NSF GRANT GK-1297 THE OFFICE OF NAVAL RESEARCH
REPORT NO. AS-67-14 AND THE
AUGUST 1967 NATIONAL SCIENCE FOUNDATION
INVISCID AND VISCOUS MODELS OF THE VORTEX BREAKDOWN PHENOMENON
by
Hartmut H. K. Bossel
Submitted in partial satisfaction of the requirements
for the degree of Doctor of Philosophy in Engineering
Distribution of this report is unlimited
FACULTY INVESTIGATOR: M. Holt, Professor of Aeronautical Sciences
August 1967
University of California Aeronautical Sciences Division
ABSTRACT
In the case of high swirl and large Reynolds number, the
Navier-Stokes equations for rotationally symmetric incompressible flow are shown
to reduce to 1) a viscous parabolic system for slender (quasicylindrical)
flows and 2) an inviscid elliptic system for expanding (or contracting) flows.
The inviscid system is solved for the case of flow with initial rigid rotation in a cylindrical stream surface. Assuming different
down-stream boundary conditions, Fourier-Bessel series solutions are computed
for the supercritical (nonoscillatory) case and plotted. For very high
swirl values, closed and open bubbles of recirculating fluid are obtained
for certain cases, where the closed bubbles resemble those observed in vortex tube experiments.
The viscous slender problem is formulated in an integral method,
using approximating functions for axial velocity and circulation which satisfy boundary and asymptotic requirements. Weighting functions are
used to generate a linearly independent set of equations of sufficient
number to determine the coefficients in the velocity and circulation
approximations as functions of one (the axial) coordinate. The formu-lation is for any order of approximation. Computational difficulties appear near the suspected breakdown point. Flows resembling breakdown flows are obtained.
Page ABSTRACT TABLE OF CONTENTS ii LIST OF FIGURES v NOMENCLATURE vii 1 - INTRODUCTION 1
2 - VORTEX BREAKDOWN OBSERVATIONS 4
A survey of experiments 5
Definition 5
Features of vortex breakdown 6
Occurrence of vortex breakdown 7
Delta wing 7
Confined vortex flows 8
Conclusions 9
3 - THEORIES OF VORTEX BREAKDOWN 10
Separation analogy 10
Hydrodynamic instability 11
Standing wave 12
Finite transition
4 - AN AMPLIFICATION CRITERION 14
5 - EQUATIONS OF MOTION AND BOUNDARY CONDITIONS 17
Equations of motion 17
Two limiting cases 19
Slender vortex flows 19
Expanding vortex flows 20
Transformation of variables 22
Boundary conditions 24
6 - INVISCID SOLUTIONS 27
A general linear solution 27
Swirling flow in a cylindrical streamtube 30
Pressure on the axis 33
Velocities 34
Nondimensional results 35
Specific cases 36
Computational note 39
7 - INVISCID RESULTS AND DISCUSSION 40
Parameters 40
Stream surface plots 41
Velocity on the axis 42
Axial velocity profiles 42
Swirl velocity profiles 43
Swirl angle profiles 44
Summarizing discussion 45
An explanation of breakdown 47
8 - VISCOUS SOLUTIONS 49
Method of solution 49
Integral relations 52
Approximations for axial velocity and circulation 54
Axial velocity W 54
Circulation K 55
Weighting functions 56
The numerical method 61
Singularities 62
Computing programs 62
Flow models 63
Results for zero swirl 67
Results for increasing swirl 67
Summary and discussion of the viscous results 70
REFERENCES 73
FIGURES 79
Inviscid results 79 - 127 Viscous results 128 - 138
LIST OF FIGURES Inviscid results
Fig. 1 Coordinate system and velocities Fig. 2 Vortex breakdown in a tube
Fig. 3 Vortex core model
Fig. 4 Boundary conditions on ip
Fig. 5 Pressure on the axis for cases A and B with
R0 = 0.4 Figs. 6 Stream surfaces, velocity on axis, and axial
velocity profiles for cases A - F with zero swirl and L = 2
Figs. 7 Stream surfaces, velocity on axis, and axial
velocity, swirl
velocity, and swirl angle profiles for cases A - F with initial swirl
v Lw
= 1.75 and L = 2c
Figs. 8 Stream surfaces, velocity on axis, and axial
velocity, swirl
velocity, and swirl angle profiles for cases A - F with initial swirl v 7w = 1.91585 and L = 2
c
Figs. 9 Stream surfaces, velocity on axis, and axial
velocity, swirl velocity, and swirl angle profiles for cases A - F with initial swirl
v/w
= 1.191585 and L = 4Viscous results
Fig. 10 Flow chart for computing program VIVO 1 Fig. 11 Flow chart for computing program VIVO 2
Fig. 12 Initial axial and swirl velocity profiles. Ke = 2.0 Fig. 13 Velocity on axis, and axial velocity profile development
for profile B, no swirl
Fig. 14 Axial velocity profile development for
profile C, no swirl Fig. 15 Velocity on axis, and axial and swirl velocity
profile development for profiles A, ke = 2.0
development for profiles A, Ke = 2.8
Fig. 17 Initial axial and swirl velocity profiles, velocity on axis,
and axial and swirl velocity profile development for profiles
A, Ke = 3.2
Fig. 18 Velocity on axis, axial and swirl velocity profile development,
and stream surfaces for profiles B, Ke = 3.6
Fig. 19 Velocity on axis, axial and swirl velocity profile development, and stream surfaces for profiles B, Ke = 4.0
Fig. 20 Velocity on axis, axial and swirl velocity profile development, and stream surfaces for profiles B, Ke = 41
NOMENCLATURE
(Capital variables are nondimensional)
a(Z)
parameters of the velocity profileb(Z)
parameters of the circulation profilecik(Z)
coefficients in the system of ordinary differentiald ,k(Z) equations in the method of integral relations
weighting functions
Bessel functions of order zero and one
j1
zeros of the Bessel functions of order onek=vr circulation
h =u r
9. reference length
p pressure
p0
= p + --
(u2+v2+w2), total pressure r radial coordinate Re Reynolds number u radial velocity v swirl velocity w axial velocity y=r2/2
z axial coordinatea(Z) exponent parameter of the velocity profile (Z) exponent parameter of the circulation profile
y stream surface angle with respect to axial direction
2 V r w
C
p density
= arctan V/W, swirl angle stream function
Subscri pts
ax on axis
c outer edge of rigidly (or nearly rigidly) rotating core
e external flow
i initial
reference
Superscri pts
differentiation with respect to z
- INTRODUCTION
Since its first recorded observation in 1957, vortex breakdown
has been the object of numerous investigations. There is no lack now of experimental observations, and several theories have been advanced to
explain the phenomenon.
Vortex breakdown can occur in the vortex systems of flight
vehicles, seriously affecting their pressure distribution. It may occur in swirl diffusers, influencing their performance. It stabilizes flames and drastically changes the flow from rotating jets. It could be used
to contain a volume of fluid within a body of different fluid with
rela-tively little mixing. The phenomenon may have something to do with the
mechanism of transition to turbulence.
The goal set for this report was to compute vortex breakdown
flows.
Observations of vortex breakdown flows and related flows under
a variety of circumstances are reviewed in Section 2. A definition of vortex breakdown, to be used throughout this report, is based on their
common features. Conclusions about the importance of different flow
parameters are drawn from the experimental results.
The different theories that have been advanced to explain
vor-tex breakdown are briefly outlined in Section 3.
In Section 4 a criterion for stability against an axial velocity
disturbance is derived for a highly simplified model of vortex core flow.
The Navier-Stokes equations for incompressible rotationally
symmetric motion are examined in Section 5. By an order-of-magnitude analysis it is found that inviscid elliptic equations apply in the
(quasi-cylindrical) vortex flow. Boundary conditions and some expected features
of breakdown flows are discussed.
In Section 6 conditions are derived under which the inviscid
equation of motion becomes linear. It turns out that the linear equation
belongs to flows having a rigid rotation at some point along their axis,
a condition which also applies approximately over the inner core of a
viscous vortex.
Solutions are found by the method of separation of variables,
and are specialized to swirling flow in a cylindrical stream tube. The
solutions considered are in the supercritical (nonoscillatory) regime,
characterized by a maximum swirl angle = 62.5. Stream function and velocities are given in a Fourier-Bessel series solution. The series coefficients are obtained for six different cases.
Numerical results for the six cases of Section 6 are obtained
and plotted in Section 7. Closed and open bubbles are obtained for very high swirl values in some cases, bulges without free stagnation points
in others. Vortex breakdown is explained as a feature of flows with very
high swirl values having a retardation of axial velocity near the axis.
Section 8 presents an integral method to solve the viscous
slender system. The system is integrated in one direction by the use of
exponentially decaying approximating and weighting functions. The
func-tional dependence of the parameters in the approximating function is then
determined from ordinary linear first order differential equations. Some
features of the method appear to be novel and could also be applied in
Section 9 discusses results obtained with the integral method
and some computational difficulties. At initial swirl values approaching the critical one, flows resembling breakdown flows are obtained.
2 - VORTEX BREAKDOWN OBSERVATIONS
A survey of experiments
In investigations on swept wings (Peckham and Atkinson, 1957;
Lambourne and Pusey, 1958) the cores of the leading edge vortices (marked
by smoke or condensation) were observed to ubell out' (Pecicham and Atkinson, 1957) at some distance downstream from the apex of the wing. These first
observations spawned a number of experimental and theoretical investigations
into the nature of the phenomenon, now referred to as vortex bursting, or, preferably, vortex breakdown, since it is not a time-dependent phenomenon.
The majority of experiments were carried out on delta wings,
where the modification of force and moment coefficients is of immediate
practical interest (Elle, 1960; Lawford and Beauchamp, 1961; Hummel, 1965).
Vortex breakdown was intensively studied by means of smoke or dye filaments
in wind and water tunnels at very low to moderate speeds (Elle, 1958;
Werlë, 1960; Maltby and Keating, 1960; Das, 1961; Lambourne and Bryer, 1961;
Lowson, 1964; Poisson-Quinton and Werl, 1967). A few observations were
made at sonic and supersonic speeds (Elle, 1960; Lambourne and Bryer, 1961;
Craven and Alexander, l96). Rolls, Koenig, and Drinkwater, 1965, report
on flight observations of vortex breakdown in a humid atmosphere. An abrupt
change in radial vorticity distribution in the trailing vortex of a straight
wing was observed by investigators of the Pennsylvania State University in
1965.
It was soon realized that the phenomenon might be of a more
funda-mental nature and would not only appear in leading edge vortices. Breakdown
was shown to occur in vortex tubes (Lambourne and Bryer, 1961; Harvey, 1962
Irani, 1965). These flows clearly showed a stagnation point and reversed
flow downstream of it.
Stagnation points and reversed flows in swirling fluids have
also been observed in the vortex whistle (Vonnegut, 1954; Chanaud, 1965),
in water flows through a convergent-divergent nozzle and in a tube (Binnie,
Hookings, and Kamel, 1957; Binnie, 1957), in flames in vortex flow (Potter,
Wong, Berlad, 1958), in vortex chambers (Nissan and Bresan, 1961;
Donaldson and Snedeker, 1962), in flows through expanding cross sections
and diffusers (Strscheletzky, 1957; Gore and Ranz, 1964; So, 1967). There
is evidence for a stagnation point and reversed flow in some tornado funnels
(Morton, 1966).
Vortex breakdown has been discussed in survey articles by KUchemann
(1965) and Hall (l966b).
Although there are debatable differences between the flows
men-tioned, their common features will be used to define vortex breakdown for
the purpose of this report.
Definition
Vortex breakdown is the rapid expansion over a short axial
dis-tance (of the order of the thickness of the rotational core) of streamtubes
centered about the axis of a rotationally symmetric vortex whose velocity
components are continuous on the axis.
The breakdown may or may not result in stagnation on the axis
and reversed flow, and it may or may not be accompanied by circumferen-tially periodic flow (uspirallinghi) or breakdown into turbulent flow. The flow may be incompressible or compressible,inviscid or viscous.
This definition covers more flows than have previously been
classified by the term "vortex breakdown."
Features of vortex breakdown
The flows mentioned have in common the features of the defini-tion just given. In some cases (supersonic vortex breakdown, for example)
it is not clear whether a stagnation point and reversed flow exist. In most other cases stagnation points on the axis and axial flow reversal have been observed. In these cases vortex breakdown would appear as
follows (see, for example, Lambourne and Bryer, 1961; Harvey, 1962 [Figure
2]; Hummel, 1965).
The axial filament of fluid is suddenly decelerated to stagnation. The flow deflects as if it were encountering a solid object. The
deflec-tion can be axisymmetric, with the fluid in the axial filament expanding
in an initially egg-shaped surface of revolution which may or may not close
downstream. This is the case in most low-speed vortex tube observations (Harvey, 1962; So, 1967). The deflection can also be asymmetric, with the axial filament deflecting with a kink into a spiral configuration about
the axis. This appears to be the more common case in delta wing flows. The spiral persists for a few turns and then breaks down to turbulence. Downstream of the stagnation point there exists reversed axial flow.
Transition to turbulence may or may not occur at the rear of the
egg-shaped bubble in the case of symmetrical breakdown. Harvey (1962)
obtained a closed bubble downstream of which the initial conditions were
apparently restored, with a second breakdown further downstream. His experiments proved reversibility of the basic phenomenon.
Hummel's 1965 measurements of velocities and pressures upstream
and downstream of breakdown show that axial and swirl velocities, which
are initially 2 - 3 times the free stream velocity, are reduced to very
small positive, or negative, or zero values near the axis after breakdown.
Occurrence of vortex breakdown
Delta wins. On a delta wing at sufficient incidence vortex sheets separate from the leading edge and roll up into conical vortices with origin at the
apex of the wing. The generation of vorticity ends at the trailing edge,
and the vortices continue as cylindrical vortices, under viscous diffusion
of vorticity.
As the incidence of the wing is increased, breakdown first occurs far downstream, moving upstream with increasing angle of attack.
Higher effective leading edge sweep, i.e., decreasing angle
between leading edge and free stream direction due to the geometric combi-nation of sweep, incidence, and yaw shifts the first occurrence of breakdown to higher angles of attack (Elle, 1958). Werlé (1960) demonstrated that breakdown occurs further downstream on a thicker wing, and on a wing with
larger leading edge radius. He also showed that on a very slender cone, a rectangular, and an elliptical wing, breakdown does not occur in the immediate neighborhood of the body.
Elle (1958) found that the breakdown position exhibits hysteresis
with respect to yaw, and Lowson (1964) established hysteresis with respect
to incidence. He found two steady flow states to exist, one with breakdown on the wing and one without. Lowson also noted that the position at which
core instability first occurred could be very different from the final state
The effects of upstream and downstream conditions on vortex
breakdown were investigated in particular by Werl (1960) and Lambourne and Bryer (1961). No significant effect of upstream flow disturbances could be
established. Objects in the path of the laminar vortex (upstream of break-down) could precipitate breakdown, or cause an already established breakdown
to move further upstream. Hummel (1965) used a plate normal to the vortex, and some distance downstream of a delta wing to fix the position of
break-down on the wing, which was then almost insensitive to the insertion of a
probe. Wer1 (1960) and Lambourne and Bryer (1961) showed that suction on the vortex axis could retard or even completely eliminate breakdown. A
jet of fluid emitted in the direction of the flow caused the breakdown to
appear further downstream and the vortex axis to curve towards the jet. The influence of Reynolds number on the burst position is appar-ently small. Werlé (1960) reports some upstream movement with increasing Reynolds numbers while Elle (1960) notes that observations at Re l0 and up to Re = 5 x 106 are in perfect quantitative agreement.
Elle (1960) reports on Schlieren observations of the vortex break-down on delta wings at M = 0.7 to 1.03. At these Mach numbers he found the breakdown position to be approximately the same as for low speeds.
Craven and Alexander (1963) found vortex breakdown on a delta wing at M = 2
to exist at incidences only slightly less than those at which breakdown
occurs at low speeds.
Confined vortex flows: As the swirl is increased in a vortex tube of cylin-drical or expanding cross-section (Harvey, 1962; Kirkpatrick, 1965; So, 1967)
or in rotating flow issuing through an expanding cross section or as a free
fluid with a stagnation point at its upstream end appears on the axis
(Figure 2). As the swirl is increased the bubble proceeds upstream until
(in the cases of Harvey, 1962; Kirkpatrick, 1965; So, 1967) a two-celled
vortex with reversed axial flow is established in the duct.
Gore and Ranz (1964) find that the critical swirl ratio for
breakdown is independent of Reynolds number in the range 20 < Re < 60 x
Concl usi ons
From these experiments the following conclusions can be drawn: Swirl is the primary variable. Breakdown will not occur at low swirl values.
The axial pressure gradient affects the breakdown position. An adverse
pressure gradient can apparently precipitate a breakdown which would
otherwise not occur.
Reynolds number effects are not important to the basic phenomenon. Mach number effects are minor.
Influence of upstream turbulence is small, but influence of downstream
disturbances is pronounced since upstream decay of disturbance is much
3 - THEORIES OF VORTEX BREAKDOWN
Several explanations of vortex breakdown have been proposed. They are variants of one of the four propositions:
Vortex breakdown is a separation phenomenon.
Vortex breakdown is the result of hydrodynamic instability. Vortex breakdown is a standing wave phenomenon.
Vortex breakdown marks a finite transition between two conjugate
flow states, analogous to the hydraulic jump.
Separation
Stagnation on the axis with reversed flow downstream implies a
separation-like phenomenon. Strscheletzky (1957) distinguishes between two
kinds of separation: 1) separation from a wall, caused by viscosity, and
2) the formation of discontinuities under the action of inertia forces, and
depending on the hydrodynamic equilibrium of the flow. Reversed flow regions in swirl diffusers are separated from the basic flow by discontinu-ity surfaces of the 2nd order (discontinuities only in the normal derivatives
of velocity), and his calculations indicate a strong dependence of the
geometry of the back flow region on the swirl distribution in the basic flow.
Talbot (1954) applied a one parameter momentum integral method in
his study of swirling laminar flow in a pipe. For large values of this parameter reversed axial flow results. Using the full viscous system of equations Vaisey (1956) computed swirling flows in a tube by a relaxation
method. Depending on Reynolds number and swirl she finds three regimes of
flow: 1) reversed flow in an annular region between axis and wall; 2) down-stream flow everywhere; 3) reversed flow in a region containing the axis.
Fraenkel (1956) finds reversed flow in an analysis of inviscid pipe flow.
Gartshore (1962, 1963) uses the Karman-Pohlhausen technique and
boundary-layer type approximations to compute the flow of (a) a viscous vortex in a uniform stream parallel to the vortex axis (1962) and (b) a viscous vortex imbedded in a concentric irrotational vortex (1963). The
occurrence of singularities for some choices of the parameters is suggested
to be the result of a violation of the boundary layer type assumptions,
and to indicate the occurrence of asymmetries or large axial gradients.
He proposes that vortex breakdown may correspond to these singularities
in a way analogous to the occurrence of separation in two-dimensional boundary layer flow.
Hall's computations of vortex flows by a finite difference method
(Hall, 1966a) and their failure near observed breakdown positions also lead
him to view the breakdown as a failure of the boundary layer type
("quasi-cylindrical") approximation in analogy to problems in the calculation of two-dimensional boundary layer separation.
Hydrodyriamic instability
Jones (1960) assumes axisymmetric disturbances to the inviscid
flow and looks for conditions for amplification. He does not obtain practi-cal results.
Ludwieg (1960, 1961) derives a stability criterion involving the radial gradients of the axial and swirl velocities for helical flow
between two concentric cylinders by assuming spiral disturbances to
the inviscid flow. The theory is supported by a large number of experiments (Ludwieg, 1964). The criterion is applied to the vortex
breakdown (Ludwieg 1962, 1965), giving good agreement with experimental
data obtained by Hummel (1965).
Standing wave
Squire (1960) looks for conditions under which a sinusoidal
cylin-drical wave of unspecified wave length could first be sustained. He
con-siders three different swirl distributions and finds critical values of the
ratio swirl velocity/axial velocity of the order of unity. This is in agreement with experimental observations.
Finite transition
Benjamin (1962, 1965, 1967; Fraenkel, 1967) considers inviscid
swirling flow. He shows that for cylindrical flow the equation of motion
in general does not possess a unique solution for the stream function p
between given endpoints. An indefinitely large number of solutions is obtained by varying the two arbitrary constants in the general solution.
Only two solutions, however, appear to be relevant; their solution curves
p(r) do not intersect except at the endpoints (r = 0, r = re). Hence
a different cylindrical flow is possible at another station of the same
duct, keeping the same distribution of Bernoulli constant and circulation
over their common stream surfaces. The two flows A and B are dynami-cally and kinematidynami-cally conjugate. With the aid of the calculus of vari-ations Benjamin shows
If flow A is supercritical (i.e., cannot support standing waves), then flow B is subcritical (standing waves can form).
The supercritical flow has a smaller value of the momentum flux
r
M = 2
f (p
w2 + p)r dr.Benjamin then explains vortex breakdown as follows: The flow
changes from a supercritical state to a conjugate subcritical state with
an increase in momentum flux M. If tM is small, it can be made up by the positive wave resistance of a stationary wave train. If M is large, the leading wave breaks, and vigorous turbulence results with a
4 - AN AMPLIFICATION CRITERION
Before going into the analysis of vortex core flows it may be useful
to look at a simplified model.
Consider the inner portion of a physical vortex (Fig. 3). Due to
the action of viscosity it will be in nearly solid body rotation close to
the axis, and its axial velocity is therefore assumed constant (w =
w)
over the radius for 0 < r <r.
Shear is then nonexistent, and the fluid is treated as inviscid. Let the axial velocity of the cylinder of fluid change by an amount Aw. The results are derived by neglecting small quan-tities of order A2By continuity the radius of the cylinder will change and
rc2wc = (r + ir)2(w + tw)
This leads to
AW
-
-(4.1)
where Ar/Aw is the expansion of the stream surface p(r) due to the
disturbance Aw.
By conservation of angular momentum we have for an element of fluid at rc
rv
= (r Ar)(v + Av)giving
V
Ar rc (4.2)
or
Av vc
AW 2w
This is the change of swirl velocity v at r = r due to the axial velocity disturbance w.
The centrifugal balance at r = rc + r is, to first order
p(Vc + Av) (w + Aw)Kw = giving 2 Kw Vc
-
2wThe initial velocity disturbance Aw will be amplified, i.e., > 1 for AW 2 vc 2wc2 > 1
--Ap
V2AW
p2w-Aw
c (4.3) (4.4)If the initial disturbance is decelerating, then further deceleration
and eventual stagnation will result for
Ar r + Ar
Replacing v and r by using expressions (4.1) and (4.2) leads to 2
- 2Wc_AW
The change in pressure will result in a new w = Kw by way of the axial momentum balance
or a swirl angle
= arctan > arctan V' 54.8°
wc (4.5)
In a viscous vortex the swirl angle increases from zero at the axis
to a maximum near the maximum for the swirl velocity, after which it again
falls off to zero. We can therefore expect that the swirl angle at this point, commonly taken as the characteristic radius of the viscous core, is of critical importance to the behavior of the vortex flow. In fact,
all observations of breakdown have maximum swirl angles close to the value
given in (4.5) (see, for example, Harvey, 1962; Hummel, 1965). It also
agrees well with the theoretical values given by Squire (1960) and by
5 - EQUATIONS OF MOTION AND BOUNDARY CONDITIONS
Equations of motion
The dimensional equations of motion for viscous incompressible
flow with rotational symmetry are in cylindrical coordinates
r Br Bz = 0 (5.1) 2 =
].a+
B2u u 1 Br Bz rpBr
2 Br Bz r =)[-_+i+--i_ v
1 r 2 Br Bz r Br Bz 2 r =_]2+
B2w1BW+B2W]
Br Bz p Bz2rBr
BrIt can be expected that this set of equations is governed by a
number of nondimensional parameters, and that it will reduce to simpler
sets of equations for certain choices of these parameters. To find the
parameters and investigate limiting forms of the set of equations, the
dependent and independent variables are non-diniensionalized. Reference
quantities are chosen such that the dependent variables and their
deriv-atives are of order unity for typical problems with continuous velocities
on the axis. We use
w = the external (free stream) axial velocity of the potential
flow outside of the rotational core at a reference station. vc = the maximum swirl velocity at a reference station. In a
viscous vortex it occurs near the outer edge of the
rota-tional core, and roughly marks the point of transition from
a (nearly) rigid body rotation in the center to (nearly) potential vortex flow outside.
set
r = a radius representing core thickness. In a viscous vortex it is convenient to take rc where v =
v.
= stream surface angle at r =
r.
We have = tan y.The dimensionless variables are
_r
U . p--
U =___
rwtan-
2 p Vc z V V-
re/tan
I Vc W w w 00Substitution into system (5.1) results in the nondimensional
-R R Z -0 (5.2) 2 V
22
1 (c)
-
BZtany
2 V2P(
1=
-ta -i
)wrc
tan ylU
22U
u -a-3R2 = ( V\(
1[V+iaV+(
any) --
22V
V -Z Rwr
'tany
Z Ruw
-CP
V 1wr
tan + i + (tan )2 W I R zThe three parameters of the problem are
wr
a core Reynolds number CV vC
a swirl parameter
-w
a core stream surface angle y
Two limiting cases
Depending on choice of core Reynolds number, swirl parameter, and
core stream surface angle, system (5.2) will reduce to various simpler sets. Two cases are of special interest to the problem at hand: Slender (nearly parallel) vortex flows and expanding vortex flows, both with swirl parameters
near unity (i.e., near the breakdown condition) and relatively large core
Reynolds numbers (core Reynolds numbers for typical breakdown cases are of
the order of io).
Slender vortex flows. Let
= 0(1)
tan y << 1
wr
cC
tan y = 0(1)
Neglecting terms of O(tan2y), the set (5.2) reduces to
U + R R Z -R U + W aV UV - V 1 1 V V
ZR -
wr
tany
R2 0U + W v 2 DP + V tan y 1 DW -
-(--)
wr
DR2 The corresponding dimensional system isr Dr Dz = 0 (5.3) r
pDr
=V[_+]i V
Dr Dz r Dr2rDrr21
Br BzpBz
DrrDr
The same set of equations, although found by different approaches,
was used by Gartshore (1962,1963), and Hall (1965,1966a), who calls it
'quasi-cylindrical
Expanding vortex flows. Let
= 0(1) tan y = 0(1)
wr
>> V Neglecting terms of 0(wrc
), the set (5.2) becomesU + BU + DW R DR BZ - 0 2 v 2 2 2 v
2BP
1)(_c
V DR BZtany
w =tanyH
UV = 0u+ww
vc 2P
The corresponding dimensional system is
= 0 r r z + u V U r
u+w+
= 0 z r-pz
The system is inviscid, showing that in the case of expanding
(or contracting) viscous vortex flows inertia and pressure terms dominate the motion.
The last three equations can be written in the vector form of
the Euler equations
ix
curl J = i-grad p0 where p r p0 =p+
-2 equation. u =Introduction of a stream-function p eliminates the continuity
In cylindrical coordinates
w
r r
(5.4)
By Kelvin's theorem the circulation is constant along a stream surface, i.e.,
Similarly
p0 =
p0(p)
Using these concepts, the differential equation for the stream
function in steady rotating inviscid flow can be derived (see, for instance,
Fraenkel, 1956). B2 =
-k---i-2dp
r 0 Br2 r Br Bz2 dtppdip
Transformation of variablesThe equations of motion in cylindrical coordinates can be simpli-fied by using the new variables
2
y=-
-h = ur
k = yr
The inviscid equation (5.5) then becomes
+2y4
-d (5.6)
The viscous slender system (5.3) transforms to
Bw + By Bz
h--
+ Bk -BW + Bw w- -= 0 (5.7) 32k 2y
-By 3w = p Bz (5.5)The equations are nondimensionalized by using the reference length
rc and a reference velocity w00. Then
wr
V Re R =r
Y = rcz
=-r VR R U =.0
H =wr
h w 00 v = K e wwr
k 00 dK dP 2K2Y2
The system of equations for viscous slender steady rotationally
symmetric rotating flow is
(5.8) (5.9) w w w 00
-
pw2
200
w rIn the inviscid case there is, of course, no Reynolds number stretching, and the same nondimensional variables are used formally with
Re = 1.
The nondimensional inviscid equation of steady, rotationally
symmetric rotating flow is then: P
3H 3Y + 3H 3Y + 3W 3W = 0 (5.10) K2 4Y2 3K 3K -
2Y1-+ W5- -3Y 3W 3W + 2(y.
)H--
+W-7
-In "divergence form," using the continuity equation:
= 0 (5.11) K2 4y2 .- (HK - 2Y - + 2K) +
-h-
(WK) = 0-(HW-2Y--) -(W2+P)
= 0 Boundary conditionsViscous axisymmetric vortex flows which have no sources or sinks on the axis require
u = 0 v = 0 , - = 0
Only inviscid flows which satisfy the same conditions on the axis will be
considered in this report.
Symmetry demands, besides (3w/3r) = 0, (3v/3r) = const (z) and (32v/3r2) = 0 on the axis, i.e., rigid body rotation and shear-free flow in an infinitesimal neighborhood of the axis of the viscous vortex. In most vortex flows the departure from rigid body rotation becomes appreciable
only when the radius becomes of the order of the viscous core radius r
defined earlier. The vortex breakdown bubble observed by Harvey (1962) has a radius of the order of the rigidly rotating part of the
vortex,indi-cating that the upstream boundary conditions for the part of the flow
experiencing the breakdown were nearly those of rigid body rotation. As the flow stagnates on the axis (u = v = w = 0) it follows from the equations of motion that
0
at the stagnation point.
The stream surface extending from the stagnation point is (in
the viscous case: initially) a surface of zero swirl velocity. The swirl
velocity distribution in a continuous solution could therefore, downstream
of the stagnation point have the forms (a) or (b).
N
/
()
'I(h)
r
An examination of the equation of motion (5.1) in the circumfer-ential direction under the conditions v = 0,
(v/r) = 0,
(2v/r2)J
>>excludes possibility (b). The conclusion is that unless the bubble surface is a surface of discontinuity (which is likely in some cases,
see Strscheletzky, 1957) the rotation inside the bubble will be opposite
V
to that before stagnation. Reversed rotation has indeed been observed by Hummel (1965) and researchers of the University of Pennsylvania (1965).
The axial velocity near the axis must, of course, be reversed
downstream of the stagnation point. The flow close to the inside of the
bubble surface must later again be in the downstream direction to satisfy
continuity requirements. From the experiments the bubble may be closed or open downstream.
Outer boundary conditions are mentioned with the respective
solutions. In the case of an unconfined vortex the flow will be that of a potential vortex far outside the viscous rotational core.
6 - INVISCID SOLUTIONS
A general linear solution
The results of the discussions of conditions in a real vortex
and of its equations of motion as it expands support the idea that something
about the mechanism of vortex breakdown may be learned from considering the
inviscid equation (5.5) or (5.6). We shall look for solutions which are linear (to make superposition of solutions possible), and which satisfy the
boundary conditions on the axis:
u=O
,v=O
,The equation of steady inviscid rotating motion, Eq. (5.6), is
-
- k+
dp0 d1p p
This partial differential equation will be linear if the highest
power of i on the right hand side is one. This permits a2 2 p
a0+a1i4+p
k2 b2 2 b 1 Substituting in (6.1) + 2y + 2 ip(b2-2a2y) = 2ya1-b1 An immediate solution is = WOQYSubstitution in (6.2) leads to the conditions
(6.1)
and
giving
p
2a2wQ,y2 = 0 hence a2 = 0
2a 2a1
-b2w,y = - 2a1y hence b2 = , or w
-b1 = 0 (6.3) 0 = a + a1iP = a0 + a1w,y 2 b = b 2 0
and, by the boundary condition on v on the axis, (demanding b0 = 0)
k2 2 2
= a1w00y = yv
Therefore v = v'a1wJ = const r = r
While =
wy gives
w = = w
The solution is therefore a rigid rotation. Equation (6.2) now reduces to
y rc w r
By superposition assume the solution
WJ+1J)1
(6.4)
Substitution in (6.4) leads to the separable partial differential equation for 11)1
v 2
1+2y
2h
= w C V 2 2 4 ) and assuming Introducing A -r Co = F(y)G(z)the equation separates into two ordinary differential equations (i.i is the separation constant)
G" + (x2 - 2)G = 0 2
F" +F =
0 with the solutions2
G(z) = A e Z + B
e2
Zand
F(y) = C /7 J1(p v'2) + D /57 Y1(1i /2i)
The general solution of Eq. (6.4) now becomes
22
(y,z) = wj+[Cv J1(p)+D
J1()](A e
-X Z + B e -x Z)In terms of r = w
(r,z) -
+[rJ1(pr)+rY1(r)](
e22
Z +
e2
Z)(6.5)
This solution was given by Long (1953) and was used by Long (1956) to compute the flow in a rotating cylinder into a point sink on the axis.Fraenkel (1956) solved Eq. (6.1) by application of Fourier
trans-forms to compute rotating flow through contractions and expansions of a
pipe, and past bodies in a pipe. In the case of an expansion he finds reversed flow far downstream near the axis.
An important observation about solution (6.5) is that it becomes
oscillatory when the argument of the square root in the exponentials changes
sign. At this point standing waves can first appear.
Equation (6.1) has a counterpart in stratified flows with
solu-tions analogous to (6.5) (see Yih, 1965, chapter 6).
Swirling flow in a cylindrical streamtube
Solution (6.5) of the equation for inviscid rotating flow (6.1)
will now be specialized to describe the swirling flow within a cylindrical
streamtube which is in solid body rotation at some axial position z. It can be expected from the preceding discussions that this solution will show
some of the features of bounded or unbounded (free or confined) vortex core
flows.
We limit the present solution to the nonoscillatory mode, requiring
> 0
This results in cosh, sinh type solutions. The oscillatory mode would be
periodic and undamped in the z-direction. Both features are not dominant
in breakdown flows, but they will be briefly discussed later.
The problem is elliptic and requires conditions on a closed
bound-ary surrounding the field where the solution is being sought. The following conditions will be prescribed (Fig. 4):
1) (o,z) = 0 w r2 2) p(r,o) = ____
wr2
3) p(r,z) = 2 4) p(r,i) =Here
The solution will be constructed by superimposing 2
wr
poand an infinite series constructed from the basic solution
/22
/22
Vp
-x
z-Vp
-x
ze + e )
0 on the axis requires D and
if
= 0 in solution (6.5).Proof:
urn lim cos
r J(r) - J(r)
Y1(r) =
Y(r)
=vl
sin For r << 1 Yn(r) (n-i)! 2 n()
1 2 Y1(r) - ( .j ) HencerY1(pr)
r 2 = 2 =finite0
for r-0.
Therefore
DrYi(pnr) = 0 for r - 0 requires 0 = 0. 2
The requirement (r,o) and the condition of nonoscillatory
behavior demand a sinh type solution in z.
The condition (rc,z) c on the outer boundary requires
vanishing of the Bessel function component at r = rc or
Hence
l.Inrc =
un
must be the zeros of the Bessel function and
where Un
The argument of the square root in the exponentials then becomes
.2
v 2 Un2_ 2 = 12 r w c C 2v 2 =L[ 2(c)]
rc2 31n " wwhere Vc/W is the swirl parameter based on the axial velocity and the swirl velocity at the outer radius at the station where the flow is in
solid body rotation.
It is evident now that the flow will change its mode and switch to oscillatory behavior when
w > ill = 3.83170597...
This corresponds to a swirl angle at radius r of in
3.8317... = arctan
2 62.5°
Under the stated conditions the solution to the problem becomes
2
wr
(r,z) + rCJ1(j1
)sinh(Jj - (2 (6.7) rc n= 1 R f(r)ri31n
) dr = 2 rsinh [j
- ( )2 9. ] c (6.6) (6.8)Vc 3.8317...
- 2
C
= 1.9158...
a standing wave of infinite wave length appears. The wavelength shrinks
with increasing swirl parameter, until a second wave of infinite wave length
is superimposed on the flow when = j12. We can therefore expect mixed solutions of the type
w r2 k sin V 2v 2 r . c .2 z 31n (r,z)
= -+ r
CnJl(ln
i n=11/2
2v / r (C)2Z
+r
E C n=k+l fl1 in ) sinh Vain - ;;-rfor swirl values exceeding = 1.9158... . These solutions may app]y
to breakdown flows where oscillatory behavior of the core filament has been
observed ahead of the breakdown. This will not be further discussed in this report.
Pressure on the axis
The pressure on the axis follows from the Bernoulli equation
p0 = p
+(u2
+v2 +w2)
with 2wr
f(r) =[(r)
2 (6.9)given by prescribing the downstream flow iI(r). In a following section
several solutions will be given for different assumptions for
A remark is in order concerning oscillatory flows. As the swirl
With 1
w-andu=0
on the axis V=0
we have (p0-p) 1/2 j 2 2v )2 ax 1 2 = w r sinh -rp12w
cn1
c (6.10) where the are given by Eq. (6.8).The pressure distribution is uniquely determined by the down-stream conditions by way of the common
C.
The problem can then be turned around in a physically more meaningful way:If the pressure on the axis is prescribed and belongs to the
set given by (6.10) the downstream flow for a given swirl follows uniquely
from Eqs. (6.7) to (6.9).
Velocities
The axial velocity at any point in the region under consideration
follows from w = r ar 2v 2' w + l c n=1 ln n o 31n ) sinh - ( -s- z j
CJ(
rwr
n w c c (6.11)rv is a constant on = const, hence using the initial (rigid
or
and
where
Fn* = V =
and the initial stream function distribution
2 = we get rv Vc
-
rc rc Wo:, 2v v c 1 = w r r co Nondimensional resultsIntroducing the nondimensional variables (5.8) (with the formal substitution Re = 1) if' R
G(Z)
= wrc2 / 2 r rc 1wr3
w V W w v wOax
z = ax -p/2w2
* .2 2v 2' F sinh n Z 2v-'--)2L
vJln w rc f(r)r i3'in i:;;- ) dr L -r C (6.12) (6.13) (6.14) (6.15)Then stream function
(R,z) = R2 + 4R
n1
i1n
R)G(Z)
Pressure on the axis
Ip
)l/2 = 1 + 2nl
31nG(Z)
ax Axial velocity W = 1 + 2in oin
R)G(Z)
n= 1 Swirl velocity 2v V =c'i'
Bubble surface (R(i=O)): given implicitly by
R2 + 4R J1(j1
R)G(Z)
= 0 n= 1provided there exists a solution for R 0
Stagnation point: for W = 0 on the axis Z must satisfy
n=l
G(Z) +
= 0Specific cases
In typical vortex breakdown observations, the phenomenon has been
characteristically linked with one or more of the following features down-stream of the breakdown:
Reversed flow on the axis An object at or near the axis
(6.16)
(6.17)
(6.18)
Unreversed flow on the axis (in the case of a closed bubble)
Cases A through F were designed to simulate these downstream
boundary conditions.
Case A simulates the case of a disk of radius R0 on the axis at Z = L. The outflow has--just like the inflow--uniform axial velocity
W over the cross sections, i.e., the highest power of R in the expression for is two.
In case B there is a slight inflow from downstream on and near the
axis--and corresponding outflow in the downstream direction--inside a surface
0 which passes through R = R0 at Z = L and which divides fluid originating from upstream from that originating downstream. The highest power in the stream function is now four for reasons of symmetry, giving a parabolic distribution of the axial velocity W at Z = L.
In case C a disk of radius R = R0 is simulated on the axis at Z = L. Related to case B, the highest power of R in L(R) is again four, with a parabolic distribution of the outgoing axial velocity W.
Case D has again a parabolic axial velocity distribution at Z = L
with the highest power of R in L(R) again four. The flow is now assumed to be in the downstream direction everywhere at Z = L, having a velocity on the axis of W = Wax
In case E the highest power of R in L(R) is six, giving an
axial velocity variation with the fourth power of R at Z = L. Slight
inflow is assumed from downstream on and near the axis, returning to downstream
inside of a stream surface P = 0, which passes through R = R0 at Z = L. In case F, as in case D, there is outflow everywhere at Z = L
with a velocity W Wax on the axis, and varying as the fourth power of R. Correspondingly the highest power of R in
In the following equations R0 is a radius 0 < R < 1 at
which the stream function becomes zero (i.e., a streamline splitting from
the axial stream line will pass here).
Case A: Downstream obstacle, quadratic relation for
For O<R<R0:
= 0 R2-R 2 For R < R < 1: = l-R0 * R0 F = -2 2 Ji(uinRo) ln (l-R0 )Case B: Downstream inflow, quartic relation for ''
2R2-R2
LFor 0<R<l:
= R l-R02 *4oin
F=
n (l-R02)j1Case C: Downstream obstacle, quartic relation for
For 0 < R < R0: '1( = For R0 < R < 1: L2 = R2 R2-R02 l-R02 (6. 20) (6.21) (6.22) Fn* =
j(iR02)
R )+R { 31n 2 in 0 0 31n R0 }Case D: Downstream outflow, quartic relation for
For 0 < R 1: = R2[W + (1_Wax)R2] * F = --- (l-W ) J (j ) (6.23) n .3 ax 0 in 3ln
Case E: Downstream inflow, sixth power relation for
44
2 R For 0 < R < 1: 'PL(R) = RlR04
(12
-
J0(j1)
*
un
F=
(l-R04) 31nCase F: Downstream outflow, sixth power relation for
For 0 < R < 1: = R2[W
)R]
ax96
(12 --- )l_Wax)Jo(jin)*
31nF=
n .3 3m Computatioa1 noteTo avoid computational difficulties, replace
sinh(aZ)
sinh(aL)
by ean(Z_L)e_an+t
1 where a\Ij
2v 2 =Cases A through E are computed by using the first 19 terms of the
infinite series expressions.
(6.24)
7 - INVISCID RESULTS AND DISCUSSION Parameters
The parameters of the solutions for cases A through F are
the swirl v 1w
the length ratio i/re = L the radius R0
the axial velocity Wax at Z = L
Solutions were obtained for different combinations of these
parameters, and computer plots were prepared for the following cases
The swirl value of = 1.91585 is just barely below the theoretical value of j11/2 = 3.8317.. ./2 which divides nonoscillatory
(supercritical) from oscillatory (subcritical) flows. All cases discussed
here belong to the nonoscillatory class.
The results of the calculations are plotted in Figures 6 through
9. They show stream surfaces 'F = const, the velocity on the axis, and
radial profiles at different Z-stations of axial velocity W, swirl velocity V, and swirl angle . Before more general conclusions are
drawn the five different kinds of plots will be discussed separately. L = 2, Vc/W = 0: R0 = 0.1, Wax = 0.2 (Figs. 6) L = 2,
v/w
= 1.75: R0 = 0.1, Wax = 0.2 (Figs. 7)L = 2, Vc/W = 1.91585: R0 = 0.0,0.1,0.3, Wax = 0.2 (Figs. 8) L = 4, Vc/W = 1.91585: R0 = 0.1, Wax = 0.2 (Figs. 9)
Stream surface plots
The stream surface plots for the case of no swirl (Figs. 6) have
the features of other nonrotating flows: Substantial upstream influence is
confined to the immediate neighborhood of the disturbance, represented by
the downstream condition on 1'. The picture changes little as swirl is
intro-duced and increased. The character of the flow remains essentially non-rotational up to very high swirl values of the order of 1.75 (or 60°) (Fig. 7). At this value a bubble has appeared in some cases. It grows
rapidly in size and upstream dimension as the swirl is further increased
to the limit value of 1.9185 or q 62.5° (Figs. 8 and 9). At the same
time the upstream influence of the downstream flow increases drastically,
as evidenced by the slope of the stream surfaces.
The free bubbles have the characteristic shape of an egg: a blunt
front, then gradually increasing diameter, finally a fairly rapid decrease
in diameter to a blunt end. These features are also evident in Harvey's (1962) photographs (Fig. 2). The forward portions of the other bubbles, with the exception of case A have the same features. In the case of the greater length parameter (L = 4) the bubbles are more carrot-shaped. They
remind one (especially in cases where L is even greater) of the tips of
certain tornado funnels with stagnation points on their axes (Morton, 1966).
It is important to note that the appearance of the bubbles is a
continuous phenomenon. This becomes expecially clear from case F: At zero
and small swirl this case has simply a velocity retardation on and near the
axis with a corresponding widening of the stream surfaces. The velocity
minimum appears at the end of the region at Z = L. As the swirl is increased a bulge forms in the surface contours, indicating the minimum in the axial
velocity has moved upstream. At some higher swirl value this minimum reaches the value zero on the axis, corresponding to a free stagnation
point, or a hlbubbleu of infinitesimal size. The bubble grows rapidly with
further increase in swirl.
Velocity on the axis
Since U and V are zero on the axis,
P
ax ax
In addition to the plots of velocity on the axis (Figs. 6 - 9), 'ax
as a function of swirl is given in Figs. 5 for cases A and B with R0 = 0.4.
Again there is a striking effect of swirl on the pressure and
velocity distributions. While the changes are slight as swirl is increased
to a value near 1.5 ( 56.3°) they become quite pronounced as it is
increased further to 1.91585. The high swirl permits a much greater upstream
penetration of downstream velocity and pressure disturbances. The effect is less pronounced in case A.
Some cases have velocity minima at some Z < L, corresponding to the bulges in the stream surfaces. Where stagnation points appear the velocities on the axis are reversed and comparatively small downstream of
the first free stagnation point.
In cases B through F the velocity on the axis decays practically
linearly to the stagnation point at high swirl values.
Axial velocity profiles
In the flow models considered the axial velocity at and near the
lower than the constant axial velocity of the solid body rotation at the
upstream end (Figs. 6 - 9). At low--and even fairly high--swirl values the velocity changes monotonically from one to the other distribution.
At very high swirl values the axial velocity near the axis is
retarded even more before it speeds up again to meet the required condi-tions at the downstream end. In some cases this leads to free stagnation points and bubbles with reversed flow near the axis. The axial velocities
in the bubble are of the order of 10% of the initial rigid core velocity,
with the highest values occurring on the axis and on the bubble surface.
Extending from the forward stagnation point there is a surface of zero
axial velocity inside the bubble.
With the exception of the profiles at the downstream end of case
A all velocity profiles are strikingly similar, especially in the bubble
region. The profiles at the breakdown point (first free stagnation point)
are virtually indistinguishable from each other for cases B through F.
Swirl velocity profiles
The initially linear dependence of the swirl velocity on R changes gradually and monotonically to the assumed downstream conditions
when the swirl is not too high (Figs. 6 - 9).
At high swirl values V/R becomes zero at a free stagnation point
on the axis. There are reversed swirl velocities in the bubble with magni-tudes of the order of up to 10% of the initial maximum swirl velocity. On the bubble surface the swirl velocity is zero. For all cases with the
exception of case A, the swirl velocity profiles through the bubble region
and therefore
Since q' = const on the outer boundary R = 1, the swirl
velocity is also constant there.
Swirl angle profiles
In the cases considered the maximum swirl angle always appears on the outer boundary (Figs. 6 - 9). Initially it is given by the swirl
parameter of the problem, i.e.,
qi =
arctan (-i)
Since the swirl velocity on the outer stream surface (R = 1) remains constant while W(R=1) increases, the swirl angle there decreases
accordi ngly.
With increasing Z the swirl angle profiles become increasingly more linear in their outer portions. In cases B through F the breakdown
point is marked by a swirl angle profile which is linear all the way to the
axis. The swirl angle is zero on the bubble surface and then increases,
having again a linear dependence in the outer portion. In the linear region
q/R
has almost the same value at all stations. In this region constant swirl angles are found on nearly conical surfaces.The linear R-dependence of the swirl at the breakdown point relates
/R at the breakdown to the swirl angle at the boundary at this station.
It turns out that, at the breakdown position, vc
vc /(R=l)
= 2w
For the case of maximum swirl the outer swirl angle at the break-down point is q 54.8°, and
q/R
0.958.At all other positions
4/R
is higher. If this minimum value of p/3R is not reached on the axis, no stagnation point will appear (seecase 0).
In case F at high swirl rigid body rotation is again restored near
the axis at the downstream end, changing to a flow with constant swirl angle
in the outer portion.
Sumarizing discussion
The computer experiments have shown that under certain conditions
supercritical inviscid flows with initial solid body rotation can have free
stagnation points on the axis. From the first of these a bubble-shaped stream surface originates which contains reversed axial flow on and near the axis and reversed swirl velocities throughout.
If a bubble with stagnation points and reversed flow does not appear, the stream surfaces may still expand and again contract to form a bulge. The
transition from the bulge to the bubble is a smooth one and is just a matter
of sufficient velocity retardation on the axis. Both phenomena are therefore denoted with the common term tvortex breakdown."
*
These relations appear plausible also from Eqs. (6.16), (6.18) and (6.19),
but no rigorous proof could be found.
The phenomenon is critically dependent on the initial swirl of
the flow and o,i the downstream flow conditions. Generally very high swirl
values close to the theoretical maximum for nonoscillatory rotating flows
are needed to obtain a breakdown bubble.
The breakdown flows obtained fall into two groups. In case A
the bubble is flat, does not extend far upstream, and its maximum diameter
is never greater than the "disk" radius R0. Cases B through F, though
having a variety of different downstream boundary conditions, have almost
identical flows in the breakdown region, and their bubbles have similar
shapes. Where free bubbles exist for L = 2 they resemble those observed by Harvey (1962).
The similarity of cases B through F is explained by considering
the expressions for the coefficients Fn* in the series solution (Eqs. 6.20 - 6.25). The dominant terms in the expressions for F for these cases are all of the order while for case A Fn* is of order
[J1(i1)/i].
Thus a different type of flow is to be expected.Another type of flow is again to be expected when an eighth (or
higher) power relation in R is assumed for 1. For q' - R8 this would
lead to coefficients Ffl of the order and
This case has not been explored further since a variation of the axial
velocity with the sith power of R appeared as too drastic an assumption.
Given a flow with high swirl and initial rigid rotation, an
essential condition for the appearance of breakdown is then a retardation
of the axial velocity, or equivalently, an increase in the pressure on the
(By the properties of the solution one implies the other.) Breakdown may then appear at a considerable distance ahead of the disturbance.
Any one of the stream surfaces could be considered a boundary
for a different inviscid flow problem, say the flow in a swirl diffuser.
On the inner surfaces the pressure increases, while on the outermost
sur-faces it actually decreases due to the increased axial velocity, and an
almost constant swirl velocity.
The size and upstream extent of the bubble or bulge is dependent
on both the initial swirl and the shape of the downstream "bucket" in the
axial velocity. A dominance of higher powers of R in the expression for
the downstream stream function enlarges the bubble radially and in the
upstream direction. The effect of increasing swirl is similar.
In case F an almost rigid rotation is restored over the inner portion of the flow. Cases are conceivable where this region is even wider, with a swirl value high enough to cause a second breakdown.
An explanation of breakdown
The results indicate that vortex breakdown is a phenomenon
peculiar to supercritical rotating flows which have swirl values very close to the critical swirl value. In the case of initial rigid rotation the critical swirl value is j11/2 (1/2 of the first zero of the Bessel func-tion corresponding to 62.5°). Squire (1960) has given critical values for several different vortex flows.
The necessary conditions for breakdown are: a swirl value close
to the critical one, and a retardation of the axial velocity near the axis at some downstream station.
The phenomenon is a smooth reversible transition of the flow,
and it may or may not lead to stagnation and reversed flow on the axis. In the case of no stagnation a bulge may form in the stream surfaces.
Bubbles may be closed or open downstream and can be of the "egg" or
"carrot" variety.
Where stagnation occurs, the phenomenon can be viewed as a
centrifugal separation in the sense of Strscheletzky (1957).
Spiralling and transition to turbulence are secondary effects which can modify the appearance of the flow.
8 - VISCOUS SOLUTIONS Method of solution
In the preceding two sections we considered special solutions to the inviscid elliptic system (5.4):
= 0 r 3r az = z r
p3r
= 0 3z r =_i._p
p3Z
In section 5 the Navier-Stokes equations were shown to reduce to
this system when the stream surface angle of a rotationally symmetric vortex
flow with large swirl and large Reynolds number became appreciable, i.e., when tan y = 0(1). This condition definitely applies in the immediate
break-down neighborhood, as can be seen from Harvey's photographs (Fig. 2) and
from the plots of the inviscid solutions.
In the next two sections, solutions will be sought to the viscous parabolic system (5.3) = 0 r r 3z 2 r
pr
= 3z r 2r3r
2 r + + 1 3w-- p3Z
v[--This limiting form of the Navier-Stokes equations was shown to
apply in regicns of negligible stream surface angle of rotationally
symetric vortex flows with large swirl and large Reynolds number.
The parabolic nature of this system permits a step-by-step
integration by "marching" downstream, beginning with given initial values. This is a vast simplification over the (generally nonlinear) elliptic
boundary value problem, and it permits the calculation of vortex flows by methods analogous to those used in boundary layer theory. In this section,
an integral method for the computation of slender vortex flows will be
developed.
It is tempting to try to compute breakdown flows by using the parabolic system, especially since the elliptic (breakdown) region is usually of small extent, and the flow immediately outside of it remains parabolic in nature. Attempts to "march" downstream with a finite dif-ference scheme have nevertheless failed (Hall, 1966a). Gartshore (1962,1963) encountered difficulties with a Karman-Pohlhausen type integral method. Difficulties must be expected in the neighborhood of breakdown, stemming
from the neglect of the first two (convection) terms in the second equation
of the Navier-Stokes system (5.1) or the inviscid system (5.4) which make those systems elliptic. In the breakdown region, they have the same order of magnitude as the remaining terms of the centrifugal pressure balance in the viscous parabolic system (5.3).
The inviscid solutions have shown that upstream influence
in-creases drastically with swirl, becoming infinite and undamped when the
inviscid flow changes from the supercritical to the subcritical state.
expected that more severe problems will be encountered in swirling flows
when computation with the parabolic system is attempted.
The best hope of achieving meaningful results despite locally
singular behavior lies with integral methods, and recent applications of
integral methods have had some remarkable success in the computation of
separated flows where a similar difficulty exists at the separation point. (Lees and Reeves, 1964; Nielsen, Lynes, Goodwin, 1966; Holt, 1966.)
The viscous slender vortex problem was therefore formulated in
an integral method which is based on work by Ritz, Galerkin, Kantorovich, Dorodnitsyri and others (for a historical outline, see Bethel, 1966). A
description of the principle of the method can be found in Belotserkovskiy
and Chushkin (1962). For the solution of a system of partial differential
equations in two coordinates the method would consist in approximating the
dependent variables by members of a closed set of functions in one inde-pendent variable, say y, and depending on n parameters
a(z)
whichare functions of the second independent variable, say z. The system of
equations is multiplied by members of a linearly independent set of
weight-ing functions (the set must be closed) to generate a number n of linearly
independent equations equal to that of the unknown parameters, and inte-grated in the y-direction. This generates n ordinary differential
equa-tions which are solved for the n unknown parameters a1(z).
The method will be written in general form for an unspecified
order of approximation N. It will then be applied to vortex flows with
varying degrees of swirl, and specifically to flows where breakdown is to
parabolic system, partly from the mathematical formulation which leaves
room for future improvements.
Integral relations
The divergence form of the equations of motion for slender viscous incompressible rotationally symmetric vortex flow (Eq. 5.11) is
3W
+
(wK) + (HK-2Y + 2K) = 0
---
(W+P)
+ --- (HW-2Y ) = 0
The third and fourth equations are coupled via the second one. To reduce the number of equations, P is eliminated by cross
differenti-ation. The resulting system is
3W +
-
= 0 (8.1) - (WK) + -- (HK-2y + 2K) = 0 (8.2) 3W K2 2 +----(Hw_2Yfi)
= 0 (8.3) 3YIn the last two equations H is now considered known from the first
equa-tion by
Y
H =
-
f
dY0
assuming no sources or sinks on the axis.
=0
2 K 4y2
Integral relations are now obtained by multiplying Eq. (8.3) by
a weighting function Eq. (8.2) by a weighting function
g(Y).
and integrating with respect to Y from 0 to .The integrated equations are
d 3W
[()2
-f
f(Y)W2dY]
+ dZ $ 2 K2dY + (Hw-2Y00
o4Y
- [f(Y)(Hw-2Y .)]+
f
f(Y)HWdY-2[f(Y)YW]
00
0 + 2 'k (Y)Y + f(Y)]WdY = 0 (8.4) +- f
g(Y)HKdY
- 2[gk(Y)Y 3KY]
0+ 2[g(Y)YK] - 2
f
[g(Y)Y+2g(Y)]KdY + 2[g(Y)K]
= 0 (8.5)0 0 0
With the assumption that weighting functions and
g(Y)
can be found which will reduce the integrated terms to zero at the limitof integration, the two equations reduce to
-f
f(Y)W2dY
+ dY+ f
fj(Y)HWdY + 2f
[fj(Y)Y
4Y2 0 0+ f(Y)]WdY
= 0 (8.6) - 2 5K[2g(Y)+g(Y)Y]dY
-5g(Y)HKdY
= 0 (8.7)These equations constitute a system of ordinary differential equations for axial velocity W and circulation K. Approximating