Variable Mass PDF
Variable Mass PDF
Variable Mass PDF
R=19980210404 2018-10-02T15:19:50+00:00Z
N AC_,-
_coJ
.,
lh'-..3
DYNAMICS OF VARIABLE MASS
SYSTEMS _._ CU+ _ -
_ _ _"
Submitted by
ABSTRACT
This report presents the results of an investigation of the effects of mass loss on the
attitude behavior of spinning bodies in flight. The principal goal is to determine whether
there arc circumstances under which the motion of variable mass systems can become
unstable in the sense that their transverse angular velocities become unbounded.
Obviously, results from a study if this kind would find immediate application in the
The first part of this study feann_s a complete and mathematically rigorous
derivation of a set of equations that govern both the translational and rotational motions of
general variable mass systems. The remainder of the study is then devoted to the
mass loss scenarios on the dynamics of increasingly complex models of variable mass
systems.
It is found that mass loss can have a major impact on the dynamics of mechanical
nozzle geometry, combustion chamber geometry, propellant's initial shape, size and
relative mass, and propellant location can all have important influences on the system's
ii
ACKNOWLEDGMENT
Although there is only one official author listed on the cover of this document, the
material contained in this report is based on research work performed by several graduate
students who were supported under this NASA grant. Hence, I wish to acknowledge the
contributions of the following individuals: Dr. Song-Min Wang; Dr. Tai-Chien Mao,
whose Ph.D dissertation work was completely supported by this grant; Mr. Enrique
Cervantes, whose M.S thesis work was based on this topic and was supported by the
grant; Mrs Asayo Harvey, whose M.S thesis was partially supported by this grant; and Mr.
Jason Mudge, whose Ph.D research work was supported by this grant until the expiration
of the grant. Each of these individuals made important contributions to the final outcome of
Officer in charge of this project, Dr. Kajal Gupta of NASA Dryden, for his frequent and
invaluable advice, support and inspiration throughout the duration of the project. His
kindness has had a profound positive influence not only on the outcome of this work but
on my whole career. I also wish to thank Mr. Stevcn Ycc, the Grants Officer, for his
am most grateful to NASA for the generous support of my research effort through this
iii
TABLE OF CONTENTS
Abstract ii
Acknowledgments iii
Chapter 1 Introduction 1
1.1 Background 1
1.2 Early Studies of Variable Mass systems 2
1.3 The Star 48 Problem 4
1.4 This Work 5
iv
Chapter 5 Motion of Axisymmetric Bodies with Mass Loss 74
INTRODUCTION
1.1 Background
Studies of free and forced motions of spinning rigid bodies of various geometries
have been, and continue to be documented in great detail in the literature. Such studies
have led to the development of important scientific instruments (gyroscopes, etc.) and to
the concept of spin stabilization of modern spacecraft. By contrast, variable mass systems
have receivedrelatively
little
attention
in theliterature,
though they play an equally
engineeringliterature.
They includecomplex systems such as aircraft,
rockets,and moving
robotspicking up or letting
go of objects,as well as simplersystems such as water
variation and those with discrete mass variation. Rockets, for example, fall in the
continuous variable mass class; and, robots picking up or releasing objects, or a moving
vehicle dropping off some of its payload in discrete chunks, fall in the discrete variable
mass system class. Because systems with discrete mass variation can be analyzed using
It is clear fi'om intuition that when the net change in mass 0_ a system, as well as its
mass variation rate are small, it is unnecessary to account for the change in mass in the
study of the system's motion. For example, automobiles arc in fact variable mass systems,
yet, no one takes mass variation into account in handling and performance studies of
ordinary automobiles. The reason is that the rate of mass variation is viewed as negligible
and rightly so. On the other hand, a system that undergoes a substantial change in
mass, especially if this occurs in a short period of time, will definitely require that mass
variation be accounted for in the study of its motion; otherwise, any predicted response of
the system will be far removed from its true behavior. The focus of this study is on the
determination of the impact of mass variation on the dynamic behavior of systems with
Scientific study of variable mass systems has been in progress for more than two
hundred years; so developments in this field have a long though sporadic history. The
earliest recorded work on the dynamics of bodies with varying mass was performed in the
18th century by BemouUi (1738). He was then studying the forces acting on a liquid jet
actually derived what may be referred to as the equation of motion in this special case. The
Czech scientist and inventor, George von Buquoy (1781-1851), was the first to pose the
general problem of the dynamics of systems with varying mass. _ 1812, he obtained his
"motion formula" for such systems, and went on to solve a large number of examples
based on his formula, von Buquoy's work can be said to mark the birth of the theory of
the dynamics of systems with varying mass. In the mean time, William Moore worked out
his mathematical theory of rocket motion in England in 1813, and, in 1819, Poisson rook a
rathermodern approach and derived the equations of motion of variable mass systems
based on Lagrange's general formula. In their book published in 1856, Talt and Steele
included a section on mass variation. They postulated that mass variation produced small
continuous impacts or impulsive forces on systems, and thus resulted in continuous change
of velocity. This work was followed, several years later, by that of Meshchersldi, whose
work spanned the period 1897 to 1904. He essentially laid the foundation for the
his 160 page master's thesis to exploring a large array of issues relevant to variable mass
problems in the field. All of these early investigations of variable mass systems were
limited in one way. They were only concerned with the study of the translational motion of
such systems. The issue of rotational motion of such systems was not addressed until the
mid 1940's.
The second world war brought with it a resurgence of interest and activity in the
dynamics of variable mass systems, mostly in connection with rocketry. At this time,
wanslational motion of such systems was relatively well understood and the main focus of
research in variable mass dynamics began to shift to the attitude motion of such systems.
Some of the scientific giants of this new era include Rosser et al. (1947), Gantmacher and
Levin (1947), Rankin (1949), Ellis and McArthur (1959). The equations of rotational
motion derived by these investigators are quite similar, and have forms similar to Euler's
equations for rigid bodies, with extra terms that account for mass variability. Thomson
stands out as a major contributor to this field through his book (1961) and the companion
papers (1965, 1966). He derived several versions of the equations of motion of variable
mass systems and his work gives a great deal of physical insight into the behavior of
rocket, he showed that transverse rotation rate depends on the ratio43f the distance of the
system mass center from nozzle exit, to the transverse radius of gyration of the rocket. If
4
thisratio is greaterthan one (the most common case), the transverse angular velocity
decreases with time; and when the ratio is less than one, the transverse angular velocity
increases with time. Warner and Snyder (1968) brought some refinements to Thomson's
work and pointed out how various simplifying assumptions can lead to drastically different
motion predictions. Meirovitch (1970) moved work on variable mass systems one step
brought to light in the early 80's, when several space missions with upper stages powered
by the Star 48 solid rocket motor were observed to exhibit anomalous behavior.
Unexpected and unexplainable rapid growth in cone angle occurred near the end of the
motor bum. The output of a typical rate gyro mounted on one of such flights is shown in
Fig. 1. We note from this figure that the flight is uneventful until about two thirds into the
motor bum, when there begins an exponential growth in transverse rate, and thus in
nutation angle. The Star 48 is the first solid rocket motor known to produce such anomaly,
and it differs from its predecessors mainly in its much larger size and the existence of a
The Star 48 problem sparked another flurry of investigations [ Eke (1983), Meyer
(1983), Mingori and Yam (1986), Flandro et al. (1987), Cochran and Kang (1991)] into
the behavior of variable mass systems, and is one of the main factors that motivated this
work.
In his Ph. D dissertation work, part of which was supported by this project, Wang
(1993) [ see also Eke and Wang (1995)] modeled rocket type variable mass systems as a
simple cylinder of varying mass, and produced elaborate closed-form expressions that
describe the attitude motion of such systems for various burn geometries. The study
showed that certain propellant bum scenarios can actually cause the transverse rates of
short and large rocket systems to diverge in a manner similar to that observed on the Star
48 flights.
, 1 L.AI
'li llll
ii
ii
'II! i
0 10 20 40
_NE R_ 161¢11_ m¢
Fig. 1.1 Transverse Rate vs. Time for a Typical Star 48 Flight
This study is an extension of Wang's work. Like Wang's work (1993), this study
utilizes modern mathematical tools in the study of the dynamics of variable mass systems,
with the general intent of making valuable contributions to the field and enhancing current
Mostinvestigatorsthathaveattemptedthederivationof dynamicalequationsfor
variable mass systems have relied heavily on heuristics. In the next chapter, we present
complete and mathematically rigorous derivations of both the translational and rotational
equations of motion of variable mass systems. The remainder of the study applies the
systems, extracting and presenting a wealth of new information on the attitude dynamics of
such systems.
Throughout this document, equations and figures are numbered in the form (a.b),
where the first number a represents the chapter in which the item appears, and the second
number b is the actual item number within the chapter. To refer to an item, the number b is
used if the referencing is done in the same chapter in which the item appears; otherwise, the
CHAPTER 2
EQUATIONS OF MOTION
OF VARIABLE MASS SYSTEMS
This chapter begins with a description of the model used to characterize variable
mass systems in this study. This model is general enough to represent a wide variety of
physical systems that gain or lose mass while subjected to general three-dimensional
motion. The complete equations of motion for both rotational and translational motion of
such systems are derived using one of the methods of analytical dynamics - Kane's
formalism (Kane and Levinson 1985). The merit of this approach is its efficiency. It
produces the equations of transitional and rotational motion in one mathematically rigorous
step, and makes it possible to clarify a lot of conceptual issues in the derivation, that have
been very difficult to do in previous work. The equations of motion that are derived are
then compared with those obtained by Wang (1993) and others, who used the Newton-
Euler approach.
determined by the closed surface, B, and its contents. The contents of B at any given
instant can be solid (R), fluid (G), or a mixture of both. B and its contents undergo general
three-dimensional motion in space, and matter can flow continuously in and out of B
8
during this motion. For example, parts of R can "dissolve" into G by combustion or other
processes; and, some of such products of combustion can then flow across the boundary
B. At any given instant of time, only the surface B and whatever happens to be inside it at
the instant constitute "the system" for that instant. Thus, the system under consideration
here evolves continuously, both as regards its location in space, and its material
constitution - and hence its mass. We will use the symbol S to designate this system.
The derivation of equations of motion for a system such as the one described above
bodies of constant mass. The reason is that the basic principles of dynamics, such as
Ncwton-Euler equations and Lagrange's equations are only valid when applied to a definite
set of particles or rigid bodies. Two choices are then possible at this point. One, is to seek
or develop new formalisms that would be valid for variable mass systems; another solution
is to model variable mass systems in a way that allows them to be viewed as constant mass
systems, and thus make them amenable to treatment by existing principles of dynamics.
Before formulating the dynamical equations for the system of Fig. 1, we start by
temporarily restricting the system in some important ways. First, we assume that the closed
boundary, B, of the system maintains a constant shape, and thus encloses a region of
further help in the equation derivation process, we introduce the concept of constant mass
We consider once more the system as shown in Fig. 1, keeping in mind that the
outer shell is now of constant shape. At some instant of time, t1, there is a definite set of
material particles inside B. Let us assume that this set of particles is contained in a closed
elastic container, B1, that is identical to B at time tj. In fact, we take the viewpoint that BI
has always enclosed the exact particles that ended up in B at time tl, and that B_ will
continue to delimit these particles. Obviously, subsequent to time tj, the shape of B_ will
deviate from that of B if it is to continue to delimit the particles that were in B at time t_,
10
since some (or even all) of these particles may have exited B. Similarly, prior to time t_, the
shape of B 1 was quite different from that of B, since only some, or maybe none, of the
particles inside B1 then were also inside B. The shape of B_ is thus seen to vary with time,
becoming identical to, and containing the same amount of matter as B at time h- We note
however, that B_ and its contents maintain the same mass at all times. We shall represent 131
and its contents with the symbol S_, which will be referred to henceforth as the constant
mass system associated with the variable mass system S at time h. Similarly, we can def'me
$2, $3, etc., as constant mass systems associated with S at times h, tj, etc. Furthermore,
we assume that there exists a special subset of the particles of S that remains within B
throughout the interval of time of interest in this study. In fact, this set is further assumed
The equations of motion of any one of the constant mass systems described above
can be formulated using any of the classical approaches (Newton's Second Law,
Lagrange's Equations, Kane's Equations, etc.) since each system is of constant mass.
Suppose, for example, that Kane's formalism is applied to the constant mass system _ and
yields
k+ r k =0 (r = 1 .....
2, nk) (2.1)
where/7, is the generalized active force on the system, F,* is the generalized inertia force,
nk is the number of degrees of freedom of _, and the subscript k simply indicates that we
are dealing with the system _. Assuming that the motion of the fluid particles of _, relative
to the rigid part of this constant mass system is known, _ has six degrees of freedom
nk =6.
u,=fl, uI.....
thi
, ql, _/6, r=1,2 ....
,6 (2.2)
where t is time, q)and u) are, respectively, the r - th generalized coordinate and generalized
speed of the constant mass system ,_,. In other words, the time derivative of a given
S., and possibly time. Eqs. (2) can be supplemented with kinematical equations of the form
.°
Eqs. (2) and (3) constitute the equations of motion of _, and can be written as
Next, we imagine that the set of differential equations (4) is solved for y: as
functions of time. We note then that y_kl,._ are the generalized motii_n variables for the
variable mass system S at time t = t,. We now consider some quantity v e that characterizes
the motion of _k in some way, and is therefore obtainable from the generalized motion
variables y_t, or is simply one of them. An example of a good candidate for _ is the
magnitude of the velocity of the mass center of _. Another example is a component of the
angular velocity of the rigid body B that is a part of _. We will call _ a characteristic
motion variable for _k" Once y_ is known for all times, v k is known, and can be plotted as a
" 12
function of time. Now, suppose we solve Eqs. (4) for various values of k, and thus
determine Xt for several constant mass systems; and all of these V.t (k = 1, 2 ..... m) are
plotted as functions of time on the same scale, but staggered for clarity as shown in Fig. 2.
Now, we consider the motion characteristic, v, of the variable mass system S, that
corresponds to v k. If, for example, v_ represents the speed of the mass center of _, then v
would represent the speed of the mass center of S, keeping in mind that different material
particles make up S at different times. Because the dynamic behavior of S at some instant
of time tj is, in fact, the dynamic behavior at time t] of _., it is clear that v is given as a
function of time by the curve labeled F in Fig. 2; or, more precisely, the projection of F
onto the plane gz-t (see v in Fig. 2). The task before us can be viewed as the determination
of an efficient method for generating the differential equations whose solutions lead directly
to the curve v of Fig. 2. The route to accomplishing this task will now be delineated.
We consider once more Eqs. (4), which are the equations of motion for the constant
mass system _L,. First, we note that this equation has the same form for various values of
k. The only items that change with k are system parameters such as mass and moments of
inertia. Setting t = tk in Eqs. (4) yields a set of algebraic equations that produce y t/_._
(r=1,2,...,12); these quantities are equal respectively to y/tj), Y2(t,) .... y12(tj), the time
derivatives of the corresponding generalized motion variables for S evaluated at time t = tk.
One of these (or some function of these) represents the slope of cul"ve v of Fig. 2 at t = t_
and is plotted as point Pk in Fig. 3. The above process can be repeated for The system S t at
time tj, the system _ at time t2, etc., and the relevant results are used to complete the plot
of Fig. 3. The equations of curves such as P in Fig. 3 are the equations of motion of the
variable mass system. Since the points of such curves axe generated from the equations of
motion of the various constant mass systems, the equations of motion of the variable mass
system S, have exactly the same form as the equations of motion of a typical constam mass
system. However, to apply such equations correctly to variable mass systems, care must
be taken to interpret mass and inertia parameters correctly. At any given instant of time,
13
these parameters take on their values for the corresponding constant mass system for the
To obtain the dynamical equations of the full variable mass system, S, all we need
to do is to derive the dynamical equations for a typical constant mass system. We will now
p,
In order to keep the mathematical developments that are going to follow relatively
compact, and to avoid symbol definitions that are interspersed throughout the document,
we give and define, in Table 1, a set of symbols that will be used repeatedly in the
¢a.br(r =1,2,3)
(2.5)
u, /v .br_3(r 4, 5, 6)
15
A symbol in boldface type signifies a vector, and a boldface symbol with a tilde above it
represents a dyadic. Several of the symbols defined here are shown in Fig. 4.
• O - an arbitrary point of R
and
v =v°+co×r+Vr (2.8)
Because the motion of the fluid particles relative to R is assumed to be described by known
functions of time, the partial velocities [see Kane and Levinson (1985)] of P are given by
.o
(2.9)
VUr -- V_r + O)Ur × r
S _
From (11),
S°
VUr "- VUr + t.OUr × p (2.13)
S_fl_y,
a=aS*+otxp+t.ox(oxp)+2o3x(vr-vrS*)+(ar-_ *) (2.16)
where dm is the mass of matter inside the paraUelepiped. (13), (16) and (17) yield
= s*
(F;)p -dm[aS*+o_xp+co×(mxp)+2cox(v,-v,.S*)+(a,.-a; )].v_;
(2.18)
S*
-dm[aS*+(x×p+cox(coxp)+2mx(vr-vS*)+(a,-ar. )]" (COu,.× p )
Hence, the generalized inertia force on _ corresponding to the generalized speed, Ur, is
S* X (0_ X X + a r - ar dV (2.19)
-{our- p px +otxp+m p)+2m Vr-V r
We note here that the integrals in (19) are volume integrals that are taken over the region
enclosed by 1_. We recall that I_ always contains all the material particles found inside B
at time tk. I_ coincides with B at the instant t_, but is different from B at any other time.
We now return tO the determination of the generalized inertia force, F;. From
.,
Similarly,
f_k
_ k
s. (2.21)
and
(2.23)
(19) is reduced to
(2.24)
(2.25)
and,for r = 4, 5, and 6,
The force per unit volume, F P, acting on the generic particle P of the system can be
written as
where F_ comes from forces external to the system, and F/t' is from the internal forces
acting on P. The generalized active force, Fr, for the system has the form
- 22
(2.28)
Now, if the internal force exerted on any particle P of the system _._ by another particle Q
and
(2.30)
s*.
Fr = Vu r F + C0Ur" M (2.31)
where F is the resultant external force on the system, and M is the sum of the moments of
all the external forces on the system about the system mass center. (5) and (31) then give
23
and
$* .
Fr = Vur F for r = 4, 5, 6 (2.33)
Recalling that Kane's equations have the form given in (1), we now combine (25)
(2.34)
Similarly, the equation of translational motion comes from (26) and (33), and is
m a s* = F
(2.35)
At this point, there are two main obstacles that prevent (34) and (35) from being
useful as given above. First, the motion of fluid particles within I_ was assumed known.
This means that both the velocity and acceleration fields within I_ are known as functions
of time. In reality, none of these functions is known, but reasonable guesses can be made
24
for the velocity field within B; that is, the velocity distribution inside ]_ for the instant t=
t_, when the constant mass system coincides with the variable mass system S. On the other
hand, we have no handle over the velocity distribution of any particles outside B; and some
such particles would normally he within ]k at instants of time different from tt. The other
problem is that we have no way of estimating the acceleration field - not even within B.
To circumvent these problems, two imlx_rtant measures are taken. One is that
attempts are made to convert all accelerations that appear explicitly in (34) and (35) to time
derivatives of velocity. If done properly, this would take care of the second problem
above. The other measure is that ways are found to convert any volume integral that
contains a velocity term, and is taken over the region 1_ to a volume integral over B, where
velocities can be estimated. This measure would resolve the first problem. We now show
In
Io
a +axr
. +cox
(mxr" xv_*+a_*
]-F (2.36)
m
(2.37)
,.
We then re-write the last term on the left hand side of (37) as
25
Rdvr Rd
(2.38)
LkP ardV = fBk P --_" dV =-_fBk P VrdV
where the left superscript, R, indicates the reference frame in which the time derivative is
taken. Next, we substitute (38) into (37) and evaluate the new expression at time tk, when
Because the last term on the left hand side of (39) contains a time derivative outside the
integral sign, the limit of the integral cannot he changed to B as was done for the previous
tenn. We are thus faced with the problem of taking the integral of an expression containing
a velocity term over a region in which the velocity field is not known - the trn'st problem
..
discussed above. Fortunately, this dilemma can be resolved by means of the Reynolds
i_I_d-tfBP "rdV ] = _-
[RdfB k p Yr dV + fs p Yr (V, " !11)d_ ] (2.40)
k t = tk t = tk
where S is the surface area of B, and n is a unit vector normal to B and pointing outwards.
t=tk
0('x')"
l t=tk
(2.41)
+ p + (,,..)_ =E(,_
Jt = tk t = tk
where m is the mass of_ at ILrne t_; that is, the mass of S at time t_. Because (41) has the
same form for all values of tk, the equation of translational motion for the full variable mass
system is
(2.42)
The equation of rotational motion for the variable mass system is obtained in a
(2.43)
" 27
,o
t=t k
I I
t= tlc
(2.44)
Thus, the equation of rotational motion becomes [from (34) & (44)]
(2.45)
where the inertia dyadic that appears in (45) is that of matter that is within B at the instant
under consideration.
The vector equations of motion, (42) and (45), obtained through the application of
Kane's equations, are identical to those obtained by other authors [ Meirovitch (1970),
Another very useful form of the vector equation of rotational motion can be
constant mass system as given in Eq. (34), and consider the expression
Rd Rd
(2.46)
A-- -_ife k [p × (W x p)] din= ---_fBk p [p ×(_× p)] dV
(2.47)
The last team on the fight hand side of (47) can be added and subtracted from (47), then by
Iv,x (co
x p)]-[p x(co
x vr)]= [cox (v,x p?] (2.48)
we have
29
(2.49)
k k
On the other hand, A, as given in (47), can be evaluated at t = t_ and expanded using
,,
(A_,_
"'f,,[p×_=×p)],,,,,
=_ +J',,,[p×_= _/
×p)](v,.n)
t= tk
'"(T.,,,)
+J's"[p×
_°×
p)]
(v,.n),_}
It = tk
(2.50)
[(Rail ¢o+i.ot+fsp[p×(mxp)](v,.n)aS }
=[I-dT ) "='k
dV -((Rd]).ei+fsp[px(o_xp)](Vr'n)dS+fakp[o_x(p x Vr)]dV/
ffBk' P P _a)X Vr)It = tk" dt
(2.51)
3O
Eqs. (34) and (51) lead to the following alternate form of the vector equation of attitude
motion:
(2.52)
Rd
.b
31
CHAPTER 3
The equations of motion derived in chapter 2 [see (2.42), (2.45) and (2.52)] are
quite complex; and very little progress can be made with these equations without some
simplifying assumptions to reduce their complexity. In this chapter, these equations are re-
examined. The significance of each of the terms is evaluated, and, the circumstances
under which some of these terms can be dro_ are explored. The goal is to arrive at a set
of equations that capture the salient features of the system under study, and that are
(3.1)
" 32
m[.°oxr'..x(.x
r')
1 (3.2)
where
(3.3)
Fc=-2 fB P(mXvv)dV
ed
(3.4)
rL=- _?fBpv,aV
and
(3.5)
FT=- fsPVr(V,.'n)dS
of translational motion for a rigid body. Thus, mass variation appears to augment the
the literature as the Coriolis force, since it derives from the Coriolis component of the
acceleration. FL is the rate at which the system's linear momentum relative to B decreases
with time because of particle motion inside B. FT represents the rate at which relative linear
momentum is lost across the boundary B, and is often referred to as the thrust vector in
rocket applications.
If those particles of the system that can move relative to B are allowed to move
within B but do not cross the boundary B, then F T becomes zero. F T would be non-zero
but negligible if either a very small percentage of the system's particles is allowed to cross
33
B, or those particles that cross B do so at very slow rate. In other words, the thrust vector
can be neglected whenever the amount of matter that is lost or gained per unit time is small.
What matters is the rate, not the total amount of matter lost or gained.
Note that we can use Reynold's transport theorem to expand the Coriolis force into
prdV
pv, dV=- 2 mx_
(3.6)
where
and
Fc2 =- 2 ¢o x _ p r (Vr.n) dS
(3.8)
There arc situations where Yr Can be considered negligible inside the boundary B
but not at an exit from B. An example is an inflated balloon with a hole. Gas motion
inside the balloon is hardly noticeable while gas exit velocity at the hole is substantial. In
cases like this, Ft. and the first term of F c (Fc_)are negligible, but Fc_ as well as the thrust
vector survive. As a matter of fact, this is not an unreasonable assumption for praclical
systems such as rockets, and can provide a way of rendering the equations of such systems
tractable, since the details of internal gas flow can be neglected. Even if v, is not negligible
"" 34
within B, but the particles within B can attain some type of steady state in their motion
relative to B, then F L and Fcj are once more negligible. After ignition, rocket systems
quickly attain an approximate steady state, and so, F L and Fcl can be ignored for such
systems.
In summary, the vector equation of translational motion for variable mass systems
m
a°+ ot x r* + m x (a) x r')] = F + FT +Fc2 (3.9)
where F is the external force, F r is the thrust vector, and Fc2 is part of the Coriolis force
and is given by (8). F c and F r can be dropped completely only when the rate of mass loss
is negligible.
Two forms of the vector equation of attitude motion were developed in chapter 2
(3.10)
(3.11)
Both of these equations reduce to rigid body rotational equations if Vr = 0 within B as well
where
(3.13)
Rd
(3.14)
MH = --_fB p(p x vr) dV
represents the rate of decrease of the system's relative angular momentum inside B, and
36
M,--fsp(pxq(v,.) s (3.15)
is the rate of loss of relative angular momentum across the bounda_, and is also equal to
the moment of the thrust vector about the system mass center.
In those situations where it is reasonable to assume that the motion of the fluid
phase relative to the solid phase has axial symmetry, and when there is no whirling motion,
we have that
This is so because, for every particle P of the system with position vector p and relative
velocity Vr, there exists another particle P' of position vector p' and relative velocity V'r
such that the vectors pXVr and p'Xv'rhave the same magnitude but oPtx_site directions.
And, since axisymmetry also implies that the immediate neighborhoods of P and P' have
the same mass density, Eq. (3.16) follows. This equation immediately leads to
M. = M r= 0 (3.17)
We arrive at the same conclusion for systems where Vr can be considered negligible inside
the boundary B but not at an exit from B, such as rockets. M x is also negligible for steady
kp mx x (3.18)
The last term on the right hand side of (18) vanishes for axisymmetric motion as well as for
negligible internal flow. The second term on the right hand side of (18) is often referred to
as the jet damping moment, because it has been shown [see Dryer (1963)] to have
attenuating effects on the angular rates in some types of rocket systems. To evaluate this
surface integral, it is necessary to know the geometric shape of those parts of the boundary
where particles are allowed to exit or enter the system, as well as the velocity profile at
these locations. This moment is thus very much dependent on the system's geometry. In
the case of rockets, for example, the longitudinal dimension of the combustion chamber
(for solid rockets) as well as the exit nozzle radius have much to do with the impact of this
The first term on the right hand side of (18) captures the contribution of inertia
variation. We thus conclude that jet damping moment and the moment due to the changes
in inertia properties have dominant effects on the attitude dynamics of the system. The
relative importance and the interaction between these two moments, essentially determine the
character of the attitude dynamics of variable mass systems. There is advantage in using
the reduced form of (11) for the study of rotational motion. If the assumptions of
symmetric and/or negligible internal motion are made, the most "troublesome" terms will
have dropped out. The only term in this equation that would contain Vr is a surface integral
over B. Vr'n is zero everywhere on the surface of B except at those places where fluid
particles can enter or leave the region delimited by B (the nozzle exit plane in the case of
38
rockets). In general, Vr.n can be approximated relatively well at ..these locations; hence the
In the study of the dynamic behavior of variable mass mechanical systems, the most
important forces appear to be the thrust vector and the Coriolis force. The jet damping
moment and the moment due to inertia variation are the dominant moments. These
quantities should be included in any dynamic studies of variable mass systems to ensure
CHAPTER 4
ATTITUDE MOTIONS
OF A VARIABLE MASS CYLINDER
4.1 Introduction
In this chapter, we begin to narrow the scope of our study to an important and
useful class of variable mass systems - space rockets. We start this process with an in-
depth study of the attitude motions of a variable mass system that is initially a solid right
circular cylinder, which loses its mass continuously through combustion as it moves
around in space. The cylinder problem is a very important and useful problem for several
reasons. First, it models rocket-type systems in a simple enough way that the equations of
motion become relatively tractable. Generally, space vehicles are designed to be more or
approximation for such systems. Hence, a thorough study of the variable mass cylinder
studies of rocket-type systems, and can lead to great insight into the dynamic behavior of
The idea of using a cylinder to study the a_tude behavior of variable mass systems
was originated by Eke and Wang (1995). In their work, they solved completely in closed
then extracted useful qualitative as wcU as quantitative information about the attitude
behaviorof thecylinderfi'omtheanalytical
expressionsof thesolutionsof theequationsof
motion. The approach here is different. The strategy here is to develop an analytical
method that yields a wealth of fundamental information about the behavior of the system
without actually solving the equations of motion. The main motiv..ation for this approach is
that such a strategy can be applied to more complicated models of variable mass systems
Consider a variable mass system of the rocket type that is initially a right circular
cylinder as shown in dotted lines in Fig. 1. This system is given an initial angular velocity
0%, and is then allowed to move freely in a torque-free environment as it loses its mass
and quantitative studies of the attitude motions of such a system. To begin, we define as
..
control region, the space enclosed in dotted lines in Fig. 1. Matter within this region at any
The attitude motions of such a system are governed by Eq. (2.52), which can be
burning process proceeds in such a way that the solid unburned part of the cylinder remains
symmetrical about the original cylinder axis at all times; (b) the fluid products of
combustion move in an axisymmetric manner relative to the solid unburned part of the
41
cylinder;, (c) whirling motion of the products of combustion relative m the solid portion of
M=I'a+_xI'_+ (_/
t dt ]'°_+ fBP [. C°x4/-n)_s
x
(4.1)
. :,,_:_ii;i:i:i-::!S!7:!:!:!-!.".
....................................
;::"...."'-.
::::::::::::::::::::::::::::
_ ......................-.....................
----i
• /
The axisymmetry assumption allows us to express the instantaneous central inertia dyadic
of the system as
.°
where bi, b2, I)3 are unit vectors fixed in B and directed as shown in Fig. 1; and I and I,
are respectively the transverse and axial central inertia scalars of the system. If the inertial
(4.5)
( _t i ) " ca = i ( o_ l + o_b 2) + Jz _b 3
(4.7)
In order to evaluate the last term of (1), an assumption must be made concerning the
velocity field of the fluid particles of the system at the surface of the control region. Here,
we assume that the vector Vr is zero everywhere on B except at one of the ends of the
vr " n = u(r)
(4.8)
u(r)
= u = constant (4.9)
then
where m is the instantaneous mass of the system, and l(t)is the distance of the mass center
from the right face of the shell, B. By substituting (5), (6), (7) and (10) into (1) and
setting M to zero, we obtain the scalar equations that describe the torque-free attitude
(4.11)
(4.12)
,,,,(,, (4.13)
Multiplying (11) by cox, (12) by (0y and adding the two resulting equations leads to
+ 0_y)+[ i-
I I &_-(to2x 6a (£ 2 + R 2/4) ](0)2 + 0_y)=0 (4.14)
44
We nowlet
= i- rh(l2 + R 2 / 4) (4.16)
and
do)z
I_--_-+ 0 _ = 0 (4.18)
and
If the functions 0(t) and ¢(t) are known, (18) and (19) can serve as basis for the
determination of the essential features of the rotational motion of the variable mass cylinder.
We note that these equations are uncoupled, so that the spin rate has no effect on the
wansvcrse angular rates. We also note that ff O(t) = O, then the spin rate remains constant at
its initial value throughout. Similarly, if ¢(t) = 0, the magnitude of the transverse angular
.--.o.xpE-
I (4.20)
45
or
(4.22)
c0xy = o)xy 0 exp
where foxy = _ is the magnitude of the transverse angular velocity, and the subscript 0
indicates initial value; that is, value at ignition. We shall use the subscript .o to indicate
values at burnout.
Looking back at (15) and (16), we observe that each of the functions 0 and q_ is
- _1 + _2 (4.23)
where
_1 = l (4.24)
and
= - th (t 2 + R 2 / 4) (4.25)
Because we are concerned here with a situation where mass is being lost, both 1[? and rh
are negative. Hence, _1 remains negative throughout the burn while _ is positive
system transverse inertia, and because it is always negative, it will tend to cause ¢Oxy to
diverge [see (22)]. On the other hand, 02 is, in this case, the so-called jet damping term,
and it does in fact attenuate ¢0xy by virtue of the fact that it is always positive. Although 0t
is determined by the bum geometry of the cylinder, 02 depends on the size of the cylinder
as well as the velocity distribution across its exit plane. All the statements made above
about O(t) apply equally to O(t) since (15) and (16) have the same form, and (20) is similar
to (22).
..
of the system from a detailed examination of the functions 0(t) and 0(t). From (20) and
(22), it is clear that for 0(t) and (_t) > 0, both oh and oxy approach zero from any initial
conditions. On the other hand, for 0(t) and _(t) < 0, both oh and taxy diverge. Profound
changes in angular velocity can occur when variations are made in 0(t) and _(t).
To make further progress with our study of the attitude hehavior of the variable
mass cylinder, we will now proceed with a close study of the functions 0(t) and _(t).
These functions require that the system mass, inertia scalars and some geometric properties
be known as functions of time; and these parameters can be determined if the burn scenario
is known. We will therefore consider, as did Eke and Wang(1995), four burn scenarios :
the uniform burn, the end burn, the radial or centrifugal burn, and the centripetal or anti-
radial burn. In uniform burn, the assumption is that the cylinder is ignited simultaneously
everywhere inside it at time to. It then burns in the same uniform manner everywhere, so
that the external dimensions remain the same throughout, but the density of matter within
the cylinder decreases uniformly in the same manner everywhere within the cylinder. The
density is thus the same function of time everywhere inside the cylinder, and the products
of combustion are expelled at one end of the cylinder. An end burning cylinder burns from
one of its ends to the other, in such a way that the intermediate shape of the system is
always a cylinder of the same radius but decreasing length. Once more, the products of
combustion are expelled from one of the ends - the burning end. In radial burn, the axis of
° 47
the cylinder is ignited at time to, and the system then burns radially outwards in such a way
of decreasing radius.
-I- -i U
and
• 48
Iz = m R2/2 (4.27)
nencc
(4.28)
and
0(t) = 0 (4.30)
at all times, and so, the spin rate, ok, will remain constant at its initial value.
From (16) and (28), and recalling that t in (16) is the same as h in Fig. 2 for this case,
Because ¢(0 > 0 at all times during the burn, the magnitude of the transverse angular
Fig. 3 shows the dimensions and intermediate shape of a cylinder in end bum. The
properties, including mass, moments of inertia and the location of center of mass. The
m=2p_R2z (4.32)
fia=2p_R22 (4.33)
(4.35)
Iz = m R2/2 (4.36)
iz = rh R2/2 (4.37)
L h
U
•---- 2 h-z----_
..
Furthermore,
50
= 0 (4.40)
rh=-pgR2u<0 (4.42)
Based on the assumption of uniform velocity profile at the exit pla,l.e, we conclude that th is
constant dtLring the time interval [ to, t**]. We also know that _r is negative in this interval.
betweenignition and burnout. Thus, ¢(t) increases monotonically from 0 to the positive
value given in (41). We have a situation where q)(t) has the form shown in Fig. 4.
¢(t)
_t
We conclude from (40), (41) and (43) that the transverse angular rate, c0xy, stays
approximately constant during the initial phase of the burn, then decreases exponentially to
zero.
As for the spin rate, (15) and (37) indicate that 0(t) = 0 forall t, so that the spin rate
is again constant.
Having established that the transverse angular velocity of an end burning cylinder is
always damped, our next task is to examine the effect of the cylinder's initial geometric
It is clear from (22) that this factor actually governs the speed with which the transverse
angular speed approaches zero from any initial value. The larger the magnitude of _P, the
_p=_=fia(4hz-4h 2) (4.45)
I m (R2[4 + z2/3 )
To compare the behavior of cylinders of the same length but different initial external radii,
we allow R to vary while keeping z, h and u invariant. A close inspection of (46) indicates
that increasing R/h results in smaller values for _F. Hence, the transverse angular velocity
magnitude, 0Jxy, converges more slowly as the ratio R/h is increased. In other words, the
transverse motion of prolate cylinders in end bum is highly dahaped while that of oblate
cylinder is only very slightly damped. In the limiting case, as R/h --_ **, _F = 0, the
then that the amplitude of lateral oscillation of a variable mass cylinder in end bum can be
influenced by the choice of its initial shape. Fig. 5 is obtained by numerical integration of
We recall that in radial or centrifugal burn, the cylinder starts burning on its axis,
and the burn propagates radially outwards with time. The intermediate configuration of
unburned propellant is a hollow cylinder, as shown in Fig. 6, and the mass and inertia
0.9
0.8
0.2
R/h_
0.1
0
0
[. h ,
I- h "I u
..........
m ..............
B ...........
m .............
'!........
.°
P
CV
Hence
rh=-4phnrf (4.50)
(4.51)
and
i7 =m r2 (4.52)
between ignition and burnout. Because r varies from 0 to R in the time interval [ 0, t**] and
from the physics of the problem, it is clear that f" > 0 in the given interval.
We thus have a situation where 0(t) has the form shown in Fig. 7
o(t)
.°
The switchover point, t*, at which the sign of 0(t) changes from positive to negative is
This implies that the spin rate decreases in the first stage of the bum, when r / R < 0.707,
but increases exponentially in the later phase of the burn when r / R > 0.707; a fact that is
clearly support_ by the plot shown in Fig. 8. This plot is obtained from numerical
0.8
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 1
r/R
.?
We now discuss the evolution of the magnitude of the transverse angular velocity,
foxy.
_(t)=Z-rh(12+g2/4)=fnR2[(1/2)(r/R)2-(2/3)(h/R)2-1/4] (4.58)
Hence
and
From (58),
The sign of 0-* depends on what can be described as the cylinder's "shape factor"- the ratio
of the initial value of its diameter to its length. Hence, whether _xy is bounded or
unbounded depends on this shape factor. To determine the value of R / h that separates the
"stable" region of _xy from the "unstable" region, we set 0(t) to zero in (60), and obtain
O(t)
1.63 I
_ t
Hence, the magnitude of the u'ansversc angular velocity of a cylinder in radial burn always
decreases with time in the early phases of the burn. If the ratio of the unburned cylinder's
diameter to its length (shape factor) is less than that of 1.63, the decrease of _xy continues
all the way to burnout. On the other hand, if the cylinder's shape factor is greater than
1.63, then there comes a time during the burn when o)xy levels off, and then starts
increasing exponentially through burnout. The initial shape of'the cylinder thus has a
critical effect on the cylinder's lateral attitude motion for radial burn. Pencil-shaped
cylinders will tend to be stable in radial burn while hamburger-shaped cylinders will tend to
cylinders.
> 1.63. This can be done by determining the value of r/R at which the sign of 0(0
YR[,=0 V3 R, +I
2 (4.63)
So, the onset of instability for O)xy depends again on the shape factor. Once a value is
specified for the shape factor (R / h), the value of r / R at the beginning of instability can be
R_ 5 r = 0.744 (4.64)
2 I =o 0.913
0.707
59
We conclude, from (64), that (Oxy begins to diverge much earlier for cylinders with higher
values of the shape factor (fat and short cylinders) than for cylinders with lower values of
R / h. Numerically obtained plots of the transverse angular velocity are shown in Fig. 10,
and are totally consistent with the above inferences. Furthermore, the above analyses
captttre a great deal more of the vital characteristics of radial burn than do these plots and all
unstable region
0.4
Gable region
0.2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r/R
In centripetal bum, the cylinder bums radially inwards without changing its length.
The intermediate shape is a solid cylinder of diminishing radius, as shown in Fig. 11. In
rn = 2 p h _ r 2 (4.65)
I_, = m r 2 / 2 (4.66)
and
I=m(r2/4+h2/3) (4.67)
!!iiiiii!iiii !i_iiii!i!iiiiiiiiii!iiiiii!iii:iiil
i!i
i:i!
il
iiiii__ii!i!_ii!iiiiiiii!ii_
ii!__iiiii
___!
!!!!
!i!!!!!!
!!!!
!
_h=4phnrf (4.68)
l, = rh rz (4.69)
and
00 = rh R 2 / 2 < 0 (4.72)
e. =-_h(R2/2)>0 (4.73)
since f- < 0 in this case. We thus have a situation where 0(t)has the form shown in Fig.
12
62
0(t)
.°
Hence the spin ram increases h-tidally, attains a peak value then de,creases rapidly with the
burn. The time of attaining the peak value of the spin rate corresponds to 0(t) = 0. That is
1.6 ! , ! ! ! ! ,
1.4
3
_q
1.2
0.8
iiii iiiii!!iiill
iiiiiiiiiiiiiii
iiiiiiiiil
.....
3 0.6
0.4
0.2
.................................
..............................
.........
...............
.......
!i.........
iii
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1 -r/R
From (77), we determine that t_o = 0 when R / h = 1.63, _P0> 0 for R / h < 1.63, and
_0 < 0 for R / h > 1.63. We conclude from all of this that (a) as long as the ratio of the
cylinder's initial diameter to its length is less than 1.63, the transverse angular velocity
magnitude decreases as the bum progress; (b) if the ratio R / h is greater than 1.63, the
magnitude of the transverse angular velocity increases initially until it reaches a peak value,
then it decreases with time for the reminder of the bum. Thus foxy never really becomes
unbounded for centripetal bum. These facts are demonstrated below in Fig. 14.
Our findings so far indicate that the stability of rotational motion of a variable mass
cylinder depends on two factors : the bum pattern and the cylinder's shape factor. Both the
spin rate and transverse angular velocity are bounded for all shape factors when the
cylinder is subjected to uniform, end, or centripetal bum. For radial or centrifugal bum, the
spin rate is always unbounded, while the transverse angular speed is bounded for the shape
64
factor (R / h) that are less then |.63, and unbounded for shape factors that are greater than
1.63.
All the results obtained in this section arc totally consistent with, and augment
...........
................................
...........................
.................
....
.............................
......................
°..............
oli................
0
.......................................
0. I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
1 -r/P,
Up to this point, all variables and parameters of interest in this study have been
out to be a very astute and efficient analytical strategy. Through this choice, we have been
able to avoid some of the myriad problems encountered by previous investigators of the
dynamics of variable mass systems. For example, practicaUy all previous investigators
65
worked directly with time as independent variable. Hence, to make progress with the
equations of motion [ for example (11)-(13)], it was necessary to specify the time
dependence of the mass (m), the location of center of mass (t) and moments of inertia (
m = m0- a t, and made "reasonable" guesses for the coefficients. They made similar
guesses for l, I and Iz. The problem is that these inertia quantities are interrelated; a fact
that is generally not correctly reflected in the guesses for the independent time functions
assigned to these quantities. All such difficulty is avoided by the clean analytical scheme of
Although avoiding the explicit use of time in our analysis so far has been quite
beneficial from analytical point of view, it is in fact still desirable to know how the angular
rates and even the inertia properties vary with time. For example, this enables us to
compare analytically predicted trends with experimental results, obtained through on-board
instrumentation. In this section, we show how some of the results obtained so far can be
We now reexamine some of the results obtained for the end burning cylinder.
_: =-u/2 (4.80)
or
66
(4.81)
which leads to
z = - (u / 2) t + h (4.82)
The inertia properties can now be converted to time functions simply by substituting (82)
where
mo = 2 p _ R 2 h (4.84)
where
(4.86)
P= rh(_+ h 2) (4.87)
and
(4.89)
S =1-_-2 rhu2
L.=2h/u (4.90)
which is independent of the radius of the cylinder. The instantaneous transverse central
I=mk 2 (4.91)
k2 R_ + hZ_hu
2 2 (4.92)
Since
(4.93)
For the time interval 0 < t < 2 h / u, k 2 is a decreasing quadratic function of time.
At burnout
68
k_ = R2/4 (4.95)
and the axial radius of gyration is a constant. Since fia and u am constants, we see that the
transverse moment of inertia is a cubic function of time while both the axial inertia and the
mass are linear functions of time for the end bum assumption.
rn =- 2 p h _ d(r2) - (4.97)
dt p_R2u
Hence
=R2u
d(:) (4.98)
dt 2h
r2=R2ut (4.99)
9h
69
Hence,
i=mlRZ+r2 +h21
4 3/ =(mo+fiat}(4R2+8R--_h t+h2 3 }
(4.101)
=Io+Bt+At 2
where
Since both A and B are negative, it is clear from (101) that I is a decreasing quadratic
i=B+2At<0 (4.105)
From (49)
70
I - ml ÷r l- mo÷m0tR2
+R2ut )
(4.106)
=Izo+2At 2
and
IT=4At<O (4.107)
where
Izo = ½ mo R 2 (4.108)
and
where
(4.111)
and
(4.112)
71
Now, the value of burnout time t.. is obtained by setting r = R in (99). Hence
(4.113)
and
(4.114)
3
k_=R 2 (4.115)
r2 = R 2- R2u t (4.116)
7h
2.h
t**---- (4.117 )
U
I=Io+Ct-At 2 (4.119)
with
(4.120)
72
I=C-2At (4.121)
where
D=rhR 2 (4.123)
and
Both k 2 and k2 are decreasing linear functions of time. We also have that
l_ = h2/3 (4.126)
and
k_** = 0 (4.127)
In conclusion, we observe that for the three cases examined, the cylinder mass is a
linear function of time and the bum duration is the same for all cases. For the end burning
centrifugal and centripetal bums, I is a quadratic while k 2 is linear in time. All of the inertia
73
parametersare decreasing functions of time except for centrifugal bum where the radius of
CHAPTER 5
In the previous chapter, the important problem of the attitude stability of rocket
systems is initiated, with such systems idealized by a right circular cylinder. In particular,
the effects of mass loss for various burning patterns are emphasized.
In this chapter, another step in the study of the attitude motions of rockets is taken
investigated here differs from the cylinder model in two basic ways : (i) the system is
modeled as a general axisymmetric system ( not just a cylinder), and includes a cylindrical
solid propellant grain as well as a constant mass, non-propellant portion; (ii) a nozzle
arrangement is also included. The new system can be reduced to the cylinder model by
The objectives in this chapter are similar to those of the previous chapter;, we are
interested in the angular rates of the system, and how they are influenced by mass loss.
75
//////
The system under consideration in this chapter is shown in Fig. 1, and consists of
two main parts, S and F. F represents the solid propellant of a typical rocket system. It is
assumed to be rigid and cylindrical. S is the main part of the system excluding fuel and
includes the payload. S is assumed to be rigid and axisymmetric, and maintains the same
mass throughout. Both S and F have the same axis of symmetry, so that the mass canters
S* of S, F* of F and B* of the combined system, all lie on the common axis of symmetry
z. The system loses mass continuously through a planar portion of its surface that we shall
refer to as the nozzle exit plane, and that is located at the right end of the system in Fig. 1.
At any given instant, the overall system will consist of everything that is bounded by the
represents the mass of the invariant part, S, and mf is the instantaneous mass of the fuel, F.
The origin of the coordinate axes x, y, z is located at, and moves with, the composite mass
76
center B*, such that the z-axis is always aligned with the symmetry axis, and the x and y
axisymmetric configuration throughout its bum, so that the central inertia matrices for S
and F are diagonal. Externally applied moments, thrust misalignment effects, aerodynamic
burning, and are thus neglected. The system is thus assumed to move in a torque-free
environment. The motion of the particles of the products of combustion as they cross the
exit plane is taken to be such that the velocity distribution perpendicular to the plane is
uniform. If it is further assumed that the motion of gas particles inside the combustion
chamber is symmetric relative to the z-axis, and that whirling motion of the gas particles
within the system is negligible, the vector equation of attitude motion for the system is
Because of the symmetry assumption, the central inertia dyadics for F, S, and the
and
77
I =I,, + If + M_ b 2 + mf a 2 (5.5)
and
J = Js + Jf (5.6)
Note also that the distances a and b of Fig. 1 are given respectively by
(5.7)
b = mf C (5.8)
(M, + mr)
As propellant mass is lost, the system mass center B* moves forward toward S*.
I =Is+ If + _t c 2 (5.9)
where
78
mf Ms
(5.10)
theinertial
angularacceleration
is
a = coxbx + _y by + Ct)z
bz (5.12)
(4)and(12)thengive
.?
where _e is the axial distance between F* and the nozzle exit plane. By combining (1) and
(13)-(16), we have that the scalar equations of attitude motion for the system are
_,-(J-_
)__ +[i-mt(
_:
+_/4
)].=0 (5.18)
(5.19)
where I and J are given by (9) and (6) respectively, and [see (9)]
80
i=if+_tc2+2gc_: (5.20)
j = .if (5.21)
In this section, we render the equations of attitude motion very compact through the
studies of these equations in a way that provides great insight into the most important
factors that govern the behavior of the system. We start by introducing the non-
dimensional time. For the system of Fig. 1, the instantaneous mass.of the cylindrical
propellant is
mf = mr0 - m t (5.22)
where
= - rh (5.23)
81
tb = mr0/_ . (5.24)
Thus, z= 0 at ignition, and x = 1 at burnout. Also, the quantity ot l has the dimension of
= ot ddx (5.26)
r /A
where the prime indicates derivative with respect to "1:, and the inertia properties I, J and
In the next phase of the non-dimensionalization process, we use mr0 and R as scale
factors for mass and length respectively, so that the scaring factor for inertia scalars
dimensional parameters :
where ks and k,z are, respectively the transverse and axial radius of gyration of S.
We now divide the numerator and denominator of the right hand side of (27) by
mm R 2 to obtain
83
_ ____/_ I +
O_= mr°R2 m_R2taY°z- mr°R2 mr°R2 (5.31)
I 0_
mfo R 2
Similarly,
(5.33)
and
-'
(_' m'
-meo T _z (5.34)
%=
84
where COx,Oy, O)z are non-dimensionalized angular velocity components, and I, J are
,,
(5.35)
and
o(_)= F_-m- _2
mfo 2 (5.36)
m E
_,r (5.37)
(34) becomes
d_z=- _. _ (5.38)
85
n i
and
(fo
I J
(5.40)
where
As in the case of the cylinder, the sign of the functions _ and 0 determine whether
the angular rates grow or decay with time. A positive sign implies decay and a negative
From Fig. 1
We recall here that p is the density of the fluid products of combustion at the exit plane.
For a uniform velocity profile at the exit plane, u(t) is constant, so that
"I
(5.45)
R 11
(5.46)
" 87
To make further progress with our analysis, we will now focus attention on two specific
burn scenarios for F that are closest to what occurs in real rockets.
The fuel burns radially outwards in a symmetric manner (see Fig. 2) so that F*
mf=pfxL(R2-r 2) (5.47)
'_2
dmf
__=_pfxLd(r2) (5.48)
dt dt =-pxR_u
Hence
d(r2)=_ dt (5.49)
fo r2 _o t
and
r2=_R12ut=_R12u x (5.50)
L Lot
From(50)and(45)
_____L
=Pf_L(R2_
:)
mr0 = 1 --rL I-_
pf _rL R 2 R2 = (5.52)
and so,
(5.53)
4 12 12 4
fromwhich we have
12 2
(5.54)
Similarly,
)-f=Jf =_(R2+,.2)
mnR2 mO--2R2 -=('l-'r)-_ =_ (5.55)
and
(5.56)
90
c=12+LI2 (5.57)
we have
C'=0 (5.58)
and
2gcc'=0 (5.59)
so that
(20) yields
.... g'c 2
2 _t'c c'
I =If+ _--- (5.62)
moR 2 mo R2
i' =_8"_t_
12 2 (V+I
__
,02
(82+ (5.63)
JB _l
J =- Jf =-x (5.64)
(5.65)
R R
m (5.66)
Hence,
(5.67)
(5.68)
(5.69)
From (69), the function 0(x) is a linear function of "1:; it has a slope of-1 and is
1
0
We conclude from the figure that the spin rate always decreases at the beginning of the
burn. If 13>- ":2", the spin rate continues to decrease all the way to burnout. For 13 < _2, the
spin rate bottoms out during the burn, and then increases during the later phases of the
o =- +132/
2=0 (5.69)
That is, at
x = 132/2 (5.70)
94
This result is consistent with, and augments that obtained for the cylinder, for which 13 = 1
[ see (4.57) and Fig. 4.8]. The same result is validated by Fig. 4, obtained by numerical
integration of (40).
¥=2
ct = 0.01
8= 10
6t=2
"_t = 1.2
y2=l
Fig. 5.4 Centrifugal burn" the effect of expansion ratio I_ on spin rate
At ignition ( "1: = 0 ),
It is clear from (72) and (73) that % > (h, and that it is possible for any of these quantities
t
to be negative or positive. These facts, coupled with the fact that 0 is negative, mean that
the three scenarios depicted in Fig. 5 ale possible, depending on the values assigned to the
parameters.
°_ II 1
centrifugal bum can (i) decrease from ignition to burnout, (ii) decrease initially, then
diverge later, or (iii) increase from ignition to burnout, depending on system parameters.
Factors such as the distance of the exit plane from the combustion chamber (51), shape
..
factor of the propellant grain (_5), location of the propellant grain (82), nozzle expansion
ratio ([5), and the relative amount of propellant in the system (_), all affect the transverse
angular velocity. The main difference between the general axisymmetric system being
considered here and the variable mass cylinder considered in chapter 4 is that case (iii) in
Fig. 5 does not exist for cylinders. In other words, the transverse angular velocity of a
variable mass cylinder always decreases initially, while a general axisymmewic system can
exhibit divergent transverse rate throughout the propellant bum. Fig. 6 shows the
_D
m-
_= I0
o_ = 0.01
5t=3
TI = 1.2
y2=l
We recall that in uniform bum, the dimensions of F remain unchanged, while its
density decreases continuously with time - see Fig. 7. In this case., mf has the same
iiiiii!iii!i!iiiiiiiii iiiiiiiiiiiiiii!iiiiiii
! 12 _-
iiiiiiiiiiiiiiiiiiiiiiii_iiiii_i!iii!iiiiiiiiiiiiiiiiiii_i_i_ii_i_i!_!_!_ii
!
"1
Hence
(5.75)
:)f
f0 d(pf)=-p 2
R2LJo t dt
98
so
R12ut (5.76)
Pf= Pro- P R2L
and so,
(5.79)
m,, R 2 mo [.4 R 2 12 R2J 12
(5.80)
and
j_ =_1. (5.82)
2
c =g2+L/2 (5.83)
c'=O (5.84)
and
and
(5.88)
(V"+ 1 - x)2
--i --t
ge
R = _'J"
R + 9.-_R = _ +_"7 (5.9O)
m'
mid = m'_ - =- 1 (5.91)
Hence,
(5.92)
(5.93)
_('_)=q52+_2+_518+6 -4I" (_+i-,t)2(82+-8-2-2}2
(5.94)
From (94), the function 0(x) is a constant. Fig. 8 shows O for various values of
13>1
13=1
O
13<1
If 13 > 1, the spin continues to decrease all the way to burnout. For 13= 1, the spin rate
always equals its initial value. For 13< 1, the spin rate increases from the beginning to the
end of the bum. The same result is validated by Fig. 9, obtained by numerical integration
of (40)
102
¥=2
Ot= 0.01
8= 10
8t=2
82=3
Yt = 1.2
tD
u'1
oq
y2=l
(:3
Fig. 5.9 Uniform Bum" the effect of expansion ratio 13 on spin rate
(5.96)
I03
(5.97)
Hence _b0> _l, always, and either of these quantities can be negative or positive. Since we
t
also have that _ is always negative, we have the same scenario obtained for centrifugal
bum. Numerical simulation results provided in Fig. lO below confmn these inferences.
o
LO
/
A
/
Lo
/
.-2-
¥-- 15
B"I
0
/ = 0.01
20_ Si=3
I
_xyO
[ Yt = 1.2
72=1
o"
-0.0 0.2 0.4 0.6 0.8 L.O
T
CHAPTER 6
FORCED MOTION
OF AXISYMMETRIC SYSTEMS
All the development of the last two chapters have been restricted to systems
subjected to zero external torque. The only forcing function present has been a perfectly
axial thrust, whose line of action passes through the system mass center at all times, and
thus produces no couple on the system. Great pains are taken in the design of real rockets
to minimize thrust misalignments. Yet, it is unrealistic to expect that the sys_m mass
center, which generally moves during rocket burn, will remain positioned on the line of
action of the resultant thrust vector throughout the propellant bum. Thus, torques due to
.?
thrust misalignment are often inevitable. There are also other potential sources of
moments. For this reason, we will, in this chapter, develop relationships that describe the
effects of externally applied moments on the attitude behavior of the same axisymmetric
The equations of forced attitude motion of the system of Fig. 5.1 are really the same
as (5.17)-(5.19), except that the right hand side of each of these equation is non-zero.
Thus, we have
.?
The forcing functions Mx, My and Mz are, in general, functions of time. This is so
because at least the contributions from thrust misalignments vary with time since the system
and
Mz = Mz / J (6.5)
" 106
(6.7)
This solution can be thought of as the sum of two components : a first term that represents
the homogenous part due to initial condition, and a second term - the particular solution -
that is contributed by the forcing function. If J, jr, rhf and Mz are known functions of
time, and the initial spin rate is known, (7) determines uniquely the spin rate ¢x_zof the body
S. Note that ¢Ozdepends only on 2L2 and Mz, and is independent of Mx and My. This is
We are now ready to investigate the transverse angular velocity components of the
Mxy = Mx + j My (6.9)
107
where
j= (6.10)
We also define
o=(J-I)/I (6.12)
and
Or
.. 108
d_xy
dt + P2(t) tOxy = Mxy (6.15)
where
(6.17)
or equivalently,
where
(6.19)
and
. 109
(6.20)
of theexpression;they arc:
(6.21)
and
We observe that in general the spin torque, Mz has no effect on the transverse angular
velocity, nor is the spin rate affected by any of the transverse torques.
110
equations of attitude motion and their solution. We exploit here, a transformation defined
0, has the dimension of angle (radians) and is related to the spin angle. It will be used as
Note that
-?
dO = Ct_z{_ dt (6.24)
SO
The starred quantities are assumed to be expressed in terms of the new variable O. This is
indeed possible because (23) gives 0 as a function of t, and can be inverted to yield the
" 111
follows :
d__+coy
de
[.
+ dJS__dmf
Lde de
. (te2 + R2 /4)]
(tk
I"
_ 1_ X
0)z (J* - I')
(6.28)
and
(6.30)
dO
112
wh¢l'c
_,(o)=,_(o)+j_o) (6.31)
_ =[di-
kdO
""
__(_:_
dO
+R_/_ I/i *
(6.32)
and
(6.33)
if
w_d_o:(o)as
(6.34)
then the solutionof (30)for COxand COyin _rms of thenew variable9 can be written
explicitly as
COx
(0)= (_o cosO-_: sinO)e -r* (0) .+
(6.35)
_z_'-?i
e [r" (s)- r" (o)] [_(_)co__o)+g(s) si_s-o)]e
113
,_(o)=
[,_o
,_o+_ co,o]
e r"(o)
+
(6.36)
fo° c[l"*{s!-I'_{O)][I_fy(s)c°s(s-O)-l_x(s)sin(s-O)]dsc0z(J
-I)
In order to obtain explicit expressions for the known functions such as I, J, Mx and
My, the function t(O) must be determined. This may or may not be a trivial task. The
integrand of (23) is a rational function of time whose integral may not have a closed form
sure that the complete solution sketched above can always be obtained by numerical means
- at least.
The above solution can be used to investigate the altitude motion of an axisymmetric
I; O)z(J* - I*)
In this form, the physical significance of the various parameters of the problem is not so
.?
obvious. Simpler and more revealing expressions for COxand o.b, are obtained for the case
¢os(o)-,,,yo
sin(O)]
+
(6.41)
1 {l_x sin{O)- l_y[1- cos(O)]}
(J" - I*)¢OzO
where
115
0 (t)= -_ _)zO
t (6.43)
from (4), (5) and (7). (41) and (42) indicate that transverse velocities increase with
u'ansverse torques and decrease with the inertia difference, J* - I*, and initial spin rate.
Spinning a rigid body about one of its principal axes provides the so-called spin
rigidity or gyroscopic stiffness; that is, a resistance to external disturbances during motion.
This is the well known advantage of spin stabilization. In a spin stabilized system, the
transverse angular rates COxand cA/ axe usually the result of unforeseen disturbance, and
are generally undesirable. If Nix and My are viewed as constants, body-fixed disturbance
torques in (41) and (42) above, then, these equations show clearly that high spin rates will
reduce the effects of the disturbance torques on vehicle motion. This indicates that the
nominal spin rate of a spin stabilized vehicle should be given the highest value compatible
with the vehicle's structural integrity. A fact that agrees with physical intuition. It can thus
be said that as the spin rate increases, the stability of a given system increases as measured
For a single spinning body, (41) and (42) also show that gyroscopic stiffness
depends in a crucial way on the system inertia property, J* - I*, or the difference between
the spin and the transverse moment of inertia of the body. Spin rigidity vanishes as the
inertia
differencetends to zero.In other words, bodies with sphericalsymmetry can have
no gyroscopic stiffness,
irrespective
of thevalue the spinrate.On the otherhand, heavily
CHAPTER 7
CONCLUSION
The study presented in this report deals with the dynamic behavior of spinning
bodies that lose mass while in motion. The study began with the determination of the
complete equations of such systems using one of the most powerful methods of analytical
dynamics m Kane's formalism. The resulting equations are then applied to the study of
the attitude dynamics of space-based variable mass systems using mathematical models that
vary from a simple cylinder to general axisymmetric systems. The emphasis of these
extract as much information as possible about the motion of the system, without actually
simple cylinder with four possible mass loss scenarios. Results obtained indicate that such
a system can fly stably in a torque free environment if the ratio of the radius m the length is
small. Large and short cylindrical vehicles can have stability problems, especially if they
are subjected to radial bum. This study goes far beyond previous work in this area by
clearly identifying the exact inertia conditions and the precise stages of vehicle motion at
117
which the character of motion begins to change. For cases where there is stability issue,
axisymmetric variable mass system. Although important differences were found in the
details of the motion of the general axisymmetric system as compared to that of the simple
cylinder, the main stability results did not change. It was found in this second example,
that in addition to the shape factor of the system propellant, the location of the propellant
grain within the system as well as the nozzle expansion ratio all have significant effects on
In the last section of this study, the combined effect of mass loss and external
transverse torques on system motion is examined, and solutions to the equations of motion
in this case are presented. Overall, this study augments considerably, previous knowledge
of the dynamics of variable mass systems and gives great insight into the effects of mass
This study shows precisely how the system's auimde response is tied to such
relative mass of the propellant. In this sense, the study will be of great use to designers in
that it indicates what must be done to avoid motion instability of the type witnessed in the
.
Ball, K. J. and G. F. Osborne (1967). Space Vehicles DYn.amics. Oxford. 153-
157.
6. Belknap, S. B. (1972). "A General Transport Rule for Variable Mass Dynamics."
10(9): 1137-1138.
1(4): 927-928.
119
10. Breuer, D. W. and W. R. Southernland (1965). "Jet damping effects: theory and
11. Buglia, J. J., G. R. Young, et al. (1961). Analytical Method of Approximating the
Motion of a Spinning Vehicle with Variable Mass and Inertial Properties Acted Upon by
12. Buquoy, G., (1816), "Exposition d'un Nouveau Principe General de Dynamique,
Paris.
13. Byrne, J. W. J. and S. Raynor (1964). "An Analysis of the Yawing Motion of a
14. Cochran, J.E. and Kang, J.Y., (1991) "Nonlinear Stability Analysis of the Attitude
15. Cohan, F. A. (1966). The Influence of Rocket Motor Geometry_ on Jet Damping.
19. Davis, L. J., 3. W. FoUin, et al. (1958). Exterior Ballistics of Rockets. New York,
20. Dryer, M. (1963). "Equations of Motion for An Arbitrary Body of Variable Mass."
21. Duan, C.Y., (1989) "Dynamics and Applications of Nonholonomic Variable Mass
China, p 81-86
22. Eke, F. O. (1982). Star 48 Coning Anomaly. JPL Internal Document No. 343-82-
989 1982.
23. Eke, F. O. (1983). Dynamics of Variable Mass Systems with ADolication to The
Star 48 Solid Rocket Motor., AAS 83-370, Advances in the Astronautical Sciences, V 54,
p 671-684
24. Eke, F. O. (1991). Angular Momentum and Multibody Dynamics. AAS 91-110,
27. Fang, B. T. (1965). "Kinetic Energy and Angular Momentum about the Variable
and Calculation of the Critical Bubble Radius." Journal of Engineering Physics 24(5): 589-
592.
29. Flandro, G.A., et al., (1987) "Fluid Mechanics of Spinning Rockets" AFRPL-TR-
30. Gadsden, C. P. (1964). "Newton's Law With Variable Mass." American Journal
31. Gadsden, C. P. (1966). "Laws of Motion for Variable Mass Systems." Arn_
NACA TM 1255.
121
33. C_re,Z.-M. and Y.-H. Cheng (1983). "The Hamilton's Principle Of Nonholonomic
Variable Mass Systems." Applied Mathematics and Mechanics. English Edition 4(2): 291-
302.
Mass System with Application." ASME Journal of Applied Mechanics 51: 435-437.
Nonholonomic Variable Mass System." ASME Journal of Applied Mechanics 49: 429-
431.
36. Gilmore, A. W. and T. L. Keller (1957). "Damping Due to Internal Flow." Jo_al
37. Gopak, K. M. and O. A. Goroshko (1975). "Vibrations and Stability of a Free Rod
of Variable Mass, Taking Account of Motion of Its Mass Along the Axial Line." So_et
38. Gopak, K. N. (1976). "Dynamic Stability of a Free Rod of Variable Mass." So_et
39. Grubin, C. (1963). "Mechanics of Variable Mass Systems_ ;' Journal of Franklin
Dissertation, UCLA.
42. Hrushow, W. K. (1969). "Euler's Rotational Equations for Bodies with the Inertia
Tensor Varying due to Mass Redistribution and Mass Loss." _ 7(2): 337-339.
44. Jarmolow, K. (1957). "Dynamics of a Spinning Rocket with Varying Inertia and
-?
" 122
McGraw-Hill, Inc.
Partially Filled with an Ideal Fluid of Variable Mass." Mekhanika Tverdodo T_la 28: 84-
88.
54. Leitmann, G. (1957). "On the Equation of Rocket Motion." Journal of British
55. Leimaann, G. (1958). "A Remark on the Free Vibration of a Variable Mass."/et
28: 197-198.
56. Leimaann, G. (1963). "On the Stability of Motion of a Vibrating Variable Mass."
59. Marcinek, J. J. (1958). "On the Free Vibration of a Variable Mass." ,[¢,l_I_gC.uLsJ_
28: 198-199.
60. Martz, C. W. (1961). Method for Approximating the Vacuum Motions of Spinning
61. Matolyak, J. and G. Matous (1990). "Simple Variable Mass Systems: Newton's
64. McPhee, J. L (1993). Two Different Methods for Simulating the Motion of
Academic. 385-390.
66. Mei, F.X. (1989) "Time Integral Theorem for Variable Mass Systems",
108
68. Menon, V. J. and D. C. Agrawal (1987). "Newton's Law of Motion for Variable
51(3): 241-243.
: 52-53,84-87.
71. Medam, J. L. (1975). I_fllalni_. John Wiley & Sons, Inc., p 393-420
72. Meshcherskii, I.V., (1904), "Equations of Motion of a Variable Mass Point in the
General Case," St. Petersburg Polytechnic University News, Vol. 1, pp. 77-118 (reprinted
73. Meyer, R.X., (1983) "Convective Instability in Solid Propellant Rocket Motors",
74. Mingori, D.L. and Yam, Y., (1986) "Nutational Instability of a Spinning
Spacecraft with Internal Mass Motion and Axial Thrust", AIAA paper 86-2271, AIAA/AAS
1103-1104.
76. Mikhailov, G. K. (1975). "On the History of Variable Mass System Dynamics."
80. Nita, M. M. (1979). "On the Equations of Motion about the Mass Center of the Jet
Aircraft Considered as Variable Mass System." Rev. Roum. Sci. Techn. - Me(:. AI_I..
81. Nita, M. M. and R. Patraulea (1990). "Rotary Damping in Aircraft Motion due to
Jet Propulsion System." Rev. Roum. Sci. Techn.- Mec. Ap91.. Tome 35(1): 3-15.
Variable Mass Gyroscope (in Russian)." Vesmik Lenin m_ad Univ. 14(19): 121-129.
83. Pan'kin, N. A. (1970). "Motion of a Train in the Start-up Phase (in Russian)." Tr.
85. Papis, T. and C. W. Bullard (1968). "Atfiuate of a Spinning Rocket under Thrust
with Statistically Varied Error Sources." Journal of Spacecraft and Rocket 5(5):
87. Poisson, S.D., (1819), "Sur le Mouvement d'un System de Corps, en Supposant
les Masses Variables, Bull. Sci. Soc. Philomat, April, pp. 60-62, Paris.
89. Pomeranz, K. B. (1964). "Newton's Law for Systems with a Variable Mass."
90. Prato, D. and R. J. Gleiser (1982). "Another Look at the Uniform Rope Sliding
over the Edge of a Smooth Table." American Journal of Physics 50(6): 536-539.
2(8): 1482.
126
94. Rankin,R. A. (1949). "The Mathematical Theory of the Motion of Rotated and
Unrotated Rockets." Philosophical Transactions Royal Society_ London Series A 241: 457-
585.
95. Rosser, J. B., R. R. Newton, et al. (1947). Mathematical Theory of Rocket Fli_t.
96. Rott, N. and L. Pottsepp (1964). "Simplified Calculation of the Jet Damping
97. Shomin, R. (1992). "A Special Variable Mass Dynamics Problem." The Physics
Position of a Solid body with Variable Moments of Inertia in its Motion About the Center
99. Sibert, H. W. (1953). "Relation of Force, Acceleration, and Momentum for a Body
100. Siegel, S. (1972). "More about Variable Mass systems." American Journal Of
101. SiUi, C. (1965). "Sulla Stabilita Dei Missili Con Massa Variable." ]_..olld,La,..hl_
20: 194-209.
of Variable Composition." Journal of Applied Mechanics and Technical Physics 13(1): 90-
94.
Mass Layers and Membranes." Journal of Applied Mechanics and Technical Physics
14(5): 715-719.
106 Squire, W. (1971). "An Example of Motion of a Body with Variable Mass."
107. Tait, P.G., and Steele, W.J., (1856), "A Treatise on the Dynamics of a Particle,"
Cambridge.
108. Tempelman, W. H. (1972). "Equations of Motion for the Variable Mass Flow-
109. Thomson, W. T. (1966). "Equations of Motion for the Variable Mass System."
111. Thomson, W. T. and G. S. Reiter (1965). "Jet Damping of a Solid Rocket: Theory
112. Thorpe, J. F. (1962). "On the Momentum Theorem for a Continuous System of
114. Trigg, G. L. (1966). "Laws of Motion." American Journal of Physics 34(1): 71.
115. Van den Akker, J. A. (1964). "Newton's Law with Variable Mass." _eaScan
Variable Mass Systems." ASME Journal of Applied Mechanics, V62, No. 4 : 970-974
118. Warner, G. G. (1968). "A Re-Evaluation of Jet Damping." Journal Spacecraft and
119. Wu, M.-S. (1986). "Note on a Conveyor-belt Problem." The Physics Teacher 24:
220.