FUTOJNLS 2018 VOLUME- 4, ISSUE- 1.

PP- 199 - 210

Futo Journal Series (FUTOJNLS)

e-ISSN : 2476-8456 p-ISSN : 2467-8325
Volume-4, Issue-1, pp- 199 - 210
www.futojnls.org

Research Paper July 2018

Performance Evaluation of PD and LQR Controller for

Coupled Mass Spring Damper System
Okubanjo, A.A*, Oyetola, O.K, Ade-Ikuesan, O.O, Olaluwoye, O.O, Alao, P.O

Department of Electrical Electronics and Computer Engineering, Olabisi Onabanjo University

Ibogun Campus, P.M.B.5026 ,Ifo, Ogun State, Nigeria

*Corresponding Author’s E-mail: [email protected]

Abstract

In automation and mechatronics applications, mass spring damper (MSD) system plays
significant role in ensuring model serviceability and safety. However, the position control of
mass spring damper system has been considered as an integral strategy to investigate the
system‘s dynamic. The system’s dynamic of two or more of degree of freedom (DOF) of
mechanical system is quite challenging to control. In this paper, the issue of performance
evaluation of two control schemes, proportional-derivate (PD) and Linear Quadratic
Regulator (LQR) which have not received attention in the literature have been addressed.
However, effective control of a system depends largely on the accuracy of the mathematical
model that predicts its dynamics behaviour. The mathematical model for the coupled mass-
spring-damper system (CMSDS) was based on a set of nonlinear second-order ordinary
differential equations and to simulate the dynamic accurately Lagrangian and Euler-
Lagrange equations were successfully derived and established. The proposed control
schemes were implemented with the aid of MATLAB and SIMULINK R2014b version 8.4
simulation tool to investigate the system performance. The simulations results revealed that
the linear quadratic regulator (LQR) controller has affinity to enhance the robustness,
transient and steady state performance of a coupled mass spring damper system than
convectional PD controller.

Keywords: PD control, LQR Linear-quadratic regulator, Augmented-Lagrange, Mass-spring-

damper system

Nomenclatures
m1, m2 Translating inertial elements (masses)
k1,k2 Spring stiffness coefficients
b1,b2 Viscous damping coefficients
L Lagrange’s function
T Kinetic energy
P Potential energy
R Rayleigh’s dissipative function
x1,x2 Generalized coordinates displacement
H Generalized forces
̇ ̇ Velocities of the masses m1 and m2
̈ ̈ Accelerations of the masses m1 and m2

199 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

1. Introduction
The momentous contributions of the mass spring damper (MSD) system are notably
reflected in the fields of Automation and Mechatronics Engineering. It is extensively used as
a compliant actuator in robot manipulator (Li & Yin, 2017),serves as a shock absorber in
vehicle suspension system (Akpakpavi, 2017; Allamraju, 2016; El-Nasser, Ahmed, Ali,
Ghazaly, & Abd El-Jaber, 2015; Katal & Singh, 2012; Srinivasan, Kumar, & Basha, 2016),
piezoelectric vibration energy harvester(Bahiah, Noor, & Ahmad, 2017; Caliò et al., 2014;
Kundu & Nemade, 2016), used as anti-vibration devices for sky-scraping building to improve
serviceability and suitable for teaching dynamic modeling in control systems. In recent times,
the deployment of the MSD system is significantly on high demand in health services to aid
the control of rehabilitation robot in reinstating the incapacitated and aged people and in
hybrid vehicle suspension system to enhance passenger trip comfort and vehicle stability.
The application of coupled mass spring damper system is common in dynamic control and
industrial mechanics for system design and analysis to mimic physical system. However,
simulation and control of MSD system are challenging issue for two or more degrees of
freedom (DOF) in engineering. The complexity in the system analysis and design are as a
result of complicated geometry of the system’s motion in solving large system. Therefore,
control engineers try to establish a mathematical model that depicts the MSD system
dynamics and manipulate certain system’s parameter so as to constrain the force to have a
desired system response and performance. For the MSD system shown in Figure 1, the goal
is to design control strategies that will stabilize the system position to the desired trajectory
under the influence of an obstacle. In addressing the position control of mass spring damper
system, several control algorithms have been proposed and tested and new control
schemes are still emerging. A Matlab user friendly interface for a MSD model is developed
such that there are possibility of selecting different control strategies through model
parameter modification in the work (Ahmad, Tarik, Bhutta, & Tariq, 2016). A comparative
analysis is performed on two degree of freedom MSD system in Matlab-Simulink for model
derived through Newton’s law and Lagrange’s equation of motion in Sivák & Hroncová
(Sivák & Hroncová, 2012). The simulation result confirmed identical solution. Furthermore, a
proportional-integral-derivative (PID) controller is implemented to minimize the body
acceleration of car automobile suspension model. (Tandel, Deshpande, Deshmukh, &
Jagtap, 2014). Kiran and Shivalingappa, (2013) investigated the effect of damper position in
system with multi-degree of freedom and observed that the displacement of an oscillator
that in direct contact with the input force show dominant energy absorption. Frankovský,
Hroncová, Delyová and Virgala, (2013); Rahman, Mushtaq, Ali and Anjam, (2011); Sharma,
Pathak and Dwivedi, (2014), presented mathematical modeling of a mass spring damper
system in Matlab-Simulink. A polynomial structures approach is proposed for position control
of coupled mass spring damper system (Rannen, Ghorbel, & Braiek, 2017). Malas and
Chatterjee, (2016) new control approach for inducing self-sustained oscillation of a single
degree-of-freedom MSD mechanical system. It was well established that the proposed
control model is adept of generating stable self-excited oscillation at the natural frequency
regardless of the value of the control gain. (Ge, Huang, & Lee, 2004) proposed backstepping
and model reference adaptive control approaches for addressing the position control
challenges in chain of multiple mass spring damper system.

The goals of this paper are to formulate mathematical models based on the Lagrangian and
Lagrange-Euler for a coupled mass spring damper system and to compare the effectiveness

200 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

of proportional-derivative (PD) and Linear quadratic regulator position control algorithms in

stabilizing the dynamic system and improve system performance.

2. Modeling of Coupled Mass Spring Damper System

Considering the mass spring damper system with two degree-of-freedom movement shown
in Figure1.

Figure 1: Two-DOF coupled- MSD system adapted from (Kiran & Shivalingappa, 2013)

This system consists of two bodies, the mass m1 and the mass m 2 linked to a parallel
spring-damper configuration with spring constants k 1 and k 2 and coefficient of viscous
friction b1 and b 2 for mass m1 and m 2 respectively. We denote by x1 , the displacement of
mass m1 , by x 2 the displacement of the mass m 2 and the force F t  , acts on mass 1 and
the energy is transferred to mass 2 through the spring and some part of the energy is
absorbed by the damper. Finally, the model assumed that there is no friction between the
masses and the floor and that both the springs and the dampers are linear with negligible
spring weight. The control goal is to design controller algorithms that will stabilize the
positions of the motion. An energy property based approach is adopted over the
conventional force based approach-Newton’s law or D’Alembert principle because it permits
significant elucidation of the geometry of the system motion for solving large complex
systems and also eliminates sketchy rewriting all forces acting on the body. In addition to the
adopted approach, Rayleigh’s dissipation function is integrated into the Lagrange equation to
compensate for dissipative force in the mechanical system and this is refer to as augmented
Lagrange equation.

Thus, the Lagrangian function L is defined as the difference of the system’s kinetic and
potential energy (Caliò et al., 2014; Tandel et al., 2014; Udwadia & Cho, 2013; Urrea &
Pascal, 2017). Hence, the system’s kinetic energy T equals the sum of the kinetic energy of
the individual mass and the system’s potential energy P is the sum of the potential energy of
mass m1 and mass m 2 . So that the Lagrangian L is defined as

L T P 1
Denoting by T 0,P 0 and T 1,P 1 the kinetic and potential energy of the mass m1 and mass m 2
respectively, we obtain

L  T 0T 1   P 0  P1  2
According to the conservation of energy, the kinetic and potential energy for each of the
system’s bodies are as follows:

201 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

.
1
T 0 m1 x1 2 3
2

P0
1
k1 x12 4
2

1
5
.
T 1 m2 x2 2
2

P1 
1
k 2  x1  x 2 
2
6
2

And the Rayleigh dissipative function R are

R0
1 .2
b1 x 7 
2
2
1  . . 
.

R 1  b2  x1  x 2  8
2  
 

Substituting equation (2-5) into the Lagrangian (2) gives

1   1
.


L   m1 x1  m2 x 2 2   k 1x12  k 2 x12  2k 2 x1 x 2  k 2 x 22  9
.
2

2  2
 

 x1  u 
Define the vector     as the generalized displacements and the matrix H    where
 x2  0 
u is the input force F (t). Then, the Augmented Euler- Lagrange equation can be written in
as:

d  L   L   R 
  H 10
dt   .       . 
   

The partial derivatives of equations [6, 7] and [8] gives:

 L 
   k1 x1 k 2 x 2  k 2 x 2 11

 1
x

 
 L  .

 .   m1 x 1

 1
x
12

202 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

 
d  L 
13
..
 m x
dt   x. 
1 1

 1

 
 R 
14
. . .
 b x 
 .  1 1 2 1 2 2b x  b x
  x1 

 L 
   k 2 x 2  k 2 x1
 x 2  15
 
 L 
16
.

 .   m 2 x 2

 2x

 
d  L 
17
..
 m x
dt   x.  2 2

 2 

 
 R 
18
. .

 .   b2 x 1  b2 x 2

  x2 

Substituting the derived partial derivatives in equation (9) leads to:

m1 x1  b1  b2  x1  b2 x2  k1  k 2 x1  k 2 x2  u

.. . .

19
.. . .
m2 x2  b2 x1  b2 x2  k 2 x1  k 2 x2  0 20
Equations (18) and (19) describe the system’s motion. Hence, for simplicity these equations
can be written in matrix form using the defined generalized displacement as column vector.
Thus, (8)-(9) can be written equally as

 .  21
M    B ,   K    Hu
..

 
 
where the matrices M   , B , , K   and Hu are
.. .

 
m 0   .  b1  b2  b2 
M     1  , B  ,      b
 0 m2     b2 
2
22
k  k 2  k 2  1
K     1  ,H   
  k2 k2  0 

203 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

To enhance accurate simulation of dynamic of complex system a higher-order differential

equations are replaced with a first-order matrix differential equations such that the system
represent a state space model. Define the column vector column vector

T
 .   . .

x   x T x T    x1 x1 x2 x2  23
   

Using the model description (18), (19) and the state variables defined as column vector in
(22). For the sake of simplicity, the state space model can be equivalently written as
.
x  Ac .x  Bc .u, 24
y  C c .x  Dc .u

Or equivalently in state space form

 . 
 x.1   x1 
x    F 
 2   Ac . x 2   Bc . 
v   v1  0
 .1 
 v 2 
 v 2  25
 x1 
  F 
y  C c . x 2   Dc . 
 v1  0
 v 2 

This system is linear and time invariant, where Ac , Bc , C c , Dc are the corresponding
matrices of the coupled mass spring damper system denoted as follows

 0 0 1 0 
  k1  k 2  k 2  b1  b2  b2   0 
 m  1 
m1  1 0 0 0 0 0 
, B   m2 , C 
m1 m1
A
0
2
0 0 1    0 1 0 0   ,D    26
 0 0 0 
 k 2  k2  k2  b2   
 m2  0 
m2 m2 m2 

The input ut  is represented by the excitation forces:

 F t 
u t     27
 0 

And the output y t  is represented by the displacements x1 t  and x2 t  :

204 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

 x1 
1 0 0 0   x2 
y t    . 
0 1 0 0  v1 
 v 2 

Once the matrices Ac , Bc , C c , Dc are computed from the parameter values, the system’s
dynamic can be simulated to any input response either in time or frequency domains. Thus,
the space-space model permit the possibility of representing the mathematical model in
MATLAB-SIMULINK environment.

3. Simulation

It is easy to simulate the dynamic of the linear coupled MSD model in MATLAB R2014b
version 8.4, a programming software with diverse functions and tool boxes for simulating
control. . The response to the systems is subjected to input functions such as a step unit,
and random function that mimic frequent experienced obstacle. The mathematical model is
modeled to accommodate for accurate dynamic simulation. In the Simulink environment, the
model is defined as a state space system with position of the masses as the output.
Subsequently, two control algorithms proportional-derivative (PD) controller and Linear
Quadratic Regulator (LQR) are implemented and tested for performance evaluation. A PD
controller combine P-action and D-action to adjust the model error such that a control signal
ut  is based on the error discrepancy between the reference input signal and the output
measure in conformity with generic time domain description of PD controller as cited in
(Sasaki, Iba, Hongu, Nakamura, & Moriwaki, 2016) as

det 
u t  k p et   k D 28
dt
et   r t   yt  29
In addition, a more robust control scheme, linear quadratic regulator LQR is proposed and
we determine feedback control law of the form u t   kx   R 1 B T Px t  , that optimally
maintain the masses position irrespective of the disturbances introduced into the model.
Invariably, LQR is based on minimizing the cost function
J   x T t Qxt   u T t Ru t d t


0
30
The function inside integral is a quadratic form and the matrices Q and R are usually
symmetric. It is assumed that R is positive definite and Q is positive semi definite and that
the relative weighting chosen for Q and R decide the relative emphasis placed upon
reducing errors and saving energy. In other words, there is always stringent requirement to
penalize the state at the expense of control energy or vice-versa. Hence, the controller K,
that minimize the cost function J is based on calculating the positive definite solution of
algebraic Riccati equation (ARE)
AT P  PA  PBR 1 B T P  Q  0 31
In our example, since we want to control the masses position so at minimize the error, so we
will choose larger values for the corresponding entries of Q such that the entries
corresponding to the masses velocities are set to zero. This implies that the entries do not
pose any limitations for this case.

205 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

Figure 2: Simulink Implementation of both PD and LQR controllers for CMSDS

Table 1: Parameters of the CMSDS (Ahmad, et al., 2016; Sivák & Hroncová, 2012)

Parameter of the Value Unit

CMSD
m1 250 kg
m2 300 kg
k1 750 N/m
k2 250 N/m
b1 50 N/ms
b2 100 N/ms
External force F t 

4. Results and Discussion

An m-file scripts is developed for the mathematical model of a coupled mass spring damper
system (CMSDS) based on the augmented Lagrange equation and a Simulink models are
built to simulate all the control algorithms. The parameter values for coupled mass-spring-
damper-system presented in Table 1 are used for the simulation. Initially, the model is
subjected to force of a unit of 10N and further replaced by a road load profile as a common
experienced obstacle in our daily activities. This is built with signal builder in Simulink
environment so as to allow possibility of controlling the positions with proposed control
algorithms. Figure 2 indicated the response of the step unit force of 10N and it is evident
from the simulation result that the system sustained continuous oscillation as a result of
system’s dynamic (poles) instability. This revealed that parts of the system’s poles lie on the

206 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

right-half-side (RHS) of the s-plane. Consequently, this dominates the transient response of
the system. However, PD and LQR controllers are implemented so as to stabilize the
system’s dynamic by shifting the poles to stable location in s-plane. As shown in figure 3, the
time response of the closed loop system with the simulated PD and LQR controllers are
superimposed to expedite easy performance comparison which reduce the system
overshoot, peak tine, settling time and the steady state error, thus, the system’s response is
improved. In this figure, the time response for the mass positions m1 and m2 of PD and LQR
controllers are blue, yellow, magnate and red respectively. In figure4, a road load profile that
mimic an obstacle is modeled as an input to the model and PD and LQR controller are
integrated as a control, so as to compensate for the position error and as well stabilize the
position to zero with minimum overshoot. It is evident from the curve that both PD and LQR
are capable of stabilizing the mass spring damper position provided that the system’s
parameters are optimal selected. It is worth to note that parameter turning of PD controller
has been a challenging tasks. Similarly, the selection of weighting matrices in LQR is also
time intense and sometimes there is need to compromise between performance and control
effort in LQR controller. Nevertheless, the simulation results have proven that the linear
quadratic regulator (LQR) controller is superior to PD controller in terms of its faster
response and overshot minimization.

In the light of this, it can be concluded that LQR controller exhibit better response and
performance in controlling system’s dynamics of coupled mass spring damper system.
Figure 5, shows graphical positions error comparison of the proposed control strategies.

Figure 2: Step Response of CMSD system with Figure 3: Step Response of CMSD system
Controller. controlled by PD and LQR controllers

207 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

5. Conclusion

Figure 4: PD and LQR position control of CMSD Figure 5: PD and LQR error signal in
system subject to road load profile input signal position control of CMSD system

In this study, the performance of a PD and LQR position controlled of a coupled mass spring
damper system has been investigated. To ensure that the model mimic physical system and
dynamics are accurately simulated, the system dynamic is explicitly formulated based on the
energy property of Lagrange approach. A comparative analysis of the proposed control
algorithms are analytically examined. It is evident from the simulation result that both the PD
and LQR controllers are active and appropriate for refining the system’s response and
suitable for hands-on controlling of the position of a coupled mass spring damper system.
Simulations results showed that both PD and LQR satisfied all control specifications.
However, LQR response was more satisfactory in terms of robustness, transient and steady
state criteria. In the future, the nonlinearity in the model spring and damper would be
considered and the study would be extended to velocity and acceleration performance
evaluation. A more robust control algorithms such as H-infinity, Model Predictive control,
MPC and combination effect of LQR+Estimator would be implemented and tested in future
research.

References

Ahmad, H., Tarik, M., Bhutta, H. A. & Tariq, M. (2016). An interactive educational tool for
double control. An: Interactive educational tool for double spring- mass-damper
system analysis & control. In: Proceedings of the 3rd International Conference on
Engineering & Emerging Technologies (ICEET), Superior University, Lahore.
Akpakpavi, M. (2017). Modeling and control of a car suspension system using P , PI , PID,
GA-PID and Auto- Tuned PID Controller in Matlab / Simulink. Journal of
Multidisciplinary Engineering Science Studies, 3(3), 1506–1513.
Allamraju, K. V. (2016). Dynamic analysis of an automobile suspension system. IPASJ
International Journal of Mechanical Engineering (IIJME), 4(7), 15–21.
Bahiah, N., Noor, M. & Ahmad, M. R. (2017). Modeling the vibrational dynamics of

208 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

piezoelectric actuator by system identification technique, 7(3), 1506–1512.

https://doi.org/10.11591/ijece.v7i3.pp1506-1512
Caliò, R., Rongala, U., Camboni, D., Milazzo, M., Stefanini, C., de Petris, G., & Oddo, C.
(2014). Piezoelectric energy harvesting solutions. Sensors, 14(12), 4755–4790.
https://doi.org/10.3390/s140304755
El-Nasser, A., Ahmed, S., Ali, A. S., Ghazaly, N. M., & Abd El-Jaber, G. T. (2015). Pid
Controller of active suspension system for a quarter car model. International Journal
of Advances in Engineering & Technology, 8(6), 899–909.
https://doi.org/10.4273/ijvss.8.1.07
Frankovský, P., Hroncová, D., Delyová, I., & Virgala, I. (2013). Modeling of dynamic systems
in simulation environment MATLAB/Simulink – SimMechanics. American Journal of
Mechanical Engineering, 1(7), 282–288. https://doi.org/10.12691/ajme-1-7-26
Ge, S. S., Huang, L., & Lee, T. H. (2004). Position control of chained multiple mass-spring-
damper systems – Adaptive output feedback control approaches. International
Journal of Control, Automation, and Systems Vol., 2(2), 144–155.
J.Rahman, M.Mushtaq, A.Ali, Y.N.Anjam, S. N. (2011). Journal of Faculty of Engineering &
Technology. 18(92), 47–57. Retrieved from
www.pu.edu.pk/journals/index.php/jfet/index%0A
Katal, N. & Singh, S. K. (2012). Optimization of PID controller for quarter-car suspension
system using genetic algorithm. International Journal of Advance Research in
Computer Engineering & Technology, 1(7), 30–32.
Kiran, J. O., & Shivalingappa, D. (2013). Study on the effect of position of the dampers in
systems with 3 degrees of freedom. International Journal of Engineering Technology
and Advanced Engineering, 3(4), 155–163.
Kundu, S., & Nemade, H. B. (2016). Modeling and simulation of a piezoelectric vibration
energy harvester. Procedia Engineering, 144, 568–575.
https://doi.org/10.1016/j.proeng.2016.05.043
Li, Z., & Yin, Z. (2017). Position tracking control of mass spring damper system with time-
varying coefficients. In 2017 29th Chinese Control And Decision Conference (CCDC)
(pp. 4994–4998). IEEE. https://doi.org/10.1109/CCDC.2017.7979380
Malas, A., & Chatterjee, S. (2016). Amplitude Controlled adaptive feedback resonance in a
single degree-of-freedom mass-spring mechanical system. Procedia Engineering,
144, 697–704. https://doi.org/10.1016/j.proeng.2016.05.070
Rannen, S., Ghorbel, C., & Braiek, N. B. (2017). Nonlinear identification and control of
coupled mass-spring-damper system using polynomial structures. International
Journal of Advanced Computer Science and Applications, 8(5).
Sasaki, T., Iba, D., Hongu, J., Nakamura, M., & Moriwaki, I. (2016). Experimental evaluation
of a control system for active mass dampers consisting of a position controller and
neural oscillator. Journal of Physics: Conference Series, 744(1), 12040.
https://doi.org/10.1088/1742-6596/744/1/012040
Sharma, R., Pathak, D. K., & Dwivedi, V. K. (2014). Modeling & simulation of spring mass
damper system in simulink environment. In XVIII Annual International Conference of
the Society of Operations Management Theme: Operations Management in Digital
Economy (pp. 205–210).
Sivák, P., & Hroncová, D. (2012). State-space model of a mechanical system in
MATLAB/Simulink. Procedia Engineering, 48, 629–635.
https://doi.org/10.1016/j.proeng.2012.09.563
Srinivasan, G., Kumar, M. S., & Basha, A. M. J. (2016). Mathematical modeling and PID

209 Okubanjo, et al., Performance Evaluation of…

FUTOJNLS 2018 VOLUME- 4, ISSUE- 1. PP- 199 - 210

controller design using transfer function and root locus method for active suspension
system. Middle-East Journal of Scientific Research, 24(3), 622–627.
https://doi.org/10.5829/idosi.mejsr.2016.24.03.23115
Tandel, A., Deshpande, A. R., Deshmukh, S. P., & Jagtap, K. R. (2014). Modeling, analysis
and PID controller implementation on double wishbone suspension using
SimMechanics and Simulink. Procedia Engineering, 97, 1274–1281.
https://doi.org/10.1016/j.proeng.2014.12.406
Udwadia, F. E., & Cho, H. (2013). Lagrangians for damped linear multi-degree-of-freedom
systems. Journal of Applied Mechanics, 80(4), 41023.
https://doi.org/10.1115/1.4023019
Urrea, C., & Pascal, J. (2017). Parameter identification methods for real redundant
manipulators. Journal of Applied Research and Technology, 15, 320–331.

210 Okubanjo, et al., Performance Evaluation of…