Strojnícky časopis – Journal of MECHANICAL ENGINEERING,
VOL 69 (2019), NO 3, 107 - 112
MODELING OF THE CONTROLLED AIR SPRING
RÁGULÍK Jiří1, SIVČÁK Michal1
1
TUL – Technical University of Liberec, Faculty of Mechanical Engineering, Department of Applied Mechanics,
Studentská 1402/2, 461 17 Liberec I, Czech Republic, e-mail:
[email protected]Abstract: The pneumatic springs are frequently used as a machine component. They are mainly used for passive
vibration damping with the possibility to regulate the working height of the springs. The paper describes the
creation of a mathematical model of a spring controlled thru the PID regulator. At conclusion are summarized the
limits of PID control and sketches the future possibility of substituting the PID controller by a neural network.
KEYWORDS: Air springs, PID regulation, mathematical model
1 Introduction
Pneumatic springs are most commonly used to suspend vehicle axles, as well as elements to
eliminate machine vibration. These springs use compressed gas inside the bellows as a
springing medium. Their main advantage over cheaper and simpler conventional springs is the
ability to change their characteristics by changing the internal pressure [1]. Appropriate
regulation of the internal overpressure can thus achieve the desired system properties, which
can be changed as needed even during the operation [2].
Fig. 1 Assembled control system
The control system formed (Fig. 1) consists of the spring itself, the frame, the weight, the
compressed air source, the SMC VY1A00 pressure valve, the optoNCDT 1402-200 laser
position sensor, and an electronic circuit that includes the controller. The lower spring lid,
through which overpressure is supplied to the spring, is clamped to the frame. The upper spring
cover is fastened to the movable part of the frame by a screw connection, the plate on which
the weight is placed. This position is read by the laser position sensor. The output of the laser
sensor is the voltage that is applied to the electronic circuit, in which the difference of measured
DOI: 10.2478/scjme-2019-0037, Print ISSN 0039-2472, On-line ISSN 2450-5471 ©2019 SjF STU Bratislava
and desired voltage is applied to the PID controller. The sum of the outputs of each controller
component is inverted and applied to the pressure regulator terminals.
2 Mathematical model of spring
In order to verify the functionality of the whole system and especially the assembled
electronic circuit in the software MapleSim, it was necessary to create a mathematical model of
the pneumatic spring, i.e. its force characteristic.
Several simplifications have been considered in deriving this characteristic. The most
important thing was considering the gas in the spring as an ideal gas. If no liquefaction
or humidification of air is expected, this model is accurate enough. Another simplification is
the consideration of spring deformation as reversible changes.
𝐹(𝑥) = (𝐴0 + 𝐴1 ∙ 𝑥 + 𝐴2 ∙ 𝑥 2 ) ∙ 𝑝𝑝 + 𝑘1 ∙ 𝑥 + 𝑘2 ∙ 𝑥 2 + 𝑘3 ∙ 𝑥 3 (1)
𝑝0 ∙ 𝑉0
𝐹(𝑥, 𝑝0 ) = ( 1 1 ) ∙ (𝐴0 + 𝐴1 ∙ 𝑥 + 𝐴2 ∙ 𝑥 2 )
2
𝑉0 − 𝐴0 ∙ 𝑥 − 2 𝐴1 ∙ 𝑥 − 3 𝐴2 ∙ 𝑥 3
(2)
+𝑘1 ∙ 𝑥 + 𝑘2 ∙ 𝑥 2 + 𝑘3 ∙ 𝑥 3
Measurements of the examined air spring DUNLOP 2 ¾ x 3, necessary to obtain unknown
constants, were made on a loading device TIRA test (Fig. 2). For the purpose of this
measurement, a pneumatic circuit was constructed to allow air inlet and outlet and pressure and
temperature measurements. Measured data was fitted to the mathematical model by least
squares method (Fig. 3), which is a standard approach in regression analysis to approximate the
solution of overdetermined systems of equations.
Fig. 2 Air spring with attached pneumatic circuit clamped in load machine
First, the measurement was carried out without air intake, thereby obtaining data to
determine the bellows stiffness coefficients k1, k2 and k3. Subsequently, the measurement was
already carried out with the supply of compressed air. The obtained data together with the
already known stiffness coefficients are used to obtain the coefficients of the effective spring
surface A0, A1 and A2 (Table 1). In equations (1) and (2), p0 stands for initial overpressure, pp
for overpressure and V0 for initial volume of the air spring. The volume of the investigated
108 ©2019 SjF STU Bratislava Volume 69, No. 3, (2019)
spring at a free length is 160 ml, which is the value given by the manufacturer. The accuracy of
this parameter has been verified by measuring with a measuring cylinder.
Table 1 Calculated coefficients
Coefficient Value Unit
k1 5202.87 N∙m-1
k2 74825 N∙m-2
k3 5403911 N∙m-3
A0 0.0022638 m2
A1 0.04305 m
A2 -0.06123 1
The accuracy of the acquired model is sufficient to verify system functionality and simulate
its behaviour. The absolute value of the approximated and measured force difference reaches a
maximum of 89 N when measured with the highest free length pressure, namely 1622 N. The
difference value is 5.5 % of the measured value. The maximum percentage difference occurs at
minimum pressure at free length, which is outside the assumed working area of the spring under
investigation. In the entire working area of the given spring, the error is up to 10 %, which can
be considered as very satisfying.
Fig. 3 Approximated (red) and measured (blue) force dependence on spring height for
individual pressures for closed system
The events that cause inaccuracy take place mainly in the material of the pneumatic spring
- rubber bellows [3] [4]. These phenomena were not taken into account in the mathematical
model to maintain its simplicity.
Volume 69, No. 3, (2019) ©2019 SjF STU Bratislava 109
Fig. 4 Model created in MapleSim used to verify system functionality
Fig. 5 Mathematical model of air spring created in MapleSim
The acquired mathematical model was created in MapleSim software and its functionality
was confirmed (Fig. 4). (Fig. 5) shows a detail of an air spring submodel.
3 Spring height control
There are a number of approaches to setting up a PID controller. From the simplest method
of trial-and-error, through the manual setup of the Wade method, to the time-tested and long-
used methods, such as the Ziegler-Nichols method, which has been used. The method consists
in bringing the system with the integration and derivative components turned off to the stability
110 ©2019 SjF STU Bratislava Volume 69, No. 3, (2019)
limit, where output of the control loop has stable and consistent oscillations. It is a state of the
system where constant amplitude of oscillations is obtained after excitation. The proportional
gain is set to a minimum and a step load is performed. Proportional gain is increased until the
system is just on the stability level. The gain of k* at which this state has been reached is one
of the two parameters needed to determine the appropriate controller setting. The second
parameter is the period of critical oscillations T* [5].
The system was excited by a free fall of a 7 kg steel weight from a constant height of
approximately 25 cm. This excitation was chosen for its simplicity and good repeatability. The
value of the critical gain k* = 1.5 and the period of critical oscillations T* = 0.24 s were
determined. The values of the gain of the individual components of the regulator were
calculated from these values. In this way a virtual PID controller created in LabVIEW
environment has been set up.
The electronic circuitry including the PID controller was assembled on a solderless field
from operational amplifiers, resistors, capacitors, cables, and several other electronic
components.
The control circuit thus formed was able to maintain a constant required height under static
load throughout the working height of the spring (Fig. 6).
Fig. 6 Time dependence of measured change of height with on (blue) and off (red) control
CONCLUSION
A mathematical model of an air spring was created and the accuracy of the model was
evaluated. The data needed to determine the unknown coefficients was obtained by
measurement on a loading machine. When creating the mathematical model of the spring, a
methodology was also developed for describing any bellows pneumatic spring operating as an
open (1) or closed (2) system.
The control loop was then assembled and amplifications of the individual components of the
PID controller were determined by well-repeatable excitations. The system is able to maintain
a constant height under static loading, but under dynamic loading, the height control speed is
insufficient. The solution could be to replace the analog PID controller with a microcontroller,
which predictively regulates spring height using neural networks. The control of pneumatic
Volume 69, No. 3, (2019) ©2019 SjF STU Bratislava 111
springs has been dealt with by the Nagai, Moran and Tamura in the article [6], which describe
the regulation of train suspension.
When a significant acceleration of altitude control and possible improvement of other
parameters, such as the amount of air required and energy consumed, is found, practical
applications could be considered wherever active vibroisolation is required. Such cases are, for
example, the driver's seat and the means by which the vibration deteriorates working conditions
and can also cause motion sickness. The great advantage of the assembled electronic circuit is
the very low purchase price of the individual components, compared to prices of PID controllers
available in stores.
ACKNOWLEDGEMENTS
This publication was written at the Technical University of Liberec as part of the project
“Research, development and application of advanced methods and technologies in technical
mechanics, biomechanics and strength and elasticity” with the support of the Specific
University Research Grant, as provided by the Ministry of Education, Youth and Sports of the
Czech Republic in the year 2019.
REFERENCES
[1] Sivčák, M., Škoda, J. “Verification of Active Controlled Air Spring Model”, Journal of
Vibroengineering 13 (4), pp. 788 – 791, 2011.
[2] Šklíba, J., Škoda, J. “About the latest possibility of stiffness reduction of ambulance
couch”, Journal of Vibroengineering 11 (3), 2009.
[3] Zigo, M. “Determination of extreme properties of springs created from two mutually
symmetrically sloped rubber prisms”, Strojnícky časopis – Journal of Mechanical
Engineering 48 (2), pp. 97 – 105, 1997.
[4] Bouazara, M. “Optimization of the suspension parameters and tire forces analysis on a
vehicle model using an analytical approach”, Strojnícky časopis – Journal of Mechanical
Engineering 56 (5), pp. 257 – 272, 2005.
[5] Votrubec, R. “Control System of One-Axis Vibration-Insulation Platform with
Gyroscopic-Stabilizer”, Mechatronics 2013: Recent Technological and Scientific
Advances. 1. vyd. Berlin: Springer International Publishing Switzerland, pp. 733 – 739,
2014. ISBN 9783319022932.
[6] Nagai, M., Moran, A., Tamura, Y., Koizumi, S. “Identification and control of nonlinear
active pneumatic suspension for railway vehicles, using neural networks”, Control Eng.
Practice 5 (8), pp. 1137 – 1144, 1997. DOI: https://doi.org/10.1016/S0967-
0661(97)00107-X
112 ©2019 SjF STU Bratislava Volume 69, No. 3, (2019)
