Journal of Mechanics Engineering and Automation 3 (2013) 257-262 D DAVID PUBLISHING

Modeling and Simulation of Hydraulic Long

Transmission Line by Bond Graph

Dragan Pršić, Novak Nedić and Ljubiša Dubonjić

Faculty of Mechanical Engineering Kraljevo, University of Kragujevac, Kraljevo 36 000, Serbia

Received: February 6, 2013 / Accepted: March 7, 2013 / Published: April 25, 2013.

Abstract: This paper addresses the issue of modeling of the hydraulic long transmission line. In its base, such model is nonlinear with
distributed parameters. Since general solution in closed-form for such model in time-domain is not available, certain simplifications
have to be introduced. The pipeline in the paper has been divided to a cascaded network of  segments so that a model with lumped
parameters could be reached. For segment modeling, a standard library of bond graphs element has been used. On the basis of models
with lumped parameters, the effect of the number of segments, pipeline length and effective bulk modulus on the dynamics of long
transmission line have been analyzed.

Key words: Long transmission line, lumped parameters,  segments, bond graph.

1. Introduction approximations using Bessel functions.

In this paper, we have used the method of description
Existence of a long transmission line (LTL) in
of a long hydraulic pipe with lumped parameters in a
hydraulic systems makes their dynamics significantly
way that it was divided to n equal segments of Ls length.
complex. This is especially emphasized with building
 and T model with lumped parameters have been used
and mining machines, agricultural machines,
whose electrical analogies are given in the paper. On
transportation machines, machine tools and other
the basis of equivalent electric circuits, adequate bond
devices where connection between the actuators energy
graphs of these circuits were made and connected into a
source achieved by a long hydraulic line. Physical
cascade of n segments, which defines the mathematical
variables, pressure and volumetric flow, featuring the
model of the hydraulic pipe. On the basis of the
energy transfer along the hydraulic line, besides the
mathematical model and simulation, we have analyzed
time coordinate, depend on spatial coordinate as well.
the impact of certain parameters on the character of the
These physical variables’ dependency on spatial
transfer process and its results are given in the paper.
coordinate conditions spatial distribution cannot be
The paper is organized as follows: Section 2 starts
neglected in long hydraulic line modeling. Therefore
from the description of one-dimensional flow of
they are described with models with distributed
compressible viscous fluid through LTL.
parameters. Models with distributed parameters are
Approximation of this model is then introduced through
described with partial differential equations. Use of
a cascade of small lumped elements described by means
such a model in analysis of dynamic behavior in time
of common differential equations. For each element,
domain is not practical because it requires work with
there is a corresponding electric circuit, i.e., a bond
transcendent transfer functions and their
graph model. Based on the models from section 2,
Corresponding author: Dragan Pršić, assistant professor,
section 3 gives simulation results for LTL with input
Ph.D., research field: modeling and simulation. E-mail: flow as an excitation variable and a hydraulic orifice at
[email protected].
258 Modeling and Simulation of Hydraulic Long Transmission Line by Bond Graph

the end of the line. Conclusions are outlined in section 4. 1 2 ... n

Fig. 1 A cascaded network representing a transmission
2. Modeling of Hydraulic Long Transmission
line.
Line
then serially connect those models into the system
One-dimensional flow of compressible, viscous
model (LTL model).
fluid through the LTL is represented by a set of
What we usually find in literature are two
nonlinear partial differential equations [1]. Applying
single-lump approximations for a short fluid line [7-8].
physical principles of mass conservation, Newton’s
Those two approximations are known as  and T
second law and energy conservation leads to
circuits. Their electrical analogies are given in Fig. 2.
A p Q
  (1a) Same elements for modeling of the lumped elements
a 2 t x
were used in both models. Using capacitor (C element),
1 Q  p
 g sin   Q 2   (1b) we model fluid compressibility, using coil (L element)
A t 2dA2 x
we model fluid inertia, and with resistor (R element) we
That is a pair of quasi-linear hyperbolic partial
model losses due to friction. Difference is in the way
differential equations describing pressure change p and
those elements are connected. In the first case, there are
volumetric flow rate Q depending on time t and
two distinctive pressures: P1 on one end and P2 on the
distance x along the pipeline. Generally, there is no
other end of the segment. Same flow rate Qi is adopted
closed-form solution for these equations. The problem
along the whole segment length. In the second case,
is particularly expressed in case of turbulent flow
there is one distinctive pressure of Pi segment and two
which introduces stochastic parameters. Models with
flows Q1 and Q2 on the segment’s ends.
distributed parameters are described by partial
Which model is to be used depends on the remaining
differential equations and, in the complex domain,
part of the circuit where the given pipeline is located.
these models are of infinitesimally high order [2-5].
Namely, each segment is connected with the
In order to find analytical or numerical solution,
surrounding with two energy flows. On one end, the
certain simplifications have to be introduced. One
energy flow is P1Q1, and on the other P2Q2. This means
approach is based on division of non-homogenous
fields (P(x, t), Q(x, t)) on segments with homogenous that interaction of segments with the environment can
fields (P(t), Q(t)) of all physical variables. In other be described with four variables. In regard to the
words, instead of models with distributed parameters segment, two values must be independent, and the
we move to the model with lumped parameters. other two dependent, provided that the independent
Exactly this is the approach used in this study. LTL is values must be on different segment ends. For example,
observed as a cascaded network of small lumped it is not allowed for the values P1 and Q1 to be
elements (Fig. 1) where dynamics of each of them is independent. Therefore, it is possible to make four
described by common differential equations. combinations of independent-dependent variables:
Each of lumped elements presents spatial Q1Q2P1P2, Q1P2Q2P1, P1Q2Q1P2 and
abstractions of distributed physical properties. Physical P1P2Q1Q2. Model of the segment, i.e. the whole
dimensions of each segment are much smaller than the pipeline, depends on the combination appearing in the
shortest wavelength of interest. Connections between circuit. It should be noted that when the segment
lumped elements represent physical constraints on the cascade is made, regardless if we use  or T model,
physical quantities associated to the elements [6]. there are blocks inside the model shown in Fig. 3 that
Practically, we need to do a model of one segment and periodically repeat.
 Modeling and Simulation of Hydraulic Long Transmission Line by Bond Graph 259

Q1 Q1 Q1 Q1
R
Qi Q2
R
L
L
P1 C/2 Pi C/2 P2
P1 C/2 C/2 P2

Fig. 4 LTL model interfaces with  segments (

(a)  circuit interfaces).
Q1 Q2
Q1 Q1 Q2
R/2 R/2
Qi+1
R/2 R/2
L/2 L/2
L/2 L/2
P1 C P2
P1 Pi C P2
Pi+1

(b) T circuit
Fig. 5 LTL model interfaces with T segments (T
Fig. 2 Equivalent electrical analogy of short fluid line. interfaces).
Qi Qi+1 C:C I:L
R

L
Pi C Pi+1
0:Pi 1:Qi+1

i=1,2,3...
Fig. 3 Block of LTL model repeating periodically. R:R
Fig. 6 LTL model bond graph repeating periodically.
Difference in models comprising of  and T
C:C/2 I:L C:C/2
segments is in the end blocks—interfaces according to
the remaining of systems symmetric in regard to
0:P1 1:Qi 0:P2
internal blocks. Equivalent electrical circuits of end
blocks of the cascade with  segments ( interfaces)
are shown in Fig. 4 and cascades with T segments (T R:R
Fig. 7 LTL model interface bond graph with  segments
interfaces) in Fig. 5. ( interfaces).
Suitable bond graphs of the electric circuits in Figs. I:L/2 C:C I:L/2
3-5 are shown in Figs. 6-8 respectively.
On the basis of drawn causality it can be seen that 1:Qi+1 0:Pi+1 1:Q 2
all internal blocks are second rate models. The end
blocks for the  cascade have integral causality for R:R/2 R:R/2
one-junction and arbitrary causality for zero-junctions. Fig. 8 LTL model interface bond graph with T segments
(T interfaces).
It is similar with the T cascade where zero-junction
has integral causality and one-junctions the arbitrary cascade if the flow is independent value. If pressure is
one. independent value, differential causality is avoided
This means that the LTL model can have minimum with use of the T cascade.
order of (n - 1)*2 + 1 and maximum (n - 1)*2 + 3. For In this study, we observe the response of the system
example, cascade of n = 5 segments can be minimally 9 on the independent flow change. Diagram of hydraulic
order and maximum 11. Order of the model depends installation is shown in Fig. 9.
on the hydraulic circuit where the given LTL is found. As a source of excitation, variable flow rate source
In order to avoid the differential causality and reduce (Q1) is used. LTL is represented as a cascade of n
the numerical problem, recommendation is to use the  segments. At the end of the circuit there is a hydraulic
260 Modeling and Simulation of Hydraulic Long Transmission Line by Bond Graph

Q1 ... Q2 7
1 2 n
P1 P2
6

P2
RP 5

4
P1

Fig. 9 Schematic of a circuit for LTL simulation. 3

2 Q2
orifice of constant resistance. Model order is
determined with 2n + 1. 1

Q1
0

3. Simulation Analysis -1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
t [s]

On the basis of  model with interface shown in Fig. Fig. 10 Transient processes for n = 1 and Ls = 16 m.
4, behavior of LTL in circuit shown in Fig. 9 is 8

simulated. The aim is to analyze the impact of several 7

parameters on the transfer process. 6 P2

The parameters at which the simulation was 5

performed are Be = 1.4 × 109 Pa, Qref = 3.5 × 10-4 m3/s, 4

P1

-2 3 -5 2
d = 1 × 10 m,  = 860 kg/m ,  = 9.7 × 10 m /s. 3

Before the beginning of simulation, we need to 2 Q2

define initial conditions for each segment’s flow and 1

Q1
pressure. It is assumed that volumetric flow rate and 0

pressures are in the steady-state regime before the -1

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
t [s]
beginning of the transition process.
Fig. 11 Transient processes for n = 4 and Ls = 4 m.
Figs. 10-12 present relative changes of the flow Q2
8

and the pressures P1 and P2 for step increase of the flow

7

Q1 and for the pipeline of the same length (L = 16 m). 6 P2

Three cases were observed: (a) n = 1, Ls = 16 m; (b) n 5

= 4, Ls = 4 m and (c) n = 16, Ls = 1 m (n × Ls = const.). 4

P1

The results show the influence of the number of 3

segments on dynamics. 2

In all three cases the transition process has the 1

Q1
character of damped oscillations. However, with the 0
Q2

increase of the number of segments the basic -1

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

harmonics of each value (Q2, P1, P2) are superposed t [s]

Fig. 12 Transient processes for n = 8 and Ls = 2 m.

with the harmonics of higher frequency. The larger
number of segments, the higher frequencies in the cases,  segments of same length Ls = 2 m are used for
frequency spectrum. These frequencies may lead to modeling. Difference is only in the number of
undesirable oscillations in the system [9]. segments.
Besides, delay in relation to the excitation value can The inertia of the system is increased with the
be clearly noticed in the response of values at the other increase of the pipeline length. The rise time and
end of the pipeline (Q2, P2). frequency of oscillations decline, whereas the damping
The following Figs. 13-15 show impact of the ratio and delay increase. Particularly noted is reduction
pipeline length on transfer processes in LTL. In all three of magnitude on pressure P2.
 Modeling and Simulation of Hydraulic Long Transmission Line by Bond Graph 261

25 5

20
4

15 P2 P2

10 3

5
P1
Q2 2
P1
0

Q1
1
-5
Q2 Q1
-10
0

-15

-1
-20 0 0.5 1 1 .5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 t [s ]

Fig. 13 Transient processes for n = 1 and Ls = 2 m. Fig. 16 Transient processes for Vg/Vt = 0.001.
10 3.5

8
3 P2
P2
2.5
6

2
4
P1
1.5
Q1 P1
2
1
Q1 Q2
0
0.5
Q2
-2 0

-0.5
-4 0 0.5 1 1.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
t [s] t [s]

Fig. 14 Transient processes for n = 4 and Ls = 2 m. Fig. 17 Transient processes for Vg/Vt = 0.01.
8 1.4

7
1.2 P2
6 P2 Q1
1

5
0.8
4
P1 0.6
P1
3

0.4
Q2
2

0.2
1
Q1
0
0 Q2

-1 -0.2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.5 1 1.5
t [s] t [s]

Fig. 15 Transient processes for n = 8 and Ls = 2 m. Fig. 18 Transient processes for Vg/Vt = 0.1.

The last parameter which influence on the transfer where Bc—bulk modulus of hydraulic pipe, Bl—bulk
process we are analyzing is effective bulk modulus. In modulus of liquid, Vg/Vt—portion of vapor or gas in
previous examples, we assumed that the elasticity of total volume, Bg—bulk modulus of entrapped gas.
hydraulic pipe wall can be neglected. We also assumed If we assume that Bc = 3.45×1010 Pa, Bl =1.4 × 109
that there is no entrapped vapor or gas in the fluid. Pa, Bg =2 × 105 Pa, then transient processes for
Effective bulk modulus can be approximately different gas percentage in the fluid look like in Figs.
determined using the expression [9]: 16-18 (n = 8, Ls = 2 m).
1 1 1 Vg 1 Simulation results show that presence of air in LTL
   (2)
Be Bc Bl Vt Bg can have great influence on dynamics. Even with small
262 Modeling and Simulation of Hydraulic Long Transmission Line by Bond Graph

volume share of air, character of the transfer process Education and Science under project TR 33026.
changes completely. Instead of oscillations, we have an
References
aperiodic transfer process and significantly slower
system response. [1] D. Matko, G. Geiger, W. Gregoritza, Pipeline simulation
techniques, Mathematics and Computers in Simulation 52
4. Conclusions (2000) 211-230.
[2] J.S. Stecki, D.C. Davis, Fluid transmission
Model of long transmission line is derived by using lines—distributed parameter models. Part 1: A review of
the state of the art, in: Proceedings of the Institution of
the method of pipeline division on final length
Mechanical Engineers. Part A: Journal of Power and
segments. Inside each of the segments, spatial change Energy, 1986, Vol. 200, No. 4, pp. 215-228.
is abstracted and only the time changes are observed. [3] J.S. Stecki, D.C. Davis, Fluid transmission
lines—distributed parameter models. Part 2: Comparison
Friction, inertia and compressibility are included in
of models, in: Proceedings of the Institution of Mechanical
the model with use of bond graph elements. Engineers. Part A: Journal of Power and Energy, 1986,
Simulation shows that accuracy of results depends on Vol. 200, No. 4, pp. 229-236.
the number of segments the pipeline is approximated. [4] D.A. Hullender, A.J. Healey, Rational polynomial
approximations for fluid transmission line models, in:
That number depends on the hydraulic circuit where Proceedings of the ASME Winter Annual Meeting—Fluid
the pipeline is located. Practically, the number of Transmission Line Dynamics, November 15-20, 1981,
segments should be chosen depending on the Washington, USA, pp. 33-55.
[5] D.A. Hullender, R.L. Woods, C.-H. Hsu, Time domain
frequency of the input signal and dynamics of other simulation of fluid transmission lines using minimum order
elements in the system. It results in obtaining a state variable models, in: Proceedings of the ASME Winter
mathematical model of finite order which can further Annual Meeting—Fluid Transmission Line Dynamics,
November 13-18, 1983, Boston, USA, pp. 78-97.
be reduced to a lower order and still describe the
[6] U. Söderman, Conceptual modelling of physical
transition process correctly. Besides the pipeline systems—A frame of reference, Ph.D. Thesis, Linköping
length, effective compressibility of fluid also greatly University, Linköping, Sweden, 1995.
[7] J. Watton, Fluid Power Systems: Modeling, Simulation,
affects the dynamics. Presence of entrapped vapor or
Analog and Microcomputer Control, Prentice Hall
gas slows the transfer process and reduces oscillations International, 1989.
in the system. [8] N. Nedić, V. Filipović, Lj. Dubonjić, Design of controllers
with fixed order for hydraulic control system with a long
Acknowledgment transmission line, FME Transactions 38 (2010) 79-86.
[9] H. Merritt, Hydraulic Control Systems, John Wiley &
This work was supported by the Serbian Ministry of Sons, 1967.