎪
⎪ 1 N

⎩ Rr = Rxx (t = rD) = xs xs+r
benefit calculation and facilitate the coincidence between N s=0
experiment data and theoretical result, some simplification
is made in the back-calculation model. Besides, there are In this equation, r = 0, 1, · · · , (N − 1), xs ; x (sD), Rxx and
some differences between the mathematical model and the Rr are the correlation functions of {xs }.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 3

Substitute Rxx (t) as Sxx (f ). The following equation can be obtained:


obtained: 
2p 2pfc 2
 2p kAV
 −ik r
1 N−1 V V
Sxx (k) =
1
Sxx f = k
1
= Rr e N . (2)
S(f ) =  2  2  
T T N r=0 2pf 2pf 2pfc 2
(8)
V V V
In this equation, Sxx (k) is the power spectral density function of
{xs }. The following equation can be obtained: kAV fc2 V
= .
2pf 2 (f 2 + fc2 )

2p
1N −1
1 N
−1
−ik r In this equation, V is the spatial frequency (rad/m); v, the
Sxx (k) = xs xs+r e N
N r=0 N s=0 angular frequency (rad); f , the time frequency(Hz); V, the
⎧ ⎫⎧ ⎫ speed of the vehicle (m/s).
⎨ 1 N−1 2p 2p
 ik s ⎬⎨ 1 N−1  −ik (s + r) ⎬ After obtaining the time-domain power spectrum, it is
= xe N x e N . needed to convert the unilateral spectrum to a bilateral spec-
⎩N s=0 s ⎭⎩N r=0 s+r ⎭
trum Sx (f ).
(3) The simulated time range is Ts . The time interval is D. The
sampling point number is Nr = Ts /D and the frequency-
It is defined j = r + s and the following equation can be domain sampling interval is Df = 1/(Nr D). It is well-known
obtained: that the power spectrum estimated by the periodogram method
is periodic and even symmetric.
2p 2p

N−1
−ik 
(s + r) N−1+s −ik j From Equations (2) and (6), it can be learned that the spec-
xs+r e N = xj e N . (4) tral modulus of time series x(n) is defined as follows:
r=0 j=s

|X(k)| = |DFT[x(n)]| = Nr2 × Sk (k)
Because of the discrete Fourier transform, the time series {xs }  
have been discretized periodically. The period is N. The follow- = Nr Sk (k) = Nr Sk (f = kDf )Df ,
ing equation can be obtained:
(k = 0, 1, · · · , Nr − 1). (9)
2p 2p

N−1+s
−ik j N −1
−ik j
xj e N = xj e N . (5) Because the time series x(n) is a random series, its spectral
j=s j=0 phase must be random. Let jn be a sequence with independent
phase. The mean of each component is zero. Moreover, the
Put Equations (4) and (5) into Equation (3), and the following FFT of a real sequence is a plural sequence (the real part of
equation can be obtained: it is even symmetric and the imaginary part is odd sym-
⎧ ⎫⎧ ⎫ metric). So, jn should be plural. Besides, |jn | = 1. It’s
⎨ 1 N 2p ⎬⎨ N −1 2p ⎬
−1
ik s 1  −ik j defined as follows:
Sxx (k) = xe N xe N
⎩N s=0 s ⎭⎩N j=0 j ⎭ jn = cos Fn + i sin Fn = exp (iFn ). (10)
(6)
1 1 Fn is a series belonging to a uniform distribution at the range
= |DFT[xs ]|2 = 2 [X ∗ (k)X(k)].
N2 N of 0  2p.
Because the real part of X(k) is even symmetric about Nr /2,
In this equation, X(k) is the spectrum of {xs }. and the imaginary part is odd symmetric about Nr /2.
s = 0, 1, · · · , (N − 1), k = 0, 1, · · · , (N − 1). Obviously, it is easy to get X(k), (k = 0, 1, · · · , Nr − 1)
Compared with several PSD functions of road irregularities, according to the symmetric properties, so it is only needed to
Zhou (2005) thought the PSD function of U.S. rail irregularities calculate the value of X(k), (k = 0, 1, · · · , Nr /2). The follow-
could better simulate the monorail rail, which is also proved by ing equation can be obtained:
Jiang et al. (2018, 2019a, 2019b) experimentally.

The existing power spectrums of track irregularities are X(k) = j(k)|X(k)| = Nr j(k) Sx (f = kDf )Df (k
spatial spectrums shown as follows. The spatial spectrum
needs to be transformed into time-domain spectrum (Wang = 0, 1, · · · , Nr /2). (11)
2009). Carry out the IFFT (inverse fast Fourier transform) of X(k) and
the following equation can be obtained:
KAv V2c
S(V) = . (7)
V (V2 + V2c )
2 2pk
1 N
r −1 i n
x(n) = X(k)e Nr (n = 0, 1, · · · , Nr − 1). (12)
Because V = V v, and v = 2pf , it can be obtained as N k=0
2pf
V= . According to the theory of energy conservation, it’s The parameters of Equation (7) are as follows:
V
known S(f )df = S(V)dV, and the following equation can be K = 0.25, Vc = 0.8245
4 Y. JIANG ET AL.

Figure 1. Comparison of the power spectral density of the inverse Fourier transformed USA rail irregularities and the theoretical value: (a) Level 6, (b) Level 5, (c) Level 4.

For the USA rail level of 4, Av = 0.5376. It can be simplified as Equation (14).
For the USA rail level of 5, Av = 0.2095.
For the USA rail level of 6, Av = 0.0339. ††
mb z b = Fz DLF + Fz DLR + Fz DRF + Fz DRR + Fy GLF

The time response×(t) is obtained by the inverse Fourier + Fy GLR + Fy GRF + Fy GRR + Fy SL + Fy SR
transform of the existing power spectrum, then the power spec- − (KS ((zb + wb lS ) − (zc + wc lS )) + KS ((zb − wb lS )
tral density is obtained. As can be seen from Figure 1, the actual † † † †
power spectral density of road irregularities obtained by inverse − (zc − wc lS )) + CS (( z b + wb lS ) − ( z c + wc lS ))
Fourier transform is basically consistent with the theoretical † † † †
+ CS (( z b − wb lS ) − ( z c − wc lS ))
value, which means this calculation process will not produce
large errors. The high simulation accuracy and the satisfactory † † † †
+ CV sin a(( z b + wb lV ) − ( z c + wc lV ))
results proved that both the calculation process for the time
† † † †
response by inverse Fourier transform of the known power spec- + CV sin a(( z b − wb lV ) − ( z c − wc lV ))),
tral density and the calculation process for the power spectral
(13)
density by the known time response will not cause many errors.
 
†† 2sD
3. Back-calculation for the road irregularities mb z b = 2KD (ze (t) − zb ) + 2KD ze t + − zb
V
The calculated result of road irregularities is obtained accord-  
† † † 2sD †
ing to the acceleration signals of the vehicle. Because, as + 2CD ( z (t) − z b ) + 2CD z e t + − zb
shown in Figure 2(a), the driving wheels on both left sides e V
and right sides of the bogie are attached closely; road exci- † † † †
− 2KS (zb − zc ) − 2CS ( z b − z c ) − 2CV sin a( z b − z c )
tation for both the left and the right driving wheels are con-
sidered the same. + Fy GLF + Fy GLR + Fy GRF + Fy GRR + Fy SL + Fy SR .
The parameters of the dynamic model are shown in Table 1: (14)
The dynamic model is shown in Figure 2(b), and the vertical
dynamic equation of the front bogie is shown in Equation (13). According to FIALA tire model, the equation of the tire side
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 5

Figure 2. Structure of the vehicle: (a) Structure of the bogie, (Zhong 2013) (b) Dynamic structure of the monorail vehicle.

force can be shown as follows. The equation of tire side force (16) and (17) can be com-
bined into one formula as follows:

⎨ sy = tan u
3mFz . (15)
⎩ s′ = sign(|sy | − s′ ) − 1
Ky
Fy = − mFz (1 − h3 )sign(sy )
2
(18)
Case 1 |sy | , s′ sign(|sy | − s′ ) + 1
+ mFz sign(sy ).
⎧ 2
⎨ h = 1 − Ky |sy |
3mFz . (16)

Fy = mFz (1 − h3 )sign(sy ) It can be simplified as follows:

Case 2 |sy | ≥ s′ 
sign(|sy | − s′ ) − 1 3
Fy = m Fz h +1 (19)
Fy = mFz sign(sy ), (17) 2

where, u denotes the sideslip angle; m, the coefficient of friction;


Ky , the cornering stiffness; and Fz , the tire radical force. by inputting the parameters into the equation, it can be
6 Y. JIANG ET AL.

Table 1. Parameters and variables of the dynamic system. following equations are defined:
Parameter Description ⎧ †
mc Mass of the car body ⎪
⎪ FGy (u,Dz)= m(KG Dz+CG Dz )


mb Mass of the bogie frame ⎪
⎪ ⎡  ⎤
zc Vertical vibration displacement of the point on the floor of the car ⎪
⎪ 3mKG

⎪ sign |tan u|− −1  3
body above the bogie ⎪
⎪ ⎢ ⎥
zb Vertical vibration displacement of the bogie frame ⎪
⎪ ⎢ K Gy ⎥ 1− K Gy |tan u|

⎪ ⎣ ⎦ +1
ze Road excitation of the driving wheels ⎪
⎪ 2 m Dz+C

Dz
cc Yawing angle of the car body ⎪
⎨ 3 (K G G )
cb Yawing angle of the bogie frame .
wc ⎪ †
Rolling angle of the car body ⎪
⎪ FDy (u,Dz)= m(KD Dz+CD Dz )
wb Rolling angle of the bogie frame ⎪

fc ⎪
⎪⎡  ⎤
Pitching angle of the car body ⎪
⎪ 3mKD
fb Pitching angle of the bogie frame ⎪

⎪ sign |tan u |− −1  3
lS Half of the lateral span of the air springs ⎪
⎪ ⎢ KDy KDy |tan u| ⎥

⎪ ⎢ 1− +1⎥
lD Half of the lateral span of the driving wheels ⎪
⎪ ⎣ 2 † ⎦
lG Half of the lateral span of the tire surfaces of the guide wheels ⎪
⎩ 3m(KD Dz+CD Dz )
lV Half of the lateral span of the hinge point which aritculate the shock
absorber and the bogie
lVc Half of the lateral span of the hinge point which aritculate the shock
(21)
absorber and the car body
sD Half of the longitudinal span of the driving wheels The excitations for the guide wheels and stable wheels on the
sG Half of the longitudinal span of the guide wheels left and right of the bogie are set as zGLe zGRe zSLe ,zSRe . The
sb Half of the longitudinal span of the bogies on a vehicle equations of the radial deformation of the guide wheels and
sc Half of the length of the vehicle
hb Height between the gravity centre of the bogie and the running rail stable wheels are defined as follows:
surface ⎧
hSb Height between the gravity centre of air spring and the gravity centre ⎪
⎪ DzGLF =yb − wb hG − cb sG +zGLe (t)

⎪ Dz GRF =−yb + wb hG + cb sG +zGRe (t)
of the bogie

⎪ 
hSc Height between the gravity centre of air spring and the vehicle floor ⎪

hV Height between the gravity centre of the bogie and the hinge point ⎪
⎪ 2sG

⎨ DzGLR =yb − wb hG + cb sG +zGLe t+
which aritculate the shock absorber and the bogie V
hVc Height between the vehicle floor and the hinge point which  . (22)

⎪ 2sG
aritculate the shock absorber and the car boday ⎪
⎪ Dz =−y + w h − c s +z t+
KS Vertical stiffness of the air spring ⎪

GRR b b G b G GRe
V
KSy ⎪

Lateral stiffness of the air spring ⎪
⎪ Dz =y + w +z
CS Vertical damping of the air spring ⎪
⎩ SL b
h
b S SLe (t)
CSy Lateral damping of the air spring DzSR =−yb − wb hS +zSRe (t)
CV Damping of the lateral shock absorber
Fz DL Fz DR Radical forces of the left and right driving wheel The vertical vibration movement of the bogie is transformed
Fy GL Fy GR Side forces of the left and right guide wheel
Fy SL Fy SR Side forces of the left and right stable wheel into Equation (23).
 
†† 2sD
mb z b = 2KD (ze (t) − zb ) + 2KD ze t + − zb
transformed into Equation (20) V
 
† † † 2sD †
+ 2CD ( z (t) − z b ) + 2CD z e t + − zb
⎡ 3mFz ⎤ e V
sign(| tan u| − ) − 1 3 † † † †
⎢ Ky Ky | tan u| ⎥ − 2KS (zb − zc ) − 2CS ( z b − z c ) − 2CV sin a( z b − z c )
Fy =m F z ⎢
⎣ 1− + 1⎥

2 3mFz + FGy (fb , yb − wb hG − cb sG + zGLe (t))
+ FGy ( − yb + wb hG + cb sG + zGRe (t))

 
sign | tan u| − 3mKz Dz −1 2sG
† Ky + FGy fb , yb − wb hG + cb sG + zGLe t +
= m(Kz Dz + Cz D z ) V
2  
2sG
 3  + FGy fb , − yb + wb hG − cb sG + zGRe t +
Ky | tan u| V
× 1− † +1 .
3m(Kz Dz + Cz D z ) + FGy (fb , yb + wb hS + zSLe (t))
(20) + FGy (fb , − yb − wb hS + zSRe (t)).
(23)
Let FGy (u, Dz) and FDy (u, Dz) substitute for the equation of the Another four variables are inputted in this equation, namely,
side force of the guide wheel (stable wheel) and driving wheel, zGLe (t), zGRe (t), , zSLe (t) and zSRe (t). The variables cannot be
respectively. There are two decisive variables, namely Dz, the solved by only one equation. More equations are needed to
radial deformation of the wheel and u, the sideslip angle. The solve these variables.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 7

The lateral vibration movement of the bogie is shown in † †


† †
Equation (24). The rolling movement of the bogie is shown − hVb 2CV cos a( yc + hVc wc − yb − hVb wb )
in Equation (25). + hG KG [yb − wb hG − cb sG + zGLe (t)]
†† † † †

mb y b = FDy (cb , ze (t) − (zb − wb lD + fb sD )) + hG CG [ yb − wb hG − cb sG + z (t))]
  GLe
2sD
+ FDy cb , ze t + − (zb − wb lD − fb sD ) − hG KG [ − yb + wb hG + cb sG + zGRe (t)]
V
† † † †
+ FDy (cb , ze (t) − (zb + wb lD + fb sD )) − hG CG [ −yb + wb hG + cb sG + z (t)]
GRe
    
2sD 2sG
+ FDy cb , ze t + − (zb + wb lD − fb sD ) + hG KG yb − wb hG + cb sG + zGLe t +
V V
  
+ 2KSy (yc + hSc wc − yb − hS wb ) † † † † 2sG
+ hG CG b − wb hG + cb sG + z GLe t +
y
† † † † V
+ 2CSy ( yc + hSc wc − yb − hS wb )   
2sG
† † † † − hG KG −yb + wb hG − cb sG + zGRe t +
+ 2CV cos a( yc + hVc wc − yb − hV wb ) V
− KG [yb − wb hG − cb sG + zGLe (t)]   
† † † † 2sG
− hG CG −yb + wb hG − cb sG + z GRe t +
† † † † V
− CG [ yb − wb hG − cb sG + z (t))]
GLe
− hG KG [yb + wb hS + zSLe (t)]
+ KG [ − yb + wb hG + cb sG + zGRe (t)] † † †

− hG CG [ yb + wb hS + z (t)]
† † † SLe
+ CG [ −yb + wb hG + cb sG + z (t)]

GRe
 + hG KG [ − yb − wb hS + zSRe (t)]
2sG †
− KG yb − wb hG + cb sG + zGLe t + † †
+ hG CG [ − yb − wb hS + z (t)] − lD KD [ze (t)
V SRe
 
† † 2sG
† †
−CG [yb − wb hG + cb sG + z GLe t + − (zb − wb lD + fb sD )]
V   
   2sD
2sG − lD KD ze t + − (zb − wb lD − fb sD )
+ KG −yb + wb hG − cb sG + zGRe t + V
V
   + lD KD [ze (t) − (zb + wb lD + fb sD )]
† †
† †
+ CG −yb + wb hG − cb sG + z GRe t +
2sG   
lD
V + lD KD ze t + − (zb + wb lD − fb sD )

V
† †
− KG [yb + wb hS + zSLe (t)] − CG [ yb + wb hS + z (t)] † † † †
SLe
− lD CD [ z (t) − ( z b − wb lD + fb sD )]
e
+ KG [ − yb − wb hS + zSRe (t)]   
† lD † † †
† † † − lD CD z e t + − (z b − wb lD − fb sD )
+ CG [ − yb − wb hS + z (t)]. V
SRe
† † † †
(24) + lD CD [ z (t) − ( z b + wb lD + fb sD )]
e
††   
IbX w b = hb FDy (cb , ze (t) − (zb − wb lD + fb sD )) † lD † † †
  + lD CD z e t + − (z b + wb lD − fb sD )
lD V
+ hb FDy cb , ze t + − (zb − wb lD − fb sD )
V − { − lS KS [(zb + wb lS ) − (zc + wc lS )]
+ hb FDy (cb , ze (t) − (zb + wb lD + fb sD )) + lS KS [(zb − wb lS ) − (zc − wc lS )]
  † † † †
lD − lS CS [( z b + wb lS ) − ( z c + wc lS )]
+ hb FDy cb , ze t + − (zb + wb lD − fb sD )
V † † † †
+ lS CS [( z b − wb lS ) − ( z c − wc lS )]
− hSb 2KSy (yc + hSc wc − yb − hSb wb ) † †
† †
† † † †
− lV CV sin a[( z b + wb lV ) − ( z c + wc lV )]
− hS 2CSy ( yc + hSc wc − yb − hSb wb )
8 Y. JIANG ET AL.

† † † †
vibration are omitted. The three equations, namely, Equations
+ lV CV sin a[( z b − wb lV ) − ( z c − wc lV )]} (23)–(25) cannot be solved with five variables. Replace
− lG FGy (fb , yb − wb hG − cb sG + zGLe (t)) yb , zb , wb , fb , cb , t in the Equations (23)–(25) with
2sb
+ lG FGy ( − yb + wb hG + cb sG + zGRe (t) yb2 , zb2 , wb2 , fb2 , cb2 , t + , and the dynamic equations of
  V
2sG the rear bogie in the vehicle could be obtained.
− lG FGy fb , yb − wb hG + cb sG + zGLe t +
V In order to get the acceleration signals of the bogies and car
  body, the acceleration sensors are mounted on the specified
2sG
+ lG FGy fb , − yb + wb hG − cb sG + zGRe t + location in the vehicle, as shown in Figure 3. Sensor c_1 and
V
sensor c2_1 are mounted on the floor of the car body beyond
− lG FGy (fb , yb + wb hS + zSLe (t)) the bogies. Sensor c_2 and sensor c2_2 are mounted on the
+ lG FGy (fb , − yb − wb hS + zSRe (t)). left side of sensor c_1 and sensor c2_1 in the distance of lc12 .
Sensor b_1 and sensor b2_1 are mounted in the centres of
(25)
the front and rear bogies respectively. Sensor b_2 and sensor
In order to avoid inputting longitudinal force variables, b2_2 are mounted on the rear end beam in the bogies, sb12
dynamic equations of pitching, yawing and longitudinal metres away from sensor b_1 and sensor b2_1 respectively.

Figure 3. Mount position of the acceleration sensor: (a) In the vehicle, (b) In the bogie.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 9

Sensor b_2 and sensor b2_2 are mounted on the left side beam which could be further simplified as Equation (28).
in the bogies, lb13 metres away from sensor b_1 and sensor ††
b2_1. ay senser XX , az senser XX represent the lateral and vertical mb z = KD [ze (t) − (zb − wb lD + fb sD )]
b
acceleration signals measured by sensor XX respectively   
(Figure 3). 2sD
+ KD ze t + − (zb − wb lD − fb sD )
The acceleration in the dynamic equations could be V
obtained as follows: + KD [ze (t) − (zb + wb lD + fb sD )]
  
2sD
+ KD ze t + − (zb + wb lD − fb sD )
⎧ †† ††
V

⎪ c = ay senser c 1
⎪ y y c2 = ay senser c2 1 †


† † †
+ CD [ z (t) − ( z b − wb lD + fb sD )]


††
z c = az senser c 1
††
z c2 = az senser c2 1

⎪ e

⎪ †† az senser c 1 − az senser c 2 †† az senser c2 1 − az senser c2   

⎪ wc = w c2 =
2
† 2sD † †

⎪ + CD z e t +

− (z b − wb lD − fb sD )

⎪ lc12 lc12
⎪ ††
⎪ †† V


⎨ b = ay senser b 1
y y b2 = ay senser b2 1
† † † †
††
z b = az senser b
††
z b2 = az senser b2 1 . + CD [ z (t) − ( z b + wb lD + fb sD )]

⎪ 1 e

⎪ †† az senser b − az senser b 3 †† a − a   

⎪ wb =
1
w b2 =
z senser b2 1 z senser b2 3
† 2sD † † †

⎪ lb13 lb13 + CD z e t + − (z b + wb lD − fb sD )



⎪ †† az senser b 1 − az senser b 2
†† az senser b2 1 − az senser b2 2 V

⎪ fb = f b2 =

⎪ − {KS [(zb + wb lS ) − (zc + wc lS )]


sb12 sb12

⎪ †† a 1 − ay senser b 2 †† ay senser b2 1 − ay senser b2 2

⎩ c b = y senser b c= + KS [(zb − wb lS ) − (zc − wc lS )]
sb12 b2 sb12
† † † †
(26) + CS [( z b + wb lS ) − ( z c + wc lS )]
† † † † † †
+ CS [( z b − wb lS ) − ( z c − wc lS )] + CV sin a[( z b + wb lV )
† † † † † †
The value of ze , zGLe , zGRe , zSLe , zSRe can be obtained by sol- − ( z c + wc lV )] + CV sin a[( z b − wb lV ) − ( z c − wc lV )]},
ving the six dynamic equations of the two bogies. It’s difficult (27)
to solve these transcendental equations. At the same time, in  
order to solve the dynamic model, 10 sensors should be †† 2sD
mb z = 2KD (ze (t) − zb ) + 2KD ze t + − zb
mounted in the vehicle and 14 acceleration signals should b V
be measured. The error of one acceleration signal will  
† 2sD † (28)
cause the equation results to deviate from the true value. + 2CD z e t + − z b − 2KS (zb − zc )
As the actual car body and bogies are not the rigid bodies V
as defined by the mathematical model, too many acceleration † † † †
− 2CS ( z b − z c ) − 2CV sin a( z b − z c ).
signals will amplify the difference between the actual
measured data and the mathematical calculation results This equation could be solved because there is only one vari-
and, accordingly, make a greater difference between the cal- able in the equation. The signals of the vibration displacements
culated results and the measured data. In this case, it is and velocities can be obtained by integrating the acceleration
insufficient to use this mathematical model in order to signals. It can be seen from the equation, only two acceleration
make the measured value agree with the calculated value. †† ††
signals, namely, z b and z c (calculated by the measured value
More complex models, such as the rigid-flexible coupling az senser c 1 and az senser b 1 ) are needed in this dynamic
model, are needed to make the calculation value coincide model. That means the calculation result could be much closer
with the actual data. However, it will be more difficult to to the measured data. Besides, the calculation process is much
carry out the back-calculation model of road irregularities. simplified.
Due to the difficulty of solution and large error, this dynamic This equation could be transformed into Equation (29):
model needs to be modified.
When the vehicle runs on a straight track, the pitching † 1 ††
z (t) = − { − mb z b + 2KD (ze (t) − zb )
movement of the bogie is relatively small. It exerts little e 2CD
influence on the vertical vibration of the vehicle. Therefore,      
2sD † 2sD †
its effect on the vertical acceleration signal can be neglected. + 2KD ze t + − zb + 2CD z e t + − zb
V V
As for the monorail dynamic system, the value of the side
† † † † †
slip angles of the guide wheels and stable wheels are equal − 2KS (zb − zc ) − 2CS ( z b − z c ) − 2CV sin a( z b − z c )} + z b .
to the pitching angle of the bogie. Because the pitching
(29)
vibration of the bogie is neglected in this paper, the value of
u is 0, which means the values of the side force of guide In Figure 4, integration for the acceleration signals is carried
wheels and stable wheels are 0 in this dynamic model. out to get the signals of velocities and displacements. Mod-
So, Equation (23) can be transformed into Equation (27), ules ‘1-D Look up Table1’ and ‘1-D Look up Table2’ are to
10 Y. JIANG ET AL.

Figure 4. Model of the monorail vehicle: (a) UM simulation model, (b) Topological graph of the UM simulation model.

carry out the linear interpolation for the measured data, It’s defined in Equation (31).
†† ††
1  
z (nDt0 ) and z (nDt0 ), respectively. Acceleration at the
b c ††
time ‘t’ could be obtained through the interpolation shown f (t) = − −mb z b + 2KD (ze (t) − zb )
2CD
in Equation (30).      
if (n − 1)Dt0 , t , nDt0 lD † lD †
+ 2KD ze t + − zb + 2CD z e t + − zb
⎧ †† ††
V V

⎪ †† †† z b (nDt0 ) − z b [(n − 1)Dt0 ] † †

⎪ bz (t) = z [(n − 1)Dt ] + − 2KS (zb − zc ) − 2CS (z b − z c )


b 0
Dt0


[t − (n − 1)Dt], † † †
†† ††
− 2CV sin a(z b − z c )} + z b .

⎪ z c (nDt0 ) − z c [(n − 1)Dt0 ]


†† ††
⎪ z c (t) = z c [(n − 1)Dt0 ] +
⎪ (31)

⎩ Dt0
[t − (n − 1)Dt]. In this figure, the displacement signal ze (t) is obtained through

(30) the integration for z (t). Through the transport delay module,
 e 
Dt0 denotes the sampling interval. The main model in Simulink lD † lD
signals ze t + and z t + could be obtained. Input
is shown in Figure 5(a). The module ‘1-D Look up Table1, 4, 2, V e V
† † †† the three variables into the dynamic Equation (31) and calcu-
3, 6’ export the signals zb zc z b z c z , respectively. The module ‘1-
b late the value of f (t). Then treat f(t) as the input signal ze (t)
D Look up Table5’ exports the force of the shock absorber. and carry out a new process.
Modules ‘Transport Delay1’ and ‘Transport Delay’ delay the
lD
input signal for a time of ; then, signals 4. Validation of the model through comparison of
  V
lD † lD theoretical value and simulated results
ze t + , z t+ are exported from the modules.
V e V
The Simulink main model is shown in Figure 5(b) and the It’s known before that the actual power spectral density of the
Simulink flowchart is shown in Figure. road irregularities obtained by inverse Fourier transform is
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 11

Figure 5. Simulink model of the back-calculation process: (a) integration of the acceleration signals, (b) Main model of the back-calculation and (c) Flow chart of the back-
calculation.
12 Y. JIANG ET AL.

basically consistent with the theoretical value. The monorail the figure, it can be seen that the simulated values of the
vehicle model is set up in the UM software as follows. Inverse power spectral density of the track irregularities are in good
Fourier transform is performed on the power spectral density agreement with the theoretical values.
function Sv0 (V) of the different tracks of the United States,
respectively. The theoretical value of the time response of the Error analysis:
road irregularities ze0 is obtained. By using it as a road exci- It can be seen from the figure that there is still a slight difference
†† ††
tation for the driving wheels, acceleration signals z b and z c between the simulation value and the theoretical value. Because
could be obtained. Input them into the back-calculation Simu- the process of the inverse Fourier transform will not cause
link model and the simulated value of the time response of the much error, there are only three possible causes:
road irregularities ze is obtained. The power spectral density of
the road irregularities ze could be calculated as Sv (V). When the 1. As mentioned above, the pitching movement of the bogie is
value Sv (V) is consistent with Sv0 (V), it means the back-calcu- neglected in the back-calculation of the road irregularities,
lation of the road irregularities is successful. but it is still taken into account in the UM simulation model.
The theoretical value and simulation value of power spectral That difference will result in a slight difference between the
density of the track irregularities are shown in Figure 6. From simulation value and the theoretical value.

Figure 6. Comparison of the theoretical value and the simulated result of the track irregularity power spectral density: (a) US rail level 6, (b) US rail level 5, (c) US rail level 4.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 13

2. The algorithms of the two software are different. PARK 3. The vehicle model set up in the UM software is much more
method is an implicit solver of the second order with vari- complicated.
able step size, for stiff ODE (ordinary differential equation)
and DAE (differential algebraic equation) in the UM soft- However, the simulated values of the track irregularity
ware. RK4 method is the 4th order Runge–Kutta method power spectral density are in good agreement with the theoreti-
with a constant step size used in the Simulink model, but cal values, which shows the back-calculation for the road irre-
the solver could not be used for DAE in the UM simulation. gularities is successful.
If the equations of motion are stiff, the PARK method is
much more efficient both for ODE and DAE, especially
when the calculation of the Jacobian matrices option is 5. Validation of the model through comparison of
turned on. However, there is no such algorithm in the soft- experiment data
ware of Simulink. That difference between the algorithms The model shall be proved correct if the track irregularity power
will result in a slight difference between the simulation spectrum density calculated through the test data of the adjacent
value and the theoretical value. two vehicles is consistent. According to standard ‘Interim

Figure 7. Comparison of the back-calculation result of the two vehicles at different speed: (a) 40 km/h, (b) 50 km/h, (c) 60 km/h, (d) 70 km/h.
14 Y. JIANG ET AL.

Provisions on strength design and test identification for railway agree with that of No.2 car, which means this process is success-
vehicles under the speed of 200 km/h and above’ (2001), the fre- ful (Figure 7).
quency range affecting the stability of bogies is 0.5–10 Hz. Band Error analysis:
pass filtering is applied to the measured acceleration data to get As shown in the figures, there are still some differ-
the signals in the frequency range of 0.5–10 Hz. The experimen- ences between the back-calculation results of two vehicles,
tal vehicles are the head car (motor car) and No. 2 car (trailer). which were supposed to have been caused by the following
The values of mb and mc are set to be different in the dynamic factors:
equations of the two vehicles. After the process of the back-cal-
culation in Simulink, theoretical values of the time responses of 1. The influence of the wheel out of round. Wheel polygon
the road irregularities according to the test data of the two exists in many vehicles, including the monorail (Lv et al.
vehicles are gotten as ze1 and ze2 . The data of No. 1 car with a 2017), which will result in the differences of calculated
time range of [t1 , t2 ] are selected for comparison. There is a data between the two vehicles.
2sc 2. Due to aging and wear of the rubber components (Wu
time delay of for No 2. car. So, the data of No.2 car with a
V  et al. 2015, 2016b), the real values of the parameters (Wu
2sc 2sc and Chi 2015) are somewhat different from the theoretical
time range of t1 + , t2 + are selected for comparison.
V V ones.
The power spectral densities of ze1 and ze2 are shown as follows. 3. Measurement error. The acceleration sensor may not be
As shown in the figures, the back-calculation results of No.1 car tightly attached to the surface of the object under test. A

Figure 8. Comparison of the amplitude–frequency response of the simulated and experimental acceleration of the vehicle at different speed: (a) 40 km/h, (b) 50 km/h, (c)
60 km/h, (d) 70 km/h.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 15

tiny crack can also lead to data errors. Because of manual 6. Validation of the model through comparison of
operation, there are some differences between the actual the experiment and the simulation
mounting position of the acceleration sensor and the theor-
This is the most crucial step. All the errors of the previous steps
etical position.
will be accumulated in this step, which will result in the great
4. The actual car body and bogie are not entirely rigid bodies
differences between the simulation result and the experimental
as shown in the back-calculation model. The vibration pro-
data. No matter how accurate the simulation model is, it cannot
cess is more complicated than that of the back-calculation
fully represent the actual process. That’s why experiments are
model.
repeated no matter how much they cost. This process is to ver-
5. The process of band pass filtering causes the differ-
ify whether the simulation model can represent the actual
ences between the calculation result and the experiment
vehicle.
data.
No. 2 vehicle is selected as the research object. Through
6. As mentioned above, the pitching movement of the bogie is
back-calculation for the experimental acceleration signals of
neglected in the back-calculation of the road irregularities,
the car body and bogie, road excitation for the driving wheels
which also contributes to the errors.
is obtained. By inputting the road excitation into the UM simu-
lation model, accelerations of the car body and bogie are simu-
Nevertheless, the back-calculation results of No.1 car are lated. When the simulated result of the acceleration is
consistent with that of the No.2 car, which proves the validity consistent with the experimental data, it means the back-calcu-
of the model by comparing experimental data. lation is precise enough.

Figure 9. Comparison of the amplitude–frequency response of the simulated and experimental acceleration of the bogie at different speed: (a) 40 km/h, (b) 50 km/h, (c)
60 km/h, (d) 70 km/h.
16 Y. JIANG ET AL.

The experimental sampling frequency is 2000Hz, but the 8. Conclusion


UM simulation model only allows for a sampling frequency
Because the monorail track beams hang high in the air above
of 100 Hz, so it is necessary to resample the experimental
the ground, it is extremely hard and dangerous to measure
data. The resample equation is shown as follows:
road irregularities for monorail. In order to establish a
   more accurate simulation model to reflect the actual working
†† †† (k − 1)f1
z (k) = z round +1 conditions, this paper introduces a back-calculation method
2 1 f2 (32) for road excitation of the monorail. Dynamic equations are
(k = 1, 2, 3 . . . ). set up in this paper. However, there exist many difficulties
in solving these transcendental equations. Besides, in order
††
In this equation, z 1 substitutes for the measured acceleration to solve the dynamic model, 10 sensors are needed to be
†† mounted in the vehicle and 14 acceleration signals need to
data; z 2 , the resampled acceleration data. f1 , f2 represent
be measured. As is known to all, even if the vehicle is
their respective frequency. The function round(x) is the round-
under the same road excitation, due to differences between
ing process for x.
the real vehicle structure, the multi-rigid body model and
In many types of research, the accuracy verification method
the mathematical expressions, acceleration results of simu-
to verify the accuracy of the simulation model is to verify
lation and experiment are different. If there are too many
whether the amplitude–frequency response of the simulation
acceleration parameters in the back-calculation formula,
results is consistent with the experimental data. Comparison
errors of each acceleration signal will accumulate into the
between simulation result and the experimental data of the
†† back-calculation result and make a large error. By neglecting
car body acceleration z c is shown as follows:
the subtle influence of the pitching angle on the vertical accel-
The comparison between simulation result and the
†† eration of the bogies, this dynamic model is greatly sim-
experimental data of the bogie acceleration z b is shown as
plified. Only two acceleration signals are needed in this
follows:
dynamic model, which means the calculation result could
Theoretically speaking, the experiment data and simulation
be much closer to the measured data. Besides, the calculation
data should be the same. However, errors in each step have
process is much simplified.
inevitably resulted in some differences. As can be seen, the
In order to validate the model in the simulation process, the
amplitudes and main frequencies of Figure 8 and Figure 9 are
monorail vehicle model is set up in the UM software. The
similar, which means the simulation is accurate enough. Differ-
inverse Fourier transform, with its advantages of less errors,
ences between the two results could be caused by the following
faster operation speed and better generality over other methods
factors.
like the white noise filtering method, secondary filtering
Error analysis:
method and trigonometric series method, is adopted to get
the simulated road irregularities. The simulated power spectral
1. Differences among the back-calculation model in Simulink,
density of road irregularities could be calculated through the
UM simulation model and the real vehicle structure will
back-calculation for simulated accelerations. In spite of the
result in simulation errors.
differences between the back-calculation model and the UM
2. The resample process will cause errors.
simulation model, differences between the algorithms of the
3. Errors in the process of inverse Fourier transform are accu-
two software and the error of the inverse Fourier transform,
mulated in this step.
the back-calculation result of power spectral density of road
4. Errors in comparing the theoretical value and the simulated
irregularities is consistent with the theoretical value, which
result also exist in this step.
shows the simplification of the back-calculation model doesn’t
5. Errors in comparison of experiment data of different
result in a large error. Through comparing the back-calculation
vehicles also exist in this step.
results of the two vehicles in a monorail, it is proved that this
6. Due to aging and wear of rubber components, real values of
method can be applied to experiments accurately. The final
parameters somewhat differ from theoretical ones.
and the most crucial step is the validation of the model by com-
7. Model application field
paring the experiment and the simulation. All the errors of the
previous steps will be accumulated in this step and result in the
According to the standard Department of Science and
differences between the simulation result and the experimental
Technology Education (2001), the frequency range of
data. Through the back-calculation of the experimental accel-
vibration affecting the stability of bogies is 0.5–10 Hz. Band
eration signals of the car body and bogie, road excitation for
pass filtering is applied to the measured acceleration data
the driving wheels is obtained. By inputting the road excitation
to get the signals in the frequency range of 0.5–10 Hz. This
into the UM simulation model, the accelerations of the car
proves that the model is valid only in low frequencies. How-
body and bogie are obtained. When the simulated result of
ever, as for the vibration in high frequencies, the rigid multi-
the acceleration agrees with the experimental data, it means
body dynamic model is not precise enough. Each component
the back-calculation is precise enough.
of the vehicle needs to be considered as a flexible body. In
Band pass filtering is applied to the measured acceleration
order to get more precise data, a much more complicated
data to get the signals in the frequency range of 0.5–10 Hz
model and more test parameters are needed. In spite of
according to the standard ‘interim provisions on strength
that, this model is good enough for the study of vehicle
design and test identification for railway vehicles under the
stability.
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 17

speed of 200 km/h and above’, which shows that the model can Gou, H., Zhou, W., and Yang, C., 2018. Dynamic response of a long-span
be valid only in low frequencies. However, as for the vibration concrete-filled steel tube tied arch bridge and the riding comfort of
monorail trains. Applied Sciences-Basel, 8, 1–22.
in high frequencies, the rigid multibody dynamics model is not
Jia, C., Xu, W., and Wei, L., 2013. Study of railway track irregularity
precise enough. Differences between the back-calculation standard deviation time series based on data mining and linear
model in Simulink, UM simulation model and the real vehicle model. Mathematical Problems in Engineering. DOI: 10.1155/2013/
structure should be taken into account. 486738.
It could be noticed in the paper that the deformation of the Jiang, Y., et al., 2018. Comparison of the curve negotiation properties of two
different articulated monorail vehicles. Proceedings of the Institution of
girder deformation is excluded, If the girder deformation was Mechanical Engineers, Part F: Journal of Rail and Rapid Transit.
considered, the vehicle–bridge coupling dynamics should be Jiang, Y., et al., 2019a. Detection and alleviation of the abnormal vibration
carried out and the FE model of the bridge should be set up. of the monorail: based on experiment and simulation. Journal of Low
With this complicated model set up, there would be too Frequency Noise Vibration and Active Control, DOI: 10.1177/
many more signals to be measured. Errors of each signal will 1461348419825605.
Jiang, Y., et al., 2019b. Multi-parameter and multi-objective optimisation
accumulate into the back-calculation result and make a large
of articulated monorail vehicle system dynamics using genetic algor-
error. In fact, unlike the suspended monorail girder, defor- ithm. Vehicle System Dynamics, DOI:10.1080/00423114.2019.156655.
mation of the straddle type monorail girder is relatively Jiang, Y., Wang, Y., and Xie, Q., 2015. Scheme and analysis of a kind of
small. Theoretically speaking, neglecting the effect of the defor- suspended monorail’s bogie structure. Electric Drive for Locomotives,
mation can somewhat cause errors, though more large errors 6, 56–59.
Jiang, Y., Wang, Y., and Xie, Q., 2017. Analysis of the dynamics and sta-
could be avoided in actual tests. The effect of other factors
tionarity of suspend monorail. Urban Mass Transit, 4, 97–100.
like an earthquake (Wu et al. 2016a) is also neglected. Kim, Y.S., et al., 2008. Dynamic model for ride comfort evaluations of the
Nevertheless, this model is accurate enough for researches of rubber-tired light rail vehicle. Vehicle System Dynamics, 46, 1061–1082.
vehicle stability. Kim, C.W., and Kawatani, M., 2006. Effect of train dynamics on seismic
response of steel monorail bridges under moderate ground motion.
Earthquake Engineering & Structural Dynamics, 35, 1225–1245.
Disclosure statement Kim, C.W., Kawatani, M., and Kanbara, T., 2013. Seismic behavior of steel
monorail bridges under train load during strong earthquakes. Journal of
No potential conflict of interest was reported by the authors. Earthquake and Tsunami, 7, 1–17.
Law, S.S., and Zhu, X.Q., 2005. Bridge dynamic responses due to road sur-
face roughness and braking of vehicle. Journal of Sound and Vibration,
Funding 282, 805–830.
Lee, C.H., et al., 2005. Dynamic response analysis of monorail bridges
This work was supported by National Key R& D Program of China [grant
under moving trains and riding comfort of trains. Engineering
number 2016YFB1200501, 2017YFB1201304-13
Structures, 27, 1999–2013.
Lee, C.H., Kawatani, M., and Kim, C.W., 2006. Dynamic response of a
monorail steel bridge under a moving train. Journal of Sound and
ORCID Vibration, 294, 562–579.
Yongzhi Jiang http://orcid.org/0000-0001-7449-2325 Li, C., 2016. Design of the lower chassis of a monorail personal rapid tran-
sit (MPRT) car using the evolutionary structural optimization (ESO)
method. Structural & Multidisciplinary Optimization, 54, 1–11.
References Lin, L., 1982. Track random excitation functions. China Railway Science, 3,
74–112.
Abdulhameed, D.O., Abdelrahim, A.B., and Dessouki, A.K., 2015. Lombaert, G., and Conte, J.P., 2012. Random vibration analysis of dynamic
Proposed design model for singly-symmetric overhanging monorail I- vehicle-bridge interaction due to road unevenness. Journal of
beams. Journal of Constructional Steel Research, 115, 346–358. Engineering Mechanics, 138, 816–825.
Bai, L., Liu, R., and Sun, Q., 2015. Markov-based model for the prediction Lv, K., et al., 2017. Influence of wheel eccentricity on vertical vibration of
of railway track irregularities. Proceedings of the Institution of suspended monorail vehicle, experiment and simulation. Shock and
Mechanical Engineers Part F – Journal of Rail and Rapid Transit, 229, Vibration, 2017, 1–10.
150–159. Ma, Z., Dong, Y., and Liu, H., 2018. Forecast of non-equal interval track
Bao, Y., Li, Y., and Ding, J., 2016. A case study of dynamic response analy- irregularity based on improved grey model and PSO-SVM. IEEE
sis and safety assessment for a suspended monorail system. ACCESS, 6, 34812–34818.
International Journal of Environmental Research and Public Health, Moghimi, H., and Ronagh, H.R., 2008. Development of a numerical model
13, 1121–1138. for bridge–vehicle interaction and human response to traffic-induced
Cai, C., et al., 2019. Dynamic interaction of suspension-type monorail vibration. Engineering Structures, 30, 3808–3819.
vehicle and bridge: numerical simulation and experiment. Mechanical Muller, S., 1978. Engineering operation and state of development of the H-
Systems and Signal Processing, 118, 388–407. Bahn system. Siemens Review, 45, 523–527.
Chen, G., and Zhai, W., 1999. Numerical simulation of the stochastic pro- Naeimi, M., Tatari, M., and Esmaeilzadeh, A., 2015. Dynamic interaction
cess of railway track irregularities. Journal of Southwest Jiaotong of the monorail–bridge system using a combined finite element multi-
University, 34, 138–142. body-based model. Proceedings of the Institution of Mechanical
Choi, C.K., and Song, M.K., 2000. Analysis of high-speed vehicle-bridge Engineers, Part K: Journal of Multi-body Dynamics, 229, 132–151.
interactions. Computer Engineering, 1240–1247. Nassif, H.H., 2003. Model validation for bridge-road-vehicle dynamic inter-
Department of Science and Technology Education, 2001. Interim action system. Journal of Bridge Engineering, 8, 112–120.
provisions on strength design and test identification for railway vehicles Perrin, G., Soize, C., and Duhamel, D., 2013. Track irregularities stochastic
under the speed of 200 km/h and above, People’s Republic of China modeling. Probabilistic Engineering Mechanics, 34, 123–130.
Railway Ministry, 21, 1–47. Pu, Q., et al., 2015. Bogie of hanged-type monorail vehicle. Electric Drive
Du, Z., Wen, X., and Zhao, D., 2017. Numerical analysis of partial abrasion for Locomotives, 2, 90–93.
of the straddle type monorail vehicle running tyre. Transactions of Tsunashima, H., 2003. Dynamics of automated guideway transit vehicle
Famena, 41, 99–112. with single-axle bogies. Vehicle System Dynamics, 39, 365–397.
18 Y. JIANG ET AL.

Wang, F., 2009. Vehicle system dynamics. China Railway Publishing House, Wu, X., Chi, M., and Gao, H., 2016b. Damage tolerances of a railway axle
15–19. in the presence of wheel polygonalizations. Engineering Failure
Wang, L., Kang, X., and Jiang, P., 2016. Vibration analysis of a multi-- Analysis, 66, 44–59.
span continuous bridge subject to complex traffic loading and vehicle Wu, X., Chi, M., and Wu, P., 2015. Influence of polygonal wear of railway
dynamic interaction. Journal of Civil Engineering, 20, 323–332. wheels on the wheel set axle stress. Vehicle System Dynamics, 53, 1535–1554.
Wang, N., and Wang, F., 2015. Prediction for track irregularity based Xiao, T., et al., 2015. Dynamic simulation of hanged-type monorail vehicle.
on gray interval prediction model. Railway Computer Application, 24, Electric Drive for Locomotives, 2, 10–13.
1–3. Xu, L., Zhai, W., and Gao, J., 2017. A probabilistic model for track random
Wang, H., and Zhu, E., 2018. Dynamic response analysis of monorail steel- irregularities in vehicle/track coupled dynamics. Applied Mathematical
concrete composite beam-train interaction system considering slip Modelling, 51, 145–158.
effect. Engineering Structures, 160, 257–269. Xu L., Zhai, W., and Li, Z., 2018. A coupled model for train-track-bridge
Wei, L., Zeng, J., and Wu, P., 2014. Indirect method for wheel-rail force stochastic analysis with consideration of spatial variation and temporal
measurement and derailment evaluation. Vehicle System Dynamics, evolution. Applied Mathematical Modelling, 63, 709–731.
52, 1622–1641. Zhang, L., et al., 2016. The influence of vehicle–tire contact force area on
Wen, X., et al., 2017. The coupling dynamic model and vibration response vehicle–bridge dynamic interaction. Canadian Journal of Civil
of straddle type monorail vehicle. Journal of Vibration, Measurement & Engineering, 43, 769–772.
Diagnosis, 37, 462–468. Zhang, N., and Xia, H., 2013. Dynamic analysis of coupled vehicle–bridge
Wu, X. and Chi, M., 2015. Parameters study of Hopf bifurcation in railway system based on inter-system iteration method. Computers & Structure,
vehicle system. Journal of Computational and Nonlinear Dynamics, 10, 114–115, 26–34.
031012. Zhong, J., 2013. The structure and analysis of the turnout of the straddle
Wu, X., Chi, M., and Gao, H., 2016a. Post-derailment dynamic behaviour type monorail vehicle. Beijing: China Communications Press.
of a high-speed train under earthquake excitations. Engineering Failure Zhou, J., 2005. Research of Chong Qing straddle-type monorail vehicle’s ride
Analysis, 64, 97–110. index. Thesis (Master). Beijing Jiaotong University.

You might also like