You are on page 1of 8

Dynamic Model for a

A. Nabi
e-mail: nebia@rafael.co.il Dome-Loaded Pressure Regulator
A generalized physical model describing dynamic behavior of a fast-acting, dome-loaded,
E. Wacholder gas pressure regulator was developed. The regulator is designed to respond quickly to
e-mail: eitanwa@rafael.co.il
command changes, and to operate over a wide range of flow rates and pressures. The
Rafael–Israel Armament Development Authority,
analytical lumped-parameter model developed consists of a set of nonlinear, first-order,
P.O. Box 2250,
ordinary differential equations with respect to time, accounting for mass and energy
Haifa, 31021, Israel
conservation at regulator outlet, command dome and internal feedback compartments. It
also accounts for the equation-of-motion for the poppet and the control piston-assembly.
The numerical solution, based on a Runge–Kutta method, is amenable to an extensive
parametric study of regulator performance, and serves as a useful analytical tool for
J. Dayan designing new pressure regulators. Several tests were performed on a fast-acting regu-
Technion–Israel Institute of Technology,
lator to verify the physical model. Good agreement between predictions and measure-
Mechanical Engineering Faculty,
ments was obtained. The effect of several parameters, geometrical and operational, on
Haifa, 32000, Israel.
regulator performance was studied. 关S0022-0434共00兲00402-0兴
e-mail: merdayan@tx.technion.ac.il
Keywords: Dome-Loaded Pressure Regulator, Regulator Dynamics, Regulator Perfor-
mance, Control, Simulation, Pneumatic Device

1 Introduction as well as stability criteria in this way, these are of very limited
value, since such models are only suitable for small deviations
Gas dome-loaded pressure regulators 共DLPR兲, having fast re-
around the selected equilibrium point. Thus, caution must be ex-
sponse and large flow rates, are essential equipment in modern
ercised if an attempt is made to extend these results to large
high-energy chemical lasers, as well as in some special industrial
changes in the working conditions.
applications. These regulators have been continuously developed
In this paper, we present a carefully built analytical model de-
and improved over the years 共in a few centers around the world兲,
veloped for a newly designed DLPR. The analytically obtained
and though the principle of operation of today’s equipment is the dynamic model and its simulated response contributed consider-
same as 30 or 40 years ago, the modern regulators are quite dif- ably to the understanding of the various effects influencing actual
ferent in appearance, accuracy and in their fast dynamics com- regulator behavior. Thereafter, for the remainder of the regulator
pared to the ‘‘old’’ regulators. Nevertheless, dynamic analysis of development, computer runs of the model were substituted for a
such regulators has not been reported for almost three decades. In large portion of the previously required physical tests, which
fact, only a few papers dealing with the dynamic and static be- saved a great deal of time and expensive resources. Moreover, the
havior of the more familiar spring-loaded pressure regulator model is now available for future designs.
共SLPR兲 have been published. Tsai and Cassidy 关1兴 introduced,
probably for the first time, a systematic analysis, linear as well as 2 Regulator Description and Operation
nonlinear, of a simple SLPR. For simplicity, they assumed a linear
viscous friction force on the command piston and on the poppet- The DLPR discussed in this paper is schematically shown in
stem seals. Dustin 关2兴, presented a nonlinear investigation of a Fig. 1. It was developed for applications requiring very fast dy-
more complicated SLPR, with more realistic nonlinear constant namic response, i.e., capability of reaching the desired value of
dry friction force, but restricted the model to choked flow through the outlet pressure within a fraction of a second from its opening,
the valve, and neglected gas temperature dynamics. Anisimikin while exhibiting stable and damped behavior.
et al. 关3兴, investigated the stability of a regulator with a linearized The DLPR operation is based on force balance. The desired
model including nonlinear friction force. Sverbilov and Ani- outlet pressure is dictated by applying a reference pressure to the
simikin 关4兴 presented a linear model for the study of a very simple upper side of the command piston. This pressure causes a dis-
SLPR, operating under an arbitrary load. placement of the poppet 共opening or closing the outgoing passage兲
It should be pointed out that, for many years, the trial-and-error and a respective change in the regulator outlet pressure. The regu-
lator has a built-in feedback control loop, since the outlet pressure
method of developing the various kind of pressure regulators was
acts on the lower side of the command piston and the poppet
the only accepted way 共see, for example, Dustin 关2兴; Tsai and
through one or more pressure-equalization holes. At steady state,
Cassidy 关1兴兲. Most probable, this was the result of the inherent
this pressure applies a balancing force on both command piston
difficulties in the modeling of these regulators, due to the nonlin-
and poppet. Regulators having this arrangement are said to have a
earities and coupling among the equations describing the dynam-
‘‘balanced poppet.’’ This special feature reduces, but does not
ics of both the fluids and the mechanical parts. The main source of
cancel, the effects of changes in inlet 共supply兲 pressure on the
nonlinearities are the gas processes, the dry Coulomb friction act-
regulated outlet pressure.
ing on the poppet-command piston assembly, and the dependency
Initiating a command to the regulator is governed by a three-
of the free cross-flow area on the poppet location. These difficul-
way solenoid valve 共Fig. 1兲. Initially, the command volume is
ties, together with the slow running nonlinear programs, led to
open to atmosphere and the regulator is closed. When the solenoid
various attempts to linearize the equations around the operating
valve is activated, it closes the opening to atmosphere, and con-
point. Though it is possible to obtain various analytical solutions
nects the upper side of the command piston to the desired refer-
ence pressure from a pressure accumulator. A needle valve con-
Contributed by the Dynamic Systems and Control Division for publication in the
JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript
trols the flow of the command gas from the accumulator to the
received by the Dynamic Systems and Control Division December 17, 1997. Asso- upper side of the command piston. This restriction 共later referred
ciate Technical Editor: Woong-Chul Yang. to as ‘‘command orifice’’兲 affects the rate of regulator opening

290 Õ Vol. 122, JUNE 2000 Copyright © 2000 by ASME Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


5 According to Potter and Levi 关6兴 the time required to obtain
significant pressure changes in the system is much greater than the
time for wave propagation indicated by Reynolds and Kays 关7兴.
Thus, the gas dynamics can be considered as continuum of quasi
steady states.
6 The compression and expansion processes are adiabatic. This
assumption is reasonable, as long as the gas dynamic processes
within the regulator are faster than the heat-transfer processes at
and through the walls.
7 The gas behaves ideally. This is probably a reasonable as-
sumption for the entire operation range of the regulator when
either air or nitrogen at moderate pressures is used. A possible
deviation from ideal-gas behavior may occur for very high
pressures.
8 The specific heats of the gases are constants, a reasonable
assumption for the considered gases at temperatures in the neigh-
borhood of standard atmosphere.
9 The volume at the regulator exit is lumped. Thus, dynamic
effects and possible pressure distribution in a long pipe connected
to the exit are neglected in the parametric study. Nevertheless,
adding a simple pipe model was also attempted in order to facili-
tate comparison between the measured and the calculated outlet
pressure time-response. 共Measurements were taken at a distance
of 2 m from the regulator exit.兲
The Governing Equations. The equation-of-motion repre-
sents the balance of the inertia, friction, pressure, and spring
forces acting on the poppet and the command piston. The sche-
matic force diagram and main nomenclature are presented in Fig.
Fig. 1 Schematics of gas DLPR and experimental arrange- 2. Positive forces are assumed to act in the upward direction,
ments
while, for the mechanical parts, downward motion, i.e., opening
the valve, is considered positive. The equation is
M ẍ⫹K s x⫹F s ⫹ f •sgn共 ẋ 兲 ⫹ 共 A s1 ⫺ ␣ A s2 兲 P s ⫹A ␯ 1 P 1 ⫹A 3 P 3
and its dynamic stability, the entering command-pressure forces,
the command-piston and poppet movements, and permits the flow ⫹ 共 A ␯ o ⫺ ␤ A o2 兲 P o ⫺A ␯ P ␯ ⫽0 (1)
of gas through the regulator. As the outlet pressure rises, both
pressures under the command piston and under the poppet rise where, ␣ and ␤ are empirical correction factors for the force act-
with it. The lag between these two pressures and the outlet pres- ing symmetrically on the poppet surface, according to assump-
sure is dictated by the size of the pressure equalization holes, and tions 2 and 3. A s2 and A o2 are the conical surface areas of the
this, in turn, affects the characteristics of the response. When poppet related to the valve opening area A p ,
equalization of forces, above and below the poppet and piston, is
reached, the regulator is operating at steady state. A s2 ⫽A s2,o ⫹A p cos ␪ (2)

A o2 ⫽A o2,o ⫺A p cos ␪ (3)


3 Analysis and Modeling of the Regulator Dynamics A p ⫽ ␲ x sin ␪ 共 d⫺x sin ␪ cos ␪ 兲 (4)
The presented analysis is based on the following assumptions.
A o2,o and A s2,o are the exit side and supply side conical surface
1 Generally, the poppet and command piston are moving as a areas of the poppet, for a closed regulator, and d is the diameter of
single unit. Separate movements of these parts are assumed to the poppet seat. 共See Figs. 1 and 2 for nomenclature and
occur only during extreme imbalance between the feedback pres- definitions.兲
sure sensed under the poppet and the pressure above the poppet. The dynamic equations for pressure and temperature at the
Assuming a single unified movement for the poppet-command regulator exit and in its internal compartments, based on the
piston assembly greatly simplifies the dynamic equations. above-stated assumptions, were derived from the mass, momen-
2 In computing the force balance on the poppet, it is assumed tum and energy balances. The analysis is applied to a general case
that a sharp boundary, perpendicular to the conic edge and located of variable control volume 共moving piston兲 with single inlet and
at the narrowest cross-section of the flow, divides the free space single outlet flows, assuming ideal gas behavior and constant spe-
around the poppet. In the space directly below this boundary, the cific heats 关8兴:
pressure is ‘‘close’’ to the inlet pressure, while above this bound-
ary, the pressure is close to the outlet pressure. The exact prox- d P ␥R ␥A P
⫽ 共 T inṁ in⫺Tṁ out兲 ⫺ ẋ (5)
imity of these pressures to the inlet and outlet pressures, respec- dt V V

冋冉 冊 册
tively, is determined experimentally. This assumption is based on
a somewhat simpler approach stated by Anderson 关5兴. dT RT 2 T in 共 ␥ ⫺1 兲 AT
3 The forces acting on the poppet are symmetric. In reality,
⫽ ␥ ⫺1 ṁ in⫺ 共 ␥ ⫺1 兲 ṁ out ⫺ ẋ (6)
dt VP T V
possible asymmetric forces and moments can be formed around
the poppet by the flow through the valve due to the single inlet where, V⫽V(0)⫹Ax.
and single outlet configuration. Neglecting these effects signifi- For the pressure and temperature at the lumped-exit volume, the
cantly simplifies the force-balance analysis. equations are
4 The seals of the command piston and the poppet produce a
d Po ␥R
constant Coulomb friction force, acting in a direction opposite to ⫽ 共 T ṁ ⫺T o ṁ o 兲 (7)
the poppet-command piston assembly movement. dt Vo s s

Journal of Dynamic Systems, Measurement, and Control JUNE 2000, Vol. 122 Õ 291

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


d P1 ␥R ␥ A ␯1
⫽sgn共 P o ⫺ P 1 兲 T ṁ ⫹ P ẋ (12)
dt V1 o 1 V1 1
V 1 ⫽V 1,o ⫺A ␯ 1 x (13)
With the above-mentioned assumptions, and by neglecting the
small volume variations due to poppet motion, the pressure dy-
namics at the feedback compartment under the poppet is given by
d P3 ␥R
⫽sgn共 P o ⫺ P 3 兲 T ṁ (14)
dt V3 o 3
The DLPR dynamics comprises a set of seven coupled, nonlin-
ear, ordinary differential equations with respect to time, namely:
共1兲, 共7兲–共10兲, 共12兲, and 共14兲. In order to solve this set of equa-
tions, the flow to and from the various volumes must be specified.
For a general case of choked and nonchoked compressible flow
through a restriction, the mass flow rate is given by 共Andersen 关5兴,
p. 20兲
ṁ⫽C d A f 共 ␥ 兲 N in,outP in/ 冑RT in (15)
where C d is the geometry dependent discharge coefficient of the
restriction,
f 共 ␥ 兲 ⫽ 冑␥ 关 2/共 ␥ ⫹1 兲兴 共 ␥ ⫹1 兲 / 关 2 共 ␥ ⫺1 兲兴 (16)
N in,out accounts for the ratio of inlet to outlet pressure conditions
共‘‘in’’ and ‘‘out,’’ respectively兲. For choked flow N in,out⫽1; else,
it is less than one, as calculated by

再 共 P out / P in兲 2/␥ ⫺ 共 P out / P in兲 共 ␥ ⫹1 兲 / ␥


1/2

冋 册
N in,out⫽N 共 P out / P in兲 ⫽ 共 ␥ ⫹1 兲 / 共 ␥ ⫺1 兲
共 ␥ ⫺1 兲 2
2 共 ␥ ⫹1 兲
(17)
Fig. 2 The schematic force diagram and main nomenclature
for the DLPR, „a… Forces acting on the poppet and on the con-
Applying Eqs. 共15兲–共17兲 to the flow through the passage con-
trol piston, „b… Poppet’s schematic diagram. trolled by the poppet movement yields:
ṁ s ⫽C ds A p f 共 ␥ 兲 N s,o P s / 冑RT s (18)

冋冉 冊 册
where N s,o ⫽N( P o / P s ) is given by Eq. 共17兲, and variations in A p
dT o RT 2o Ts
⫽ ␥ ⫺1 ṁ s ⫺ 共 ␥ ⫺1 兲 ṁ o (8) are controlled by the poppet displacement, according to Eq. 共4兲.
dt VoPo To The flow through the exit nozzle is
For the command-dome volume, the pressure and temperature ṁ o ⫽C doA noz f 共 ␥ 兲 N o,envP o / 冑RT o (19)
are given by
where N o,env⫽N( P env / P o ) is, again, determined by Eq. 共17兲.
d P␯ ␥R ␥A␯ 共P env is the environment pressure at the nozzle exit.兲
⫽sgn共 P ␯ o ⫺ P ␯ 兲 T ⬘␯ ṁ ␯ ⫺ P ẋ (9)
dt V␯ V␯ ␯ The flow through the restriction 共needle valve兲 into the com-

冉 冊
mand dome is
dT ␯ RT 2␯ T ␯⬘ 共 ␥ ⫺1 兲 A ␯
dt
⫽sgn共 P ␯ o ⫺ P ␯ 兲
V␯P␯
␥ ⫺1 ṁ ␯ ⫺
T␯ V␯
T ␯ ẋ ṁ ␯ ⫽C d ␯ A c f 共 ␥ 兲 N ␯ P ␯⬘ / 冑RT ␯⬘ (20)
(10) where
where subscript ␯ o denotes conditions upstream of the command
orifice and subscript ␯ conditions at the command dome. N ␯ ⫽N ␯ , ␯ o ⫽N 共 P ␯ / P ␯ o 兲 , P ␯⬘ ⫽ P ␯ o ,
It should be noted that fluid can enter and exit through a single and
port only. The direction of flow through this port affects the tem-
perature and pressure of the fluid in the dome. Thus, T ␯⬘ ⫽T ␯ o for T ␯⬘ ⫽T ␯ o for flow into V ␯
positive sgn(P␯o⫺P␯), which applies to the case when fluid enters while N ␯ ⫽N ␯ o, ␯ ⫽N( P ␯ o / P ␯ ), P ⬘␯ ⫽ P ␯ , and T ␯⬘ ⫽T ⬘␯ for flow out
the dome, while T ␯⬘ ⫽T ␯ for negative sgn(P␯o⫺P␯) applies to fluid of V ␯ .
leaving the dome.
It should also be noted that the volume in the command dome is Because of the wide range of flow regimes into the command
changing due to the command piston movement, and is given by volume 共from choked flow to zero兲, C d ␯ varies, but can be ap-
proximated by the linear correlation 共based on 关5兴兲
V ␯ ⫽V ␯ ,o ⫹A ␯ x (11)
C d ␯ ⬇1.0⫺0.7• 共 P ␯ / P ␯ o 兲 (21)
The pressure dynamics at the feedback compartment under the
command piston is evaluated in a similar manner to that of the Similarly, the flows into the two feedback compartments are
ṁ 1 ⫽C d1 a 1 f 共 ␥ 兲 N 1 P 1⬘ / 冑RT o
exit volume. Neglecting temperature variations by assuming T 1
(22)
⬇T o , and recalling that the motion considered here is in the op-
posite direction with respect to the coordinate x 共which causes a ṁ 3 ⫽C d3 a 3 f 共 ␥ 兲 N 3 P 3⬘ / 冑RT o (23)
change in sign for the displacement and the velocity terms兲, the
equations are where, for entering flow

292 Õ Vol. 122, JUNE 2000 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Table 1 Construction data of the DLPR

Symbol Value Units


Poppet angle ␪ 45 deg Degree
Main valve orifice diameter d 67.5 mm
Range of effective diameter of dc 0.8–8 mm
the restriction to dome
Pressure-equalization holes d 3 ⫽d 1 4 mm
共4 in command piston, 1 in poppet兲
Area of upper side of command piston A␯ 1.08⫻10⫺2 m2
Area of lower side of poppet A3 3.47⫻10⫺3 m2
Outlet nozzle area A noz 7.04⫻10⫺4 m2
Volume of outlet tube Vo 3.5⫻10⫺3 m3
Initial command volume V ␯ ,o 4.0⫻10⫺5 m3
Piston and poppet mass M 1.46 kg
Spring constant Ks 2.68⫻105 N/m
Spring preset force Fs 3.80⫻103 N
Friction force 共varied in the f 0–500 N
parametric study兲

Fig. 3 Simulation of regulator reaction to a 60.5 bar step results was studied. A representative constant correction factor of
change in command pressure „inlet pressure 122 bar…. Run ␣ ⫽0.8 and ␤ ⫽0.4 共affecting Eq. 共1兲兲 was used throughout the
conditions: ␣ Ä0.8, ␤ Ä0.4, f Ä100 N, d c Ä1.06 mm, d 1 Ä d 3 calculations.
Ä4 mm, F s Ä3800 N.
It can be seen that the regulator exhibited stable and generally
damped response. It achieved steady state after approximately
0.37 s, and the outlet pressure reached 46 bar. The feedback pres-
N 1 ⫽N o,1⫽N 共 P 1 / P o 兲 , P 1⬘ ⫽ P o , sure, P 1 , was almost identical to the outlet pressure, P o , through-
out the valve opening process, due to four pressure-equalizing
N 3 ⫽N o,3⫽N 共 P 3 / P o 兲 , holes 共4 mm in diameter兲 through the command piston. On the
other hand, there is only a single hole, of the same diameter, in the
and poppet, and this caused a delay in the response of pressure P 3 .
P 3⬘ ⫽ P o Except for the small overshoot during the initial stages of regu-
lator opening, the displacement of the poppet was overdamped. A
while, for exiting flow 0.02 s delay time between poppet movement and command exists
N 1 ⫽N 1,o ⫽N 共 P o / P 1 兲 , P 1⬘ ⫽ P 1 , due to the build-up of the required minimum pressure 共7.7 bar兲
within the dome necessary to overcome the preset force of the
N 3 ⫽N 3,o ⫽N 共 P o / P 3 兲 , spring. The temperature within the command dome rose, during
the first 0.02 s 共coinciding with the poppet delay time兲, to a maxi-
and mal value of approximately 380 K, due to the adiabatic compres-
P 3⬘ ⫽ P 3 . sion process. This was followed by a short undershoot due to an
abrupt increase in the command volume as the poppet began to
4 Simulation of Regulator Dynamics move, and then remained almost constant. However, the tempera-
ture at the regulator exit rose to 360 K, but then returned slowly to
The dynamic response of the DLPR is simulated by solving the its initial value of 293 K.
model equations, for a given set of initial conditions and a forcing
function 共the command pressure at the dome as function of time兲.
The numerical integration was carried out by a fourth-order 5 Experimental Testing of the DLPR
Runge–Kutta algorithm 关9兴. Several tests were conducted on a DLPR designed and devel-
The physical constraints limiting the movement of the poppet oped at Rafael for high gas-flow rates 共up to 10 kg/s兲 and quick
were also included in the program. These constraints zero the opening. Its geometrical and operational data 共corresponding to
value of the coordinate x in case it becomes negative during the Figs. 1 and 2 nomenclature兲 are given in Table 1. The general
calculations, since the poppet obviously cannot pass its seats. In scheme of the experimental set-up is also shown in Fig. 1. The
addition, the velocity ẋ of the poppet is set to zero when x⫽0. experiments were conducted with air source 共pressurized cylin-
Physically, this means that the poppet has been completely ders兲 of 200 bar maximal pressure. In the experimental set-up, the
stopped by its seat 共plastic collision兲 and the regulator is closed. regulator exit arrangement consisted of two parallel tubes, each of
Similarly, another constraint limits the maximal opening of the 30 mm diameter and 2 m long, discharging through a common 30
valve, and x is set to x max while the poppet velocity is, again, set mm diameter orifice.
to zero (ẋ⫽0) when this limit is exceeded. As described in Section 2, command pressurized air is obtained
The simulated model solutions for the regulator operation pre- from a pressure accumulator through the adjustable needle valve
sented here, always start with a closed valve (x⫽ẋ⫽0) and with 共Fig. 1兲. 共The effective needle valve adjustment is determined by
atmospheric pressure at both regulator command-dome and exit. separate tests of filling a known volume with air flowing in
The simulation of the regulator response to a step change in choked-flow regime.兲 In all of the experimental runs, a step input
command pressure is presented by the curves in Fig. 3. The simu- was applied to the regulator command-dome by means of the
lation was conducted for the regulator whose geometrical mea- three-way solenoid valve.
sures and other characteristic data are given in Table 1, and for Initially, this solenoid valve was closed allowing the regulator
command pressure of 60.5 bar and inlet gas pressure of 122 bar. command-dome to be ventilated to the atmosphere. Then the so-
The effective needle-valve flow diameter 共command orifice diam- lenoid valve was opened, shutting off the atmospheric vent of the
eter兲 at the inlet to the command dome was 1.06 mm. An approxi- dome, and allowing the inside pressure to build up to the accumu-
mate friction force of 100N has been included in the model. This lator pressure and to operate the regulator. Two to three seconds
parameter is not known exactly, and its influence on the calculated later, after steady-state conditions were definitely achieved, the

Journal of Dynamic Systems, Measurement, and Control JUNE 2000, Vol. 122 Õ 293

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Table 2 DLPR test data for initially closed valve. P ␯ o —dome
final pressure, P s —supply upstream pressure, P o —outlet pres-
sure at the regulator, P o ,noz—pressure at exit nozzle,
d c —command orifice diameter, t d —dead time, t f —settling time.

Run P ␯o Ps dc Po P o,noz tf td
# bar bar mm bar bar s s
1 61.5 125 1.06 — 40.5 0.44 0.02
2 60.5 122 1.06 — 40.2 0.46 0.02
3 60.5 122 1.21 — 41.2 0.35 0.02
4 60.7 118 1.02 — 41.5 0.44 0.02
5 60.7 116 0.94 — 42.0 0.63 0.02
6 20.0 114 1.06 — 9.6 0.34 0.06
7 6.0 113 1.06 — 1.0 * *
8 10.0 113 1.06 — 3.6 0.17 0.11
9 83.0 112 1.06 — 66.0 0.8 0.016
10 85.0 198 1.06 61.5 55.0 0.40 0.025
11 82.5 165 1.06 66.7 57.0 0.46 0.025

*Valve did not open.


Fig. 4 Empirical correlation for the correction factors ␣ and ␤
as function of outlet to inlet pressure ratio assuming ␣ Ä ␤
experiment was shut off by closing the solenoid valve. Eleven
experiments were made, for which the data are summarized in
Table 2.
During the first nine runs, the pressures were measured at four
different locations: P ␯ 0 in the accumulator, P ␯ at the command
dome, P s at the inlet to the regulator, and P o,noz at the exit orifice
共the far end of the 2 m long pipe, just before the opening to
atmosphere兲. For comparison with the experimental results of the
first nine runs a simple model for the pipe, which takes into ac-
count the effect of the pipe on the pressure drop and time response
关8兴, was added to the regulator model. Fig. 5 Regulator response in run no. 2 for command pressure
of 60.5 bar and needle valve openings of 1.06 mm. Run 2 con-
For the last two experiments, the pressure, P o , at the immediate ditions: P ␯ o Ä60.5 bar, P s Ä122 bar, d c Ä1.06 mm.
vicinity of the regulator outlet 共at the beginning of the 2 m long
pipe兲 was also measured.
The first five test runs were conducted with almost identical
command pressure but the needle valve opening was changed to
study its effect on the regulator response. The rise time became
shorter as the command orifice diameter was increased. It varied
from 0.63 s for a 0.94 mm orifice diameter in run 5 down to 0.35
s for a 1.21 mm orifice diameter in run 3.
Runs 6 to 9 were conducted with fixed valve opening but with
different command pressures, to study the regulator response at
different operating points. As the command pressure was lowered, Fig. 6 Regulator dynamics at low command pressures of 20
the rise time became shorter, going down to 0.17 s for a command bar for run no. 6. Run 6 conditions: P ␯ o Ä20 bar, P s Ä114 bar,
pressure of 10 bar in run 8, but at the same time, the pressure d c Ä1.06 mm.
response delay time was increased up to 0.11 s for the same run.
Note that in run 7, with a 6 bar command pressure, the valve did
not open because it did not reach the required minimum pressure
to overcome the spring preload 共see Section 4 and Table 1兲. sure inside the dome. However, in most of the experimental runs,
Runs 10 and 11 had command orifice diameter and command especially those with high outlet pressures, there are some ‘‘struc-
pressure similar to those in run 9, but the inlet supply pressures tural’’ discrepancies between the predicted and the measured tran-
were higher 共198 bar and 165 bar, respectively兲. As a result, sient response of the outlet pressure. Right after the beginning of
shorter rise times 共0.4 s and 0.46 s兲 were obtained, compared to a the run, the actual pressure rises slower and in a more gradual
rise time of 0.8 s in run 9 for inlet pressure of 112 bar. way, exhibiting ‘‘S-shape’’ behavior, compared to the predicted
The values of the steady-state outlet pressure were used to de- exponential-shape 共similar to first-order lag response兲 pressure
rive an empirical estimate for the correction factors ␣ and ␤ of the rise. Substantial effort was invested to identify the source of this
force acting on the poppet’s upper face, in Eq. 共1兲. Figure 4 pre- discrepancy. Some of these efforts, such as adding the above men-
sents a semilog plot of ␣ and ␤ as functions of outlet to inlet tioned pipe model, assuming isothermal operation, and studying
pressure ratio. These values were derived with the rather artificial the effect of very high friction, did not provide an acceptable
assumption of ␣ ⫽ ␤ . Note that these correction factors are not explanation. In all these attempts, the calculated pressure response
constant. They approach 1.0 at the two limits of pressure ratios 0 has an exponential-shape similar to the one in Fig. 5. It is pos-
and 1.0, and have a minimum at pressure ratio of about 0.2. sible, however, that the pressure distribution 共namely, ␣ and ␤兲,
The experimental results for three of the test runs were plotted hence, the force on the upper side of the poppet, is dynamically
共dotted curves兲, along with the theoretical calculated predictions changing as a function of the ratio between the inlet and outlet
from the model 共continuous curves兲. Figure 5 describes regulator pressures, similar to Fig. 4 indication 共see further discussion in
response in run 2 representing the general behavior for all the first Section 6兲.
five runs. Only rise time, which was affected by the different Figure 6 describes the regulator dynamics at low command
needle valve opening, was different as discussed earlier. The pressures for run 6. 共Run 8 yielded similar behavior.兲 Again, the
model reconstructs the initial delay time behavior quite accu- model reconstructs quite accurately the initial behavior and the
rately. It also correctly predicts the transient of the command pres- long delay time.

294 Õ Vol. 122, JUNE 2000 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


for the DLPR under study, that enlarging the diameter d 3 共even up
to 12 mm兲 in order to shorten the delay of P 3 , improved the
overall DLPR response but with only minor effect on the
damping.
The preload force applied to the poppet by the spring, and the
spring constant selection are determined at an early design stage,
mainly according to considerations dealing with preventing leak-
age between the poppet and its seals. The effect of initial spring
load and spring constant on the regulator dynamics was studied by
Fig. 7 Dynamic response of the regulator for a d c Ä2 mm reducing the preload force by a factor of 10 with respect to the
command-orifice diameter actual load used in the experiments. The overshoot in poppet
movement increases substantially, and the forces acting on the
low stiffness spring may cause it to fail 共as actually did happen
6 Parametric Study of the Nonlinear Dynamic Model during experiments兲.
Increasing command pressure increases overshoot of the poppet
The general agreement between the model predictions and the movement and lowers its damping. At the same time, the rise time
experimental results indicates that the model is reasonably correct,
is also increased 共a trend supported by the experimental findings
and is capable of predicting the general dynamics of the regulator.
of Table 2兲. However, the dead time delay is shortened from 0.025
Thus, the effects of the physical and geometrical parameters,
s at 20 bar to 0.005 s at 90 bar command step input.
f ,d c ,d 3 ,d 1 ,F s 共pre-applied spring force兲, P ␯ o , and P s , were
Studying the effect of the inlet supply pressure shows that the
studied by simulation.
outlet pressure overshoot increases with increase in the supply
The effect of friction on the DLPR time response was studied
by varying the value of f in the model between 0 to 500 N. The pressure, but the rise time decreases. For the same command, the
dome pressure and the outlet pressure showed damped response final steady state regulated pressure increases if the supply pres-
during the entire opening process 共similar to that in Fig. 3兲 even sure is reduced—a phenomenon that was more pronounced in the
for zero friction. Only the simulation of the poppet displacement experiments than in the predictions
showed some small-amplitude high-frequency vibrations 共⬃310 Finally, a cross comparison summary of the various parameter
Hz兲 at the beginning of the process. The minor role of the friction effects on regulator dynamics is presented by the curves in Fig. 9.
in damping the pressure response is probably an indication that All these curves are given as a function of the restriction size at
most of the DLPR damping is related to the pneumatic process. the inlet of the command dome inlet, d c . Figure 9共a兲 summarizes
Figure 7 shows the regulator dynamic response for a 2 mm the dependence of the poppet movement overshoot at the begin-
command orifice diameter, which is larger than the 1.06 mm ori- ning of regulator opening. This dependence is given for several
fice of the ‘‘nominal case’’ in Fig. 3. The forcing function 共60.5 cases of initial spring loading and different pressure-equalizing
bar step input to the command dome兲 and other DLPR parameters holes. The command pressure in all these cases was a 60.5 bar
are the same as for the nominal case. The regulator equipped with step input. Figure 9共b兲 presents the overshoot of the controlled
the larger opening shows a much faster response compared to the outlet pressure P o as a function of the restriction diameter d c , for
regulator response with nominal opening,—settling within 0.15 s various sizes of the pressure-equalizing holes. Effect of supply
versus 0.37 s. The command dome pressure and the outlet pres- pressure is also shown for one size of pressure-equalizing holes.
sure show small overshoots. The dome pressure oscillates slightly Figure 9共c兲 depicts the settling as function of d c . For d c
during the transient, while the poppet responds with an initial ⭐3.5 mm, the opening time is inversely proportional to the re-
sharp overshoot—about six times larger than in Fig. 3—changing striction diameter. This inverse proportionality is steeper for d c
to high frequency 共⬃700 Hz兲 with high but decaying amplitude. ⭐1.5 mm, while for d c ⬎1.5 mm the dependence is more gradual,
Near the steady state, the poppet movement turns into a lower but overshoot appears at the regulated pressure. For d c ⬎4 mm,
frequency 共⬃60 Hz兲 and lower amplitude decaying movement. the regulated pressure starts developing oscillations that increase
Further simulations for larger orifices indicate that overshoots of the time to reach steady state. These phenomena are related to the
the outlet pressure, poppet movement and the oscillations ampli- time lag in the dynamics of P 3 , the pressure under the poppet.
tude are all intensified. The shape discrepancy between the measured and predicted
The effect of reducing the size of the equalizing holes is de- outlet pressure response led to an intense study of the validity of
picted in Fig. 8, for d 1 ⫽d 3 ⫽2 mm 共physically, d 1 ⫽d 3 ⫽4 mm兲. the model assumptions. Four different modifications have been
This causes increased overshoot in the outlet pressure response, tried:
and increased settling time. Nevertheless, smaller equalizing ori-
共a兲 Expanding the model by adding a simple pipe dynamics of
fices cause increased damping of the poppet movement. It should
several stirred tanks in series 关8兴 instead of assumption no. 9.
be noted that the feedback pressure P 1 under the command piston
共b兲 Assuming an isothermal, rather than adiabatic dynamic pro-
reaches P o faster than the pressure under the poppet P 3 , due to
cess 共assumption no. 6兲.
the four pressure equalizing holes under the piston compared to
共c兲 Assuming very fast dynamics in the feedback volumes V 1
only one hole through the poppet. However, P 3 is the pressure
and V 3 , with respect to V ␯ and V o . This simplification is justified
dictating the rate at which P o reaches steady state. It was found,
for very small feedback volumes and very large pressure-
equalizing holes, causing P 1 (t)⬇ P 3 (t)⬇ P o (t), but it requires a
minimal amount of friction ( f ⬎170 N) to stay stable.
None of the above three modifications provided the expected
S-shape response during the middle section of the transient. They
all stayed rather exponential in shape and some even deviated
from the correctly predicted initial and steady-state responses of
the original model.
共d兲 Finally, the model had been tried with varying correction
factors 共reinspecting assumption no. 2兲. Values for ␣ and ␤ were
obtained from Fig. 4 for P o (t)/ P s (t) during the calculations of
Fig. 8 The effect of changing the size of the equalizing holes the dynamic response 共note that ␣ and ␤ were derived from ex-
„ d 1 Ä d 3 Ä2 mm… perimental measurements at steady-state conditions兲. Figure 10

Journal of Dynamic Systems, Measurement, and Control JUNE 2000, Vol. 122 Õ 295

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 10 Effect of variable correction factors ␣ and ␤ „Fig. 4… on
the model predictions for conditions in run 2

depicts the results of such calculations for run No. 2. Obviously,


the predictions are much closer to the entire measured dynamical
and steady-state data than those of Fig. 5.
The new assumption, namely, that the pressure distribution over
the upper face of the poppet is changing dynamically, can be
explained by an analogy to an over-expanded variable-area
choked-nozzle. The pressure distribution and the pattern of pres-
sure recovery shock waves downstream of the nozzle throat shifts
location as a function of pressure ratio in quite a complicated
manner. This explanation is even more applicable to the annulus
shaped ‘‘nozzle’’ created between the poppet and the seat. Thus, a
time dependent force, which is changing along with the transient,
is acting on the poppet surface and is different 共lower兲 than the
force predicted by assumption no. 2.

7 Summary and Conclusions


Theoretical analysis and experimental testing of a dome loaded
gas pressure regulator were presented. This is a high-flow fast-
responding regulator, for which design and development are usu-
ally costly and time consuming. A nonlinear dynamic model de-
scribing the response of the major regulator variables was
suggested, and its predictions were experimentally validated.
The following conclusions and observations have resulted:
1 The dynamic response of the regulator is nonlinearly depen-
dent on the magnitude of the command-dome pressure 共the DLPR
set point兲. For increasing command pressures, the system damping
decreases and the poppet displacement displays greater overshoot.
On the other hand, opening time is shortened as command-
pressure magnitude decreases.
2 The size of the inlet command orifice, d c , is a very important
factor in determining the rate at which the regulator opens, the
magnitude of the poppet-movement overshoot and the regulated
outlet pressure. For the investigated regulator, only a narrow
range of needle-valve opening (d o ⫽0.95⫺1.8 mm) produces fast
共0.15–0.6 s兲, damped opening with tolerable overshoot in the
regulated pressure and the poppet movement.
3 The pressure equalization holes in the poppet and command
piston affect poppet damping and overshoot in the regulated outlet
pressure. Too small holes may cause overshoot in the outlet pres-
sure because of delay in the feedback. Too large holes reduce the
outlet pressure overshoot, but also reduce the amount of damping
of the poppet. The model predicts improved response for this
regulator if d 3 is increased to 6–10 mm.
4 The preload force applied to the poppet by the spring mainly
affects the overshoot of the poppet opening at the beginning of the
Fig. 9 Cross comparison of the various parameter effects on regulator response. Large spring force reduces the overshoot.
regulator dynamics as a function of the restriction opening, d c .
„a… Dependence of overshoot in poppet movement on d c for
Thus, it is recommended to select the spring and its initial com-
different initial spring loading and pressure equalizing holes. pression not only by static consideration 共prevention of gas leaks
„b… Overshoot in outlet pressure, P o , as function of d c for vari- while the regulator is closed兲, but also by dynamic considerations.
ous sizes of the pressure equalizing holes and for different 5 The regulator response is almost unaffected by the ‘‘inter-
supply pressures. „c… Settling time as a function of d c . nal’’ seals friction force in the 0–500 N range.

296 Õ Vol. 122, JUNE 2000 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


6 High supply pressure causes large overshoot in the regulated References
outlet pressure, but reduces the regulator opening time 共settling 关1兴 Tsai, D. H., and Cassidy, E. C., 1961, ‘‘Dynamic Behavior of Simple Pneu-
time兲. If, during the regulator operation, the inlet stream 共supply兲 matic Pressure Reducer,’’ ASME J. Basic Eng., 83, p. 253.
pressure drops 共e.g., because of source depletion兲, the final outlet 关2兴 Dustin, M. O., 1971, ‘‘Analog Computer Study of Design Parameter Effects
pressure rises slowly 共creeps upward兲. The actually detected pres- on the Stability of a Direct-Acting Gas Pressure Regulator,’’ NASA TN
D-6267.
sure rise was even higher than that predicted. 关3兴 Anisimikin, Yu. S., Kvasov, V. M., Kravchenko, Yu., Suslin, E. I., Shorin, V.
7 The experimental correction factors ␣ and ␤ for the force P., and Chibizov, V. V., 1976, ‘‘Stability Analysis of Gas Pressure Regulator
acting on the poppet are not constants, but vary with the ratio of Operating Under High-Temperature Conditions Over a Wide Flow Rate
outlet to inlet pressures. This is explained by the complicated Range,’’ Sov. Aeronaut., 19, No. 1, p. 94.
关4兴 Sverbilov, Y. Ya, and Anisimikin, Yu. S., 1981, ‘‘Evaluation of Multi-Purpose
dynamics of the flow expansion process through the DLPR annu- Gas Pressure Regulator Stability,’’ Sov. Aeronaut. 24, No. 3, p. 62.
lus nozzle. Using variable coefficients improves the model predic- 关5兴 Andersen, B. W., 1967, The Analysis and Design of Pneumatic Systems,
tion of the transient pressure behavior. Wiley, New York.
关6兴 Potter, J. H., and Levy, M. J., 1961, ‘‘The Free Expansion of Dry and Moist
It should be mentioned that in the course of this study, a lin- Air,’’ ASME J. Eng. Ind., 83, No. 1, Feb., p. 97.
earized model was also derived and verified by experiments. Ob- 关7兴 Reynolds, W. C., and Kays, W. M., 1958, ‘‘Blow-Down and Charging Process
in a Single Gas Receiver With Heat Transfer,’’ Trans. ASME, 80, p. 1160.
viously, the results are valid only for small variations in continu- 关8兴 Nabi, A., 1986, ‘‘The Dynamics of Fast Responding Dome-Loaded Pressure
ous steady-state operation. This additional work will be described Regulator,’’ M.Sc. thesis, Technion, Haifa.
in a paper under preparation. 关9兴 White, F. M., 1974, Viscous Fluid Flow, McGraw-Hill, New York.

Journal of Dynamic Systems, Measurement, and Control JUNE 2000, Vol. 122 Õ 297

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like