Journal of Mechanical Engineering Vol 17(2), 79-91, 2020

Vibration Suppression Control of

Three Inertial Systems

Duong Minh Duc*, Nguyen Manh Linh, Dao Quy Thinh, Do Trong Hieu
Hanoi University of Science and Technology
*
[email protected]

ABSTRACT

This paper concerns with vibration suppression control for three inertia
resonance system which can be used as the model for many moving
mechanisms widely used in industry such as steel rolling mills, flexible arms,
large-scale space structures, etc. Since the mechanical resonance which
causes vibration is unavoidable, especially at high speed operation, vibration
suppression plays a key role in improving the accuracy of the system. In this
research, a linear quadratic regulator-based speed controller with an
integrator that can effectively suppress the torsional vibration is applied to
control the system. Instead of using conventional state observer that is
sensitive to noise and model uncertainties, the extended state observer is used
to overcome the disturbance and uncertainty problems. Moreover, this
observer also gives us the load torque information that is used to improve the
load response. Simulation results show the effectiveness of the proposed
controller. In addition, the comparison to conventional PID controller also be
done to verify the advantages of the proposed approach.

Keywords: Three inertial system; extended state observer; linear quadratic

regulator; vibration suppression

Introduction

Nowadays, high precision and fast response motor driver systems are widely
used in industry such as steel rolling mill, robot manipulator, electrical vehicle,
etc. Generally, these moving mechanisms can be regarded as multi-inertia
systems with several inertia moments, gears and springs. Theoretically,
vibration is unavoidable in above mentioned systems where motion is involved
due to the mechanical resonance. For most motions, conventional
proportional-integral-differential (PID) controller is sufficient to meet the
desired requirements due to the fact that the nature frequency of the mechanical
___________________
ISSN 1823-5514, eISSN2550-164X Received for review: 2019-12-22
© 2020 Faculty of Mechanical Engineering, Accepted for publication: 2020-03-20
Universiti Teknologi MARA (UiTM), Malaysia. Published: 2020-07-15
Duong Minh Duc et al.

system is quite high in comparison with the motion maneuver. However, for
applications where both fast dynamic and high precision positioning are
required, vibration suppression plays a key role in improving the system
performance.
To deal with the vibration problem, most researches treat the motion
mechanism as a 2-inertia system which comprised of mass, spring and
damping. Then, various control methods are employed to suppress the
vibration such as resonant ratio control [1], state feedback control [2,3], linear
quadratic Gaussian (LQG) control [4,5], linear quadratic control with extended
state observer (ESO) [6], active disturbance rejection control [7], fractional
order PID-k controller [8], model predictive control [9], back-stepping position
control [10], adaptive speed control [11]. Since the simplified model and the
real system may be different in the number of resonant frequencies, i.e., the
number of links of a manipulator is three or more, the effectiveness of the
control system based on the 2-inertia model may be degraded in practice.
To further improve the system performance in term of vibration
suppression, multi inertia model, i.e., three and more inertia system, is
considered [12]. In [13], a three-inertia system is controlled by using PI/PID
control. In addition, a modified integral plus proportional plus derivative (m-
IPD) controller is used to suppress vibration in 3-inertia system. Fuzzy
controller with differential evolution is also employed to control this 3-inertia
system [12]. Despite of improved performance, the tuning procedure of the
aforementioned methods are quite complicated due to the presence of the
modeling error and uncertainties.
In this research, a linear quadratic regulator (LQR) control with an
extended state observer (ESO) is proposed to handle the vibration problem of
the 3-inertia system. The proposed method is not only effective in torsional
vibration suppression but also robust against parameters variation since the
ESO is employed to estimate the immeasurable state variables and load
disturbance. In advanced, the tuning procedure of the method is simplified
which plays a very importance role in practical applications. Particularly, the
method can also be extended to higher order inertia systems.
To this end, the paper is organized as follows. The mathematical model
of the 3-inertia system is introduced in section II. Section III shows the design
procedure of the ESO and the LQR. Numerical simulations which verify the
validity and effectiveness of the proposed strategy are shown in section IV.
The conclusions are shown in the last section of the paper.

Three-Inertia System Model

A typical 3-inertia system, which consists of three rigid inertias and two
torsional shafts, is shown in Figure 1.

80
 Vibration Suppression Control of Three Inertial Systems

Figure 1: Three inertial system.

In this figure, wm is the motor angular speed, wL1 is the angular speed
of load 1, wL2 is the angular speed of load 2, Jm is the motor inertia, JL1 is the
inertia of load 1, JL2 is the inertia of load 2, Tm is the motor torque, TL is the
load torque, T12 is the torsional torque of shaft 1, T23 is the torsional torque of
shaft 2, Ks1 is the stiffness of shaft 1 and Ks2 is the stiffness of shaft 2. The 3-
inertia system can be modeled by the following equations:
"#
!
⎧𝐽! "$ = 𝑇! − 𝑇%&
⎪"'"# = 𝐾 (𝜔 − 𝜔 )
(% ! )%
⎪ "$
"#$"
𝐽)% = 𝑇%& − 𝑇&* (1)
⎨"' "$
#%
⎪ "$ = 𝐾(& (𝜔)% − 𝜔)& )
⎪ "#$"
⎩𝐽)& "$ = 𝑇&* − 𝑇)

Equation (1) can be rewritten in state-space form as:

𝑥̇ = 𝐴𝑥 + 𝐵% 𝑇) + 𝐵& 𝑇!
(2)
𝑦 = 𝐶𝑥

where
𝑥 = [𝜔! 𝑇%& 𝜔)% 𝑇&* 𝜔)& ]'
0 −1/𝐽! 0 0 0
⎡𝐾 0 −𝐾(% 0 0 ⎤
⎢ (% ⎥
𝐴=⎢ 0 1/𝐽)% 0 −1/𝐽)% 0 ⎥ (3)
⎢ 0 0 𝐾(& 0 −𝐾(& ⎥
⎣ 0 0 0 1/𝐽)& 0 ⎦

81
Duong Minh Duc et al.

0 1/𝐽!
⎡ 0 ⎤ ⎡ 0 ⎤
⎢ ⎥ ⎢ ⎥
𝐵% = ⎢ 0 ⎥ 𝐵& = ⎢ 0 ⎥
⎢ 0 ⎥ ⎢ 0 ⎥
⎣−1/𝐽)& ⎦ ⎣ 0 ⎦
𝐶 = [1 0 0 0 0];

From Equation (2), the transfer function from Tm to wm is:

, ,
!+ (#) (# , &!-. )(# , &!-, )
%+ (#)
=' , &! , )(# , &! , )
(4)
+ (# /. /,

In which, wa1, wa2 are anti-resonant frequencies and wr1, wr2 are resonant
frequencies of the system and are given by.

% 2&" 32&# 2 2&" 32&# 2 2&" 2&#

𝜔0% = @( + 5 &#) − A( + 5 &#)& − 4 (5)
√& 4$" $# 4$" $# 4!

% 2&" 32&# 2&# 2&" 32&# 2&# & 2&" 2&#

𝜔0& = @( + ) + A( + ) −4 (6)
√& 4$" 5$# 4$" 5$# 4!

2 2&" 32&# 2
D 4&" + 4$"
+ 5 &#E −
% ! $#
𝜔6% = C 2 2 32 2 4 34 34
(7)
√&
A( &" + &" &# + &#)& − 4 ! $" $# 𝐾(% 𝐾(&
4! 4$" 5$# 4$" 4$#

2 2&" 32&# 2
D 4&" + 4$"
+ 5 &# E +
% ! $#
𝜔6& = C 2 2 32 2 4 34 34
(8)
√&
A( 4&" + &"4 &# + 5 &#)& − 4 ! 4 $"4 $# 𝐾(% 𝐾(&
! $" $# $" $#

These resonant frequencies cause vibrations and therefore degrade the

system performance. The controller must be designed not only for
speed/position tracking, but also for vibration suppression.

Control of 3-Inertia System

State and Disturbance Observer

To implement the state feedback controller, all the state variables should be
known. However, not all state variables are available for measure because of

82
 Vibration Suppression Control of Three Inertial Systems

the high cost, mounting constraints or noise. In this case, a state observer is a
good solution to estimate the immeasurable state variables. In this paper, the
ESO [14] is chosen because of its better performance over other observers such
as high-gain and sliding-mode observers in term of robustness against
disturbance.
Consider the fifth order system (2), in order to estimate the states and
loads disturbance of this system by the ESO, the following transformation
matrix T is employed:

1 0 0 0 0
⎡0 −𝐽! 0 0 0 ⎤
⎢ 4! ⎥
⎢1 0 2&"
0 0 ⎥
𝑇=⎢ 4$" 4! ⎥ (9)
⎢0 −(𝐽! + 𝐽)% ) 0 −
2&"
0 ⎥
⎢ 4! 4! 34$" 4$" 4! ⎥
⎣1 0 2&"
+ 2&#
0 2&" 2&# ⎦

Let 𝑥 ∗ = 𝑇 8% 𝑥. Then, Equation (2) can be written as:

∗ ∗ %
⎧𝑥̇ % = 𝑥& + 4! 𝑢
⎪𝑥̇ &∗ = 𝑥*∗
⎪
⎪𝑥̇ ∗ = 𝑥 ∗ − 2&" 𝑢
* 9 4#
! (10)
⎨𝑥̇ 9∗ = 𝑥:∗
⎪ ∗ 2# % %
⎪𝑥̇ : = 𝑎(𝑡) + #&" D + E 𝑢
⎪ 4! 4! 4$
⎩𝑦 = 𝑥%∗

with
4! 34$" 34$# 8% 2&" 32&# 2&" 2&" 2&#
𝑎(𝑡) = (−𝐾(% 𝐾(& 4! 4$" 4$#
)𝑥&∗ + (4 + 4$"
+ 4!
)𝑥9∗ − 4 𝑇) (11)
$# ! 4$" 4$#

By treating a(t) = x6*(t) as an augmented state variable, h(t) as the

derivative of a(t) which is unknown, the system state-space model in Equation
(10) becomes:

83
Duong Minh Duc et al.

∗ ∗ %
⎧ 𝑥̇ % = 𝑥& + 4! 𝑢
⎪ 𝑥̇ &∗ = 𝑥*∗
⎪ ∗ ∗ 2&"
⎪ 𝑥̇ * = 𝑥9 − 4! # 𝑢

𝑥̇ 9∗ = 𝑥:∗ (12)
⎨
2# % %
⎪𝑥̇ :∗ = 𝑥;∗ + 4#&" D4 + 4 E 𝑢
⎪ ! ! $
⎪ 𝑥̇ ;∗ = ℎ(𝑡)
⎩ 𝑦 = 𝑥%∗

To estimate the new states variable 𝑥 ∗ , the following ESO is used:

𝑒 = 𝑧% − 𝑦
⎧ %
𝑧̇ = 𝑧& − 𝛽% 𝑔% (𝑒) + 4 𝑢
⎪ % !
⎪𝑧̇& = 𝑧* − 𝛽& 𝑔& (𝑒)
⎪ 2
𝑧̇* = 𝑧9 − 𝛽* 𝑔* (𝑒) − 4#&" 𝑢 (13)
!
⎨𝑧̇ = 𝑧 − 𝛽 𝑔 (𝑒)
9 : 9 9
⎪ #
⎪𝑧̇: = 𝑧; − 𝛽: 𝑔: (𝑒) + 2&" D % + % E 𝑢
⎪ #
4! 4! 4$
⎩𝑧̇; = 𝛽; 𝑔; (𝑒)

where 𝛽< (i=1,..,6) are observer gains, 𝑧< (i=1,..,6) are estimated values of 𝑥<∗ ,
and 𝑔< (𝑒) (i=1,..,6) can be either linear or nonlinear functions. In this paper,
𝑔< (𝑒) = 𝑒 is chosen. With appropriate values of 𝛽< , the estimation of 𝑥<∗ and
the system states 𝑥N = 𝑇𝑥N ∗ can be obtained. Then the estimation of load torque
𝑇O) can be determined by

4 34 34 8% 2 32 2 4 4 4
𝑇O) = D−𝑎(𝑡) + D−𝐾(% 𝐾(& !4 4$"4 $# E 𝑥&∗ + (4 + &"4 &# + 4&")𝑥9∗ E !2 $"2 $#
! $" $# $# $" ! &" &#
(14)

Controller Design
To extend the bandwidth of the motion mechanism, vibration caused by
mechanical resonant must be suppressed. Besides, other requirements of a
standard controlled system such as stability, minimal tracking error, etc., must
also be fulfilled. To achieve the aforementioned goals simultaneously, a state
feedback control with an extended integral of error state variable is utilized to
eliminate the steady-state error. The design procedure is as followings.

84
 Vibration Suppression Control of Three Inertial Systems

The block diagram of the state feedback control system is shown in

Figure 2 in which Ki is the integral gain, Kd is feed forward gain to compensate
the load torque, and F = [f1 f2 f3 f4 f5] is the state feedback gain.

Figure 2: Proposed controller schema

In order to determine the integral gain and the state feedback gain, we
first define a new state variable S as:
$
𝑆 = ∫= (𝜔! − 𝜔∗ )𝑑𝜏 (15)

Then, by differentiating both sides of Equation (15), it yields:

𝑆̇ = 𝜔! − 𝜔∗ = 𝐶𝑥 − 𝜔∗ (16)

Combining Equation (2) and Equation (16), a new state-space model in which
the torque Tm is substituted by the control signal u is derived as follows:

𝑥̇ 𝐴 0 𝑥 𝐵 𝐵 0 𝑇)
T ̇U = T U T U + T &U 𝑢 + T % UT U (2)
𝑆 𝐶 0 𝑆 0 0 −1 𝜔∗

In steady-state, 𝑥̇ = 0 and 𝑆̇ = 0, then the steady-state solution xs, Ss and us

must satisfy the following equation:

𝐵% 0 𝑇) 𝐴 0 𝑥( 𝐵
T U T ∗U = − T U T U − T & U 𝑢( (18)
0 −1 𝜔 𝐶 0 𝑆( 0

Substituting Equation (18) into Equation (17), it gives:

𝑥̇ 𝐴 0 𝑥 − 𝑥( 𝐵
T ̇U = T UT U + T & U (𝑢 − 𝑢( ) (19)
𝑆 𝐶 0 𝑆 − 𝑆( 0

85
Duong Minh Duc et al.

Define a new state variable Z as:

𝑍% 𝑥 − 𝑥( 𝑥̇
𝑍=W X = T𝑆 − 𝑆 U , 𝑍̇ = T ̇ U , 𝑞 = 𝑢 − 𝑢( (20)
𝑍& ( 𝑆

Then Equation (19) becomes:

𝐴 0 𝑍% 𝐵
𝑍̇ = T U W X + T & U 𝑞 = 𝐴̅𝑍 + 𝐵\𝑞 (21)
𝐶 0 𝑍& 0

Choose the performance index J as follows:

> ∗ )& ∗ )& ∗ )&
𝐽 = ∫= [𝛼(𝜔! − 𝜔! + 𝛽(𝜔)% − 𝜔! + 𝜀(𝜔)& − 𝜔! + 𝛿(𝑆 − 𝑆( )& +
𝛾(𝑢 − 𝑢( )& ]𝑑𝑡 (22)

or
>
𝐽 = ∫= (𝑍 ' 𝑄𝑍 + 𝑅𝑞& )𝑑𝑡 (23)

where

𝛼 0 0 0 0 0
⎡0 0 0 0 0 0⎤
⎢ ⎥
0 0 𝛽 0 0 0⎥
𝑄=⎢ and 𝑅 = 𝛾
⎢0 0 0 0 0 0⎥
⎢0 0 0 0 𝜀 0⎥
⎣0 0 0 0 0 𝛿⎦

Then, the state feedback gain K = [K1K2] of the LQR control for system in
Equation (21) subject to performance index in Equation (22) can easily be
computed. Consequently, the control input q is:

𝑞 = −𝐾𝑍 = −𝐾% 𝑍% − 𝐾& 𝑍& or𝑢 − 𝑢( = −𝐾% (𝑥 − 𝑥( ) − 𝐾& (𝑆 − 𝑆( )(24)

Since 𝑢( = −𝐾% 𝑥( − 𝐾& 𝑆( in steady state, Equation (24) becomes:

$
𝑢 = −𝐾% 𝑥 − 𝐾& 𝑆 = −𝐾% 𝑥 − 𝐾& ∫= (𝜔! − 𝜔∗ )𝑑𝜏 (25)

By considering Figure 2, it can be realized that F = [f1 f2 f3 f4 f5] = K1 and Ki =

K2. Since the load torque of which the sudden change may cause speed drop
and initiate hazardous torsional vibration is observed by the ESO, the feed
forward compensation gain 𝐾" = 1 is sufficient.

86
 Vibration Suppression Control of Three Inertial Systems

With the ESO, the state estimation errors can approach zero with
arbitrarily dynamic [15]. Then, the stability of the control system can be
guaranteed by the LQR controller [16].

Numerical Simulation
In this section, numerical simulations are carried out to verify the effectiveness
of the proposed control strategy. The systems parameters are provided in Table
1.

Table 1: Systems parameters

Name Symbol Value

Motor inertia Jm 1552 Kgm2
Inertia of load 1 JL1 1000 Kgm2
Inertia of load 2 JL2 542 Kgm2
Stiffness of shaft 1 KS1 5.93x106 Nmrad-1
Stiffness of shaft 2 KS2 5.93x106 Nmrad-1
Load Disturbance TL 0.2x106 Kgm2

After being well tuned, weighting matrices Q and R in Equation (23)

are chosen with α = 1000, β =107, ε =7.106, δ = 1013 and γ = 1. Then, the state-
feedback gain is K = [3,0560.105 2,2702 2,0347.104 -1.0770 2.7274.103
3,1623.106]. The gains of ESO are β1 = 6ω0, β2 = 15ω02, β2 = 20ω03, β4 = 15ω04,
β5 = 6ω05 and β6 = ω06 where ω0 = 1000.
The performance of the proposed control strategy is first investigated
by step-response analysis. A constant reference speed, i.e., 30 rad/s, is used
and the corresponding systems response is shown in Figure 3. It can be
observed that the speed of the motor and loads quickly track the reference
speed without vibration. In addition, the robustness of the controller against
the load disturbance is also verified in this simulation. In detail, an external
disturbance is introduced at time instance 1s. It can be seen that the influence
of the disturbance is quickly compensated by the controller in about 0.1 s.
In order to show the validity of the ESO, the error between the real and
the observed speeds of motor, load 1 and load 2 are calculated and shown in
Figure 4. It can be realized that these errors are extremely small even in
transient-state, and quickly converge to zero in steady-state.
To show the advantage of the proposed control strategy over the
conventional PID controller, comparative simulation is also carried out. A PID
controller of which parameters are well tuned by particle swarm optimization
(PSO) technique is used for system (2). The comparative speed responses of
the PID controller and the proposed controller are provided in Figure 5. It can
be seen that although being optimized, the PID controller still shows poor
performance, i.e., large overshoot with torsional vibration. In contrast, the

87
Duong Minh Duc et al.

proposed controller shows much better performance with smooth transient-

state whilst the vibration is completely removed.

Figure 3: Velocity responses of the system with the proposed controller

Figure 4: Speed estimation errors

88
 Vibration Suppression Control of Three Inertial Systems

(a) Motor velocity

(b) Load 1 velocity

(c) Load 2 velocity

Figure 5: Comparison between PID and the proposed controller

Finally, the robustness of the proposed control strategy against the

parameter variation is test. The moment of inertial of the loads used in
simulation is two times larger than the one used in control design. The
simulation result in this case is shown in Figure 6. Although the system model
is inaccurate, the system speed still tracks the reference one without vibration.
However, the overshoot is unavoidable in this case. The reason may come from
the transient response of the ESO.

89
Duong Minh Duc et al.

Figure 6: Velocity responses of the system with the proposed controller in

case of parameter uncertainty.

Conclusion

In this paper, we proposed the LQR controller based ESO for a three-inertia
system in order to achieve vibration suppression and disturbance rejection. The
unknown states and load torque are estimated using ESO. Then, the state
feedback and feedforward compensation for load torque are employed in
control design. Theoretical analysis and numerical simulations show that the
proposed control strategy guarantees the system stability, good performance
and robustness with parameters variation as well as load disturbance.
Particularly, the control strategy can easily be extended to be applied to any
order inertia systems.

References
[1] Hori, Yoichi, Hideyuki Sawada, and Yeonghan Chun, "Slow resonance
ratio control for vibration suppression and disturbance rejection in
torsional system," IEEE Transactions on Industrial Electronics 46.1, 162-
168 (1999).
[2] Song, Seung-Ho, et al., "Torsional vibration suppression control in 2-mass
system by state feedback speed controller," Proceedings of IEEE
International Conference on Control and Applications IEEE (1993).
[3] Hori, Yoichi, Hiroyuki Iseki, and Koji Sugiura, "Basic consideration of
vibration suppression and disturbance rejection control of multi-inertia
system using SFLAC (state feedback and load acceleration
control)," IEEE Transactions on Industry Applications 30.4, 889-896
(1994).
[4] Ji, Jun-Keun, Dong-Choon Lee, and Seung-Ki Sul, "LQG based speed
controller for torsional vibration suppression in 2-mass motor drive

90
 Vibration Suppression Control of Three Inertial Systems

system," Proceedings of IECON'93-19th Annual Conference of IEEE

Industrial Electronics. IEEE (1993).
[5] Ji, Jun-Keun, and Seung-Ki Sul, "Kalman filter and LQ based speed
controller for torsional vibration suppression in a 2-mass motor drive
system," IEEE Transactions on industrial electronics 42.6, 564-571
(1995).
[6] Zhang, Ruicheng, and Chaonan Tong, "Torsional vibration control of the
main drive system of a rolling mill based on an extended state observer
and linear quadratic control," Journal of Vibration and Control 12.3, 313-
327 (2006).
[7] Zhao, Shen, and Zhiqiang Gao, "An active disturbance rejection based
approach to vibration suppression in two‐inertia systems," Asian Journal
of Control 15.2, 350-362 (2013).
[8] Ma, Chengbin, and Yoichi Hori, "Backlash vibration suppression control
of torsional system by novel fractional order PIDk controller," IEEJ
Transactions on Industry Applications 124.3, 312-317 (2004).
[9] P. J. Serkies and K. Szaba, “Model predictive control of the two-mass with
mechanical backlash,” Computer Applications in Electrical Engineering,
170–180 (2011).
[10] M. Mola, A. Khayatian, and M. Dehghani, “Backstepping position control
of two-mass systems with unknown backlash,” 2013 9th Asian Control
Conference, ASCC (2013).
[11] S. Brock, D. Luczak, K. Nowopolski, T. Pajchrowski, and K. Zawirski,
“Two Approaches to Speed Control for Multi-Mass System with Variable
Mechanical Parameters”, IEEE Transactions on Industrial Electronics, 64
(4), 3338–3347 (2017).
[12] Ikeda, Hidehiro, and Tsuyoshi Hanamoto, "Fuzzy controller of three-
inertia resonance system designed by differential evolution." Journal of
International Conference on Electrical Machines and Systems 3 (2),
(2014).
[13] G. Zhang and J. Furusho, “Control of three-inertia system by PI/PID
control”, Trans of IEE Japan 119-D (11), 1386-1392 (1999).
[14] Huang, Yi, and Wenchao Xue, "Active disturbance rejection control:
methodology and theoretical analysis," ISA transactions 53.4, 963-976
(2014).
[15] D. Yoo, S. S. T. Yau, Z.Gao, “On convergence of the linear extended
observer,“ Proceedings of the IEEE International Symposium on
Intelligent Control, Munich, Germany, 1645–1650 (2006).
[16] Anderson, B. D., & Moore, J. B., “Optimal control: linear quadratic
methods,” Courier Corporation (2007).

91