Professional Documents
Culture Documents
Wiggert
Department of Civil and Sanitary Engineering,
The Effect of Elbow Restraint on
Michigan State University,
East Lansing, Mich. 48824 Pressure Transients
Transient pressure in piped liquid is a function of structural restraint at elbows. If
R. S. Otwell supported rigidly, an elbow causes no appreciable alteration of the pressure
McNamee, Porter and Seeley, transient generated by rapid valve closure. However, if the support is relaxed,
Ann Arbor, Mich. significant alteration is observed. Motion of the elbow, driven by axial stresses in
the pipe and by the liquid pressure, causes the alteration. Experimental data are
presented to support these contentions. Two mechanisms are proposed as causes of
F.J.Hatfield the observed interaction of transient pressure and pipe motion. Verification that the
Department of Civil and Sanitary Engineering, mechanisms have been identified correctly is given by comparison of the ex-
Michigan State University, perimental data to predictions of an analytical model that incorporates the
East Lansing, Mich. 48824 mechanisms.
Introduction
Problem Description. Piping systems used for transfer of radial inertia of the liquid, mass of the pipe wall, and axial
pressurized liquids operate under time-varying conditions and bending stresses in the wall.
imposed by pump and valve operation. Traditionally, to There are several other ways in which pipe properties affect
analyze the unsteady behavior of the liquid, the equations of the dynamic response of contained liquid. The following
motion and continuity of the liquid are solved without regard discussion explores two such mechanisms that appear to be
to motion of the piping. The transients propagate at the significant but are neglected in contemporary engineering
acoustic velocity, or wavespeed, of the liquid in the pipe. The practice.
diameter, wall thickness, and elastic modulus of the piping are
used in computing the wavespeed, after which the liquid is Strain-Related Coupling. Strain-related, or Poisson,
assumed to be flowing through a straight, rigid pipe. coupling results from the transformation of circumferential
Recently there has been concern that the transient behavior strain caused by internal pressure to axial strain:
of liquid in a piping system that is neither rigid nor straight tx=-vee (1)
may differ from that predicted by a traditional rigid pipe Skalak [2] extended Joukowsky's formulation to include
analysis. It is reasoned that the dynamic forces exerted by the Poisson effects. The extension consisted of treating the pipe
liquid at fittings where flow direction or area changes (elbows, wall as an elastic membrane to include the axial stresses and
tees, valves, and reducers) can set the pipe in motion and that axial inertia of the pipe. For sudden valve closure, a tension
such motion can alter the liquid transient. Some investigators wave was found to propagate in the pipe wall at a wavespeed
have suggested that this alteration is either negligible or that near that of the pipe material; hence, a "precursor" wave
rigid pipe analysis provides a conservative estimate of the travels ahead of the main pressure wave in the liquid. The
transient pressure because of the transfer of energy out of the axial tension is a Poisson effect in response to pipe dilation
liquid and into the structure. However, experiments have caused by the pressure transient. An increase in liquid
shown that in some systems the elasticity of the piping am- pressure due to the tension wave was identified, but this in-
plifies transient pressure in the liquid to a significant degree. crease was small because it was caused by the small con-
Motion of piping is caused by dynamic forces in the liquid traction in pipe diameter due to the tension wave; it is a
and the pipe wall. The amplitude and frequency of the pipe second-order Poisson effect. Thorley [3] completed a study
motion also are functions of the mechanical properties and similar to Skalak's, including experimental validation of the
support conditions of the piping. Accurate analysis of theory. Williams [4] also conducted a similar study and found
pressure transients in piping requires formulation of the that longitudinal and flexural motion of piping caused
mechanisms by which liquid and piping are coupled, and damping that was greater than the viscous damping in the
incorporation of parameters describing the pipe structure. liquid.
Background. In the late 1800s Joukowsky [1] determined Walker and Phillips [5] formulated a one-dimensional,
that the wavespeed of liquid in a pipe, and hence the speed at axisymmetric system of six equations that included the radial
which liquid discontinuities propagate, is related to the and axial equations of motion of the pipe wall, two con-
relative circumferential stiffness of the pipe. He assumed that stitutive equations for the pipe wall, and the equations of
pressure is uniform across any cross-section and neglected the motion and continuity for the liquid. They reported that the
method "retains much of the rigor of the axisymmetric, two-
Contributed by the Fluids Engineering Division for publication in the JOUR-
dimensional approach" of Lin and Morgan [6]. They found
NAL OF FLUIDS ENGINEERING. Manuscript received by the Fluids Engineering their method ideal for transients where the generation of the
Division, May 17, 1984. pressure pulse occurs in several microseceonds, and that the
402/Vol. 107, SEPTEMBER 1985 Copyright © 1985 by ASME Transactions of the ASME
p2A, s A , ±C-f=±-^-Ql
f (6)
Fig. 3 Elbow
and the second two are the pipe wavespeeds:
Experimental observations were terminated at the onset of
column separation, about 80 ms. \3A=±C,= ±-j-HQ2 (7)
Experimental Uncertainty. For the pressure transducers
and accelerometers, the uncertainty is estimated by the where
manufacturer at ± 3 percent full scale, resulting in an ac- « , = ( E / P / ) 11/2
/2
,i „ _— {j<rin_\\n
af=(K/ Pf) (8)
curacy within ±45 kPa for pressure and ±0.015 m/s for l/2
1/2
2rK 1i „ r. r 2rK I
velocity. Another possible source of error is the conversion of , H= 1 + — (9)
the analog voltages to digital quantities by the A / D converter. '-['• eE* J L eE J
The uncertainty of that process is 0.03 percent. '2
Analytical Development
«.-[
= 1+
ept(l-q) J Q2
' -V ePlq(\-q) J
(10)
2rkl
To support the hypothesis that the experimental results q=(ai/af)2[ 1+ (11)
demonstrate strain-related and pressure resultant coupling, eE J
the observations were compared to the predictions of an The compatibility equations are:
analytical model incorporating those mechanisms. The model
utilizes the following assumptions: dp „ dV _ „ du da
±p ±PtCfGf (12)
1. Piping is prismatic, round, thin-walled, linearly elastic, ^i ^jt ^r ~di
homogeneous, and isotropic. dx
2. Caviation and column separation do not occur. which are valid along = ± Cf, and
3. Damping is negligible.
~di
4. A phenomenon may be neglected if its fundamental d V da
frequency is two or more orders of magnitude greater than the *
— ±^P/C,r - \ rr±p,C,GMl~ (13)
fundamental frequency of the liquid column. For the ex-
perimental piping, this assumption permits neglect of trans-
verse pressure variation, inertia associated with deformation which are valid along —— = ± C,
dt
of the pipe cross section, and pressure and density differences
along the length of an elbow. where
5. Transverse motion of a length of pipe is represented
adequately by its first flexural mode. Gf=~(l-Ql2), G, = -[l-(.a,/a/)2H2Q22] (14)
rv rv
These simplifying assumptions are valid for the experimental Figure 2 shows the characteristic representation in the x-t
piping but not necessarily for all pipe systems. Therefore, plane. Equations (12) and (13) are integrated along their
caution is advised in applying the analytical model to other respective characteristic lines; the two lines with positive slope
systems. are designated as C+ characteristics and the two with
Four-Equation Model. Internal pressure stresses the pipe negative slopes are C - characteristics. The four resulting
circumferentially and axially due to the Poisson effect. A finite difference equations are solved simultaneously for
0.36 m|<
t- L3
-O- L1
Case B -®- Case D _ 1000-
L2
0.36 m L
L J
® Valve L3
• Stiff Support
0-
Fig. 4 Structural restraint for four experimental cases
2000-•
20 40 60 80
TIME (ms)
.1000- 0.25-1
2040 60 80
TIME (ms)
Fig. 5 Comparison of experimental and predicted pressures for
Case A
discrete points along the pipe. At an elbow, additional in-
formation is necessary.
Elbow Boundary. An elbow is shown schematically in Fig. i r
3. Only motion in the plane of the elbow is considered since 0 20 40 60 80
out of plane motion is not coupled to internal pressure. TIME (ms)
Fig. 6 Comparison of experimental and predicted pressures and
Because of the comparatively short length of the elbow, the elbow velocities for Case B
variation of pressure and density from point 1 to point 2 is
negligible, and flow conservation may be expressed: restrained by two rigid supports. Figure 5 shows the predicted
and experimental pressure responses at the valve. The
( K , - « x ) = (K2 + «„) (15) pressure spike at 4 ms on the experimental trace is from valve
Flexural stiffness of the upstream and downstream pipe flutter and does not affect the subsequent pipe response. The
reaches connected to the elbow are denoted Ky and Kx, remainder of the trace is essentially flat, with only small
respectively. For the support conditions of the experiment, the amplitude, high frequency oscillations. If the pressure pulse
effective masses associated with flexural displacements were had been reflected at the elbow, it would have arrived at the
found to be comparatively small and therefore are neglected. valve in approximately 20 ms, but this is not apparent. In the
The equations of motion of the elbow simplify to: 40-60 ms range larger oscillations occur. These are from the
upstream elbow that could not be supported as rigidly. Notice
piAf-axA,=Kxuxj)1Af-ayAt=KyUy (16) the predicted response is not flat as would be predicted by a
Equations (15) and (16), the pressure invariance condition, traditional two-equation model, the variation being caused by
and the compatibility equations (12) and (13) are solved the second-order Poisson effect. For the analytical prediction,
simultaneously to determine pressure, stresses and velocities velocities of the elbow, ux and uy, are identically zero.
at an elbow.
Case B: Axial Stiffness: The supports on both sides of
elbow 1 are now removed, and the elbow is restrained in its
Comparison of Experimental and Numerical Results plane by the axial stiffness of leg LI in the x-direction and the
Four different conditions of structural restraint were in- axial stiffness of leg L2 in the y direction. An examination of
vestigated, both experimentally and analytically. Figure 4 Fig. 6 shows that the pressure response at PI is significantly
shows the restraints. The first setup, Case A, has elbows 1 and different from that of Case A. Elbow motion is initially in the
2 totally restrained. For the second setup, Case B, the stiff negative x-direction and is driven by the precursor stress
supports were removed at elbow 1 so that it is restrained only wave; this action results in an increase in the liquid pressure.
by the axial stiffnesses of legs LI and L2. The third setup, The pressure increase propagates to the valve and is recorded
Case C, has two restraints placed 0.36 m from elbows 1 and 2. at 13 ms. The elbow motion at 10 ms is the result of the
Axial translation of L2 is restrained by the bending stiffness primary pressure pulse driving the elbow in the positive x and
of the two short lengths. This arrangement is more flexible /-directions, resulting in a decrease in pressure as shown on
than Case B because pipes are much stiffer axially than in the PI curve at 20 ms. The elbow continues to vibrate at the
flexure. The fourth setup, Case D, represents a combination natural frequencies of the piping, and the pressure response is
of Case Band CaseC. a combined effect of the resultant velocities. The maximum
Responses were predicted by a computer program in- recorded pressure, occurring at 44 ms, is 22 percent above that
corporating the solution described in the preceding section. which would occur in a system with no elbow motion as in
The flexural stiffness coefficients, K, were calculated by the Case A. The maximum elbow velocity recorded is 0.27 m/s
stiffness method [19]. Flexural stiffnesses of the long lengths and the maximum displacement is approximately 0.5 mm.
(LI and L2) are comparatively small and were neglected. For
Case C, the two 0.36 m lengths each contribute a stiffness of Case C: Bending Stiffness. The two elbows and the con-
94800 N/m for translation in the /-direction. For Case D, K necting reach L2 are allowed to translate in the /-direction by
was determined to be 88700 N/m for the one short length. the flexibility of the attached piping. Because of the large
Axial stiffness is included in the four-equation model. axial stiffness of these short attached lengths, translation in
the x-direction is negligible. As shown in Fig. 7, the frequency
Case A: Immobile Elbow. Motion of elbow 1 was of vibration in the/-direction is much lower than that of Case
% 1000- 3 1000-
0-
20 40 60 80
TIME (ms)
TIME (ms)
Fig. 7 Comparison of experimental and predicted pressures and
elbow velocities for Case C