Kron1934 PDF
Kron1934 PDF
Kron1934 PDF
ELECTRICAL MACHINERY
By GABRIEL KRON*
SYNOPSIS
a.) It is shown that the charges through the terminals (brushes, taps,
etc.) of the innumerable types of rotating electrical machines (except
of the alternator) are not true Lagrangian coordinates and the usual
form of the Equation of Motion of Lagrange
#t(:~) - :~ + :~ =fa
as interpreted by Maxwell for closed electric circuits moving in a mag-
netic field is not valid for them, neither is the Equation of Voltage (derived
from the former equation by ignoring the geometrical variables.)
ea = Ra~if3 + d<pa
dt
= R ;f3 + d(L a{3i f3 )
all' dt
since the coordinate axes are not connected to the moving conductors.
The generalized form of the Equation of Motion valid for all electrical
machinery and in general for non-holonomic dynamical systems has been
given by Boltzman and Hamel as
!i (aT)
a
_ aT + aT (act _ ac~) C k C n i1' + of = f
dt a:t ax a aj;6 ax" ax k a l' aj;a a
which includes as a special case the usual form. The generalized form
of the Equation of Voltage for rotating electrical machinery becomes
- R ·tJ
ea - a~l
+ d<pa
dt + 'l'ta~
.1, .t _
-
R '8
a;<l
+ d(LaBif3)
at
+L 1'6
act C k • t
axt alaV
where: 1.) the induced voltage La~itJ/dt is due to the variation of the
currents 2.) the Coriolis-yoltage (dLall;'dt)i tJ is due to the motion of
the coordinate a.us and 3.) the last term is a voltage due to the motion
of the conductors.
b.) The theory of non-holonomic dynamical systems is developed in
* Engineering General DepL, General Electric Co" Schenectady, N, Y.
103
104 GABRIEL KRON
tensor symbolism and it is found that the usual explicit form of the
Equation of M olion for holonomic dynamical systems, that is
d2x~ dx~ dx'Y dxf3
aaf3 dt2 + [!3'Y, al dt de + Ra/J dt = fa
de" = Z"fjdif3
These forms show that the equations of rotating machinery are identical
with those of stationary networks if ordinary differentiation is replaced by
absolute differentiation. Or in other words the appearance of the
gravitational mass of the rotor during acceleration in a purely electro-
magnetic system chang(ls all ordinary drivatives into absolute derivatives
(agalljaX'Y=O) anditendowes the electncal coordinates x'l' with cylin-
drical properties. The Equation of Small Oscillations becomes
calculations may leave out the sections dealing with physical inter-
pretations.
As expected, if the various terms in the generalized Equation of Mo-
tion are reinterpreted as physical quantities (flux-density, vector-po-
tential, etc.) the new forms are nothing else but the Field Equations of
Maxwell generalized for moving bodies and moving coordinate axes. It
is intended to treat the various forms of the Field Equations (wave,
impulse-energy, etc.) in another publication and to interpret physically
the various curvature quantities (Kall' K) and tensors (stress--energy,
electromagnetic, etc.) as they apply to rotating electrical machinery.
d.) It has been most gratifying to find that while other methods require
often many months of preparation and thought and calculation to set up,
even the steady-state performance of new types of machines (and that can
be done usually only by a few engineers experienced in such mental disci-
pline) the dynamical method presented in this paper gives identical results
by a routine calculation in only a few hours and it can be employed by
anyone who knows how to multiply matrices and find their inverse. Even
in cases where other methods are practically helpless as in case of most
machines with asymmetrical windings or magnetic structure and unbalanced
1·mpressed voltages or in cases where various types of machines are inter-
connected in any arbitrary manner, this method gives correct results with
equal facility.
Other advantages of this method are:
1.) In learning the performance calculation of one machine at the
same time the performance calculation of all other machines is learned.
2.) First the transient performance of machines during acceleration
is analyzed The transient performance calculation with the speed
maintained constant and the steady-state performance calculation
follow as special cases.
3.) The fundamental equations derived for the transient analysis of
rotating electrical machinery are identical with the fundamental equations
of Dynam1·cs and of multidimensional Differential Geometry, hence the
reseaJ:ches of these sciences can be applied to the study of electrical
machinery.
4.) In introducing any future refinements into the analysis, for
instance the effect of space-harmonics, multiple phases, slot-openings,
brush-currents, etc. all fundamental equations set up in this paper
remain unchanged, only the value of the constants to be substituted
into the equations are to be changed.
3.) A powerful mathematical tool is acquired which can be used not
112 GABRIEL KRON
only in the study of rotating machinery, which is the case with all
other tools, but immediately can be applied in the most advanced
studies of mathematical physics.
e.) (In a previous paper the sudden short-circuit performance with
the speed maintained constant and the steady-state performance have
been analyzed in greater detail from purely physical considerations in
the vector and dyadic notation of Gibbs. In a second paper small
transient variations in speed and steady hunting have been covered
from the same point of view. Important labor-saving devices such as
X-matrices, complex vectors etc. have been also introduced in them.)
Anyone interested only in the dynamical aspect of the theory of
t'lectrical machinery may consult the following sections: II, III, IV, V,
VII, VIII, IX, XXIII, XXIV, XXX.
THE REPRESENTATIVE MACHINE WITH MOVING COORDINATE AXES
I. Hyper-complex Numbers
a.) The quantities the Absolute Calculus deals with are generalizations
of the complex number A +jB. A more compact notation for complex
numbers is (A, B). A three-dimensional complex number is iA +
jB
+ kC, which may also be written as (A, B, C). An evident extension
is to supply each number with two indices as Ajj +
Bjk +
Ckj +
Dkk
+ .... This may be denoted by the compact form
A,B, ....
C,D,
Adq, AJq or A~q. The order or position of the indices can not be inter-
changed, since A dq and AJq belong to different sets of n 2 quantities.
One term of a triadic is written as A dqj or AJ/ etc.
An equation such as ea = ~ Za~if3 represents an equation along axis a
f3
where the index {3 assumes all the possible values of the indices repre-
senting the various axes d, q, f, etc. There is one big difference between
this notation and the usual scalar equations however. Wht'le with the
usual scalar equations each coordinate axis has a different equation, in
tensor notation the equatiolls for all coordinate axes are identical. The
index a is simply replaced in turn by d, q, f, etc. for the various axes.
It should be noted that the indices to which the summation sign
applies appear twice in the same term (once as an upper and once as a
lower index). It is an accepted corwention to dispense with the summa-
tion sigll and to write the above equation as Ca = Zasif3. If a is say d,
then
Zaf3i~ = Zddid + Zdqi + Zd/if + ....
q
The summation sign may occur twice or more in the same term. For
instance ~~ Ra{3iai~ ~ R a{3i a i{3.
a{3
It should be noted that:
1.) The two indices that are written with the same letter are called
"dummy" indices, the others "free" indices.
2.) Any letter may be used in the same term as a dummy index.
For instance r a{3'1~'{3 = r a"'Ii".
3.) One of the dummy indices must be an upper, the other a
lower index.
4.) Since most equations of this paper are vector equations, in every
term of the equations only one of the indices will be a free index. This
index must be denoted by the same letter in each term of the equation
and it must be an upper or a lower index in all terms.
q d f3 b a
q ~ 0 0 0 0
-- -- -- --
d 0 0 0 0
- - rdd - - - - - -
Rap = b 0 0 0 0
- - - - rbb - - - -
a 0 0 0 0
- - - - - - raa - -
0 0 0 0 r tt .... , ....... 1
2.) The second set of numbers represents the self and mutual induc-
tances of all windings along the axes considered, and the moment of
inertia of the rotor. The inductance of all rotor windings is expressed
as a function of the rotor angular displacement 0 = xt. The set is
represented by a square array and is denoted by a symmetrical dyadic
La{J. This dyadic will be called the "metric tensor" (equ. 2).
q d Ii b a
q Lqq Lqd Lqb (0) Lqa (0) 0
--
d L dq Ldd Ldb (8) Lda (0) 0
--
Laf3 = b. L bq (8) Lbd (0) Lbb (8) Lba (0) 0
--
a Laq (8) Lad (0) Lab (OJ Laa (0) 0
--
0 0 0 0 L tt ... 2
The knowledge of these two sets of numbers Ra/1 and La/1 1'S sufficient to
find the transient and steady-state performances of any machine assuming
no magnetic saturation and no iron losses.
116 GABRIEL KRON
,d
FIG. 1
Rmn a 0 ~ 0 0
d. o 0 0 1'", I' 0
o 0 0 0 1'" 1.•••..•...• .4
cs a.
Z.
d. d. q; q~ ~ q~
I
d~ ~ d~ d~ q .. q~ q;
d; 0 0 - Md -M. d'• 0 0 0 0 ~/G H./2 0 0
If d~ 0 0 -M. -~d "~ 0 0 0 0 M~/Z Md/2 0 0
d; 0 -l:" L~d 0
d~ 0 0 0 0 L;rl;.Ya L;.d-l~)IZ -M.!z -M,/Z
o 0 -l"'d L~~ 0 0 0 0 _d:
L~d-L~C! L~cL~ liZ :M../2 -M
L: Lt, 0 0 fna,c 9~ M.~ (L;'-l:')/, L~~-L! )/2 0 rvz 0 0 0
, L;, L; 0 0 9; MoIZ M l~4-L~J/2 a.; -L;.11Z 0 0 0 0
M M. 0 0 'I; 0 0 -M./z -MJZ 0 0 0 0
1'1. M, 0 0 9: 0 0 -M,,2 -N./Z 0 0 0 0
a-
d'• d~ d~ d; '1~ 'I'- qs q;
d; "'d+L~p L~~p Io\jP MJP -M~pe -/lot/l,)9 0 0
" d~ L~dP r;. +L:dP M~p ~p - Motp& -M.p& 0 0
d; ~p M.p r: T L:4P L~p - L:41> 9 -L!~p& 0 0
d~ M.rp M.p l~4P r:+L~dP -L~jpe - L:~p& 0 0
1If":
'1~ 0 0 L:,p& l~,pe ~;+L~.p L:.p 1'1,,, M.p
'I: 0 0 L:"p& L!,p& L!.p r:+L:'P M,p M.p
9, 0 0 M.P9 M,pfr M,P H,r rl,·L:,r l:. p
~ 0 0 M.~e H,pe Mqp M,p U.1'l fol+ ll.p
THE" CON5T~NTS or TItE' RE"PRESEN TA livE MACHINES WITH TWO STA TOR
AND 1"«10 ROTOR LAYERS C1' WINIlNG5
TABLE I.
118
ffi f6l FIG.2 - flC.3.
J. J ' ' ' '
~
,
els , ....9.' q.,
d; d'" . : as Os "r q;~.
3<ts q. 0 0 0 I 'Ie 0 0 0 I
.
ASYMMETRICAL INDUCTION HOTOR SALIENT-POLE SYNCHRONOUS MACHINE,
(SPlIT·'"ASE. SIN'lE-PHASE, POLYPHA5 r ETC.) SLIP-RING IIiDUCTION MOTOR
FIG. 5.
S
flG.i.
7f
d. ~ d; cI~ q~ ~ ~
f11
~
····;:;:1··· d.,,+---+-:--+:- • d5 ~IO 0 0 0
b .
...
• • oc.
".
a
~
.. a 0
C... =b
r-
~o::....j--:--+--t~
, «[llc" a
=a 0 ~osO( sill 0 0
otd:oo
0(
010
:3 9, j...:...~~-:-~
~• ...:..0.L....:.0-L..--'_ 9~OO 001
ne;.6 riG. 7
~
.f ~
f!1 C:=7t-:+--=--l.~:""-
d.
'" :'\:. P"
'-'---....:...~::.:::::::r.J
fIG.S
fiG. 9.
~~
,7r I I
d's d~" q~ d, q. qs
~~
q~
oI
c-.::
d. I 0 0 0 0 I 0
b cose " q.
0 -sine I 0 oI I "
b"': . a.' f 0 0 0 n f 0 OIOln
TASlE 1I.
119
fIG.IO. flG.11
~C:';~~+-4- ~
. J'd,7<,
i
• '1 r ,.
d C:.. d,~
: f f~
g
PHASE AOVAI'ICER COMPENSATED REPULSION MOTOR
F"lG.'Z.~ flG.13.
~ c>rOO
I
@
d, ~ : cI.: "cI~ ,
". ~.:' "a coses",n9
: 0 0 I
.. ··. /::~f d,C,,= b sine coseJ : f
0 I J
b
::
qr
IS.
'It L-.:........_ - ' 3<t s
~ d; cI~ d: riG
J( ~ 14q, q~
'. .::... ' ( n ,osC( 0 0 sinG( 0 f
\, ....')\:~... e" t< a 0 0 cose sine 0 0
d:
tIC.
0 0 0
:: or
.. boO -sine cose 0 0 ~: 0 0 0 0
8'\s qs 0 0 0 0 0 n 3 0 0 0 0 0 n
n'NN-WE1CHSEL MOTOR
riG. IS f'IG./6
::?§3
f ~
FIG. 17. J(
~
... d.." . . ' d.. .I, I cos~ -sini 0 0 0 0 0 f l o , 0 0 0 0 oj
">.:.. 'I'm '.:.:.. tm
: '. ,Lf/t. '. "M' 11
0( ct,
d
0 sin& C"~ I 0
0 0 0 0 I
0 0 0
0 0 0
"'I
cia.
0 I 0 I 0 0 0 01
0 0 0 0 I 0 0 oj
. CjI( '" ,,:: or 'ls' 0 0 0 I Q 0 I 0 01
--=
0 0 0 0 0 I 0 0
-::::> d,.,0 0 000010 cia. 0 0 0 0 0 0 I 01
-=q" ...=:;1"!,0 Cf.. a 0 0 0 0 0 0 I ~ 000 0 0 0 0 I
But
n
1 aLmnimi = Lmkim
2 ai k
and
n
1 aRmnimi Rmk10m
"2 aik
122 GABRIEL KRON
Also
m
d(Lmki ) = L dim + dLmk im = L ,dim + aLmk dxn im
dt mk dt dt mk dt ax" dt
= L mk
dim
7ft
+ aLax" mk 'm 'n
z ~ .
Substituting
ek -_ R mk~'m dt + (iJL"'k
+ L dim mk ax" 1 aLmn)
- 2 axk 'm 'n
Z 1 •••••••••• 6
*
Motion will be assumed to be represented in any coordinate system
by replacing [mn, kJ by r/l'Y. a representing any function of La/l
d'/l
ea = Ra/l ill + La/l + r/l'Y. a i/li'Y I .••.•...•.•.•• 10
the actual calculation of voltages is cut into two and the physical
definitions are expressed in simpler forms
b.) In calculating r mn , k from equ. 11 it should be noted that one of
the indices must be t since Lmn is a function of Xl = (J only. That is
from equ. 11
k
d. a b q.
d. 0 - Md sin fJ - }V!d cosfJ 0
m
a - Ma sin fJ 2(L rq - L rd ) sin fJ cos fJ (L rq - L rd ) (cos2 fJ - sin 2 fJ) MqcosfJ
r;'m.t = b - Md cos 0 (L rq - L ) (cos 2 0 - sin 2 O) - 2(L rq - L ) sin 0 cos 0 - MqsinO
rd rd
b 0 M q cosO - MqsinO 0 .. 15
n
a b
d. 0 (M d sin 0)/2 (Md cos 0)/2 0
GABRIEL KRON
m
a (Md sin 0)/2 (L rd - LrJ sin 0 cos 0 (L rd - Lrq)(cos2 0 - sin 2 O)/2 - 'M q cos 0)/2
r;'m.t = b (Md cos 0)/2 (L rd - Lrq) (COS2 0 - sin 2 0) /2 (Lrq - L rd ) sin 0 cos 0 (M q sin 0)/2
b 0 - (Mq cos 0)/2 (Mq sin 0)/2 0 .. 16
NON-RIEMANNIAN DYNA:\lICS 125
+ Lmk d';t + r.
m
ek = Rmd" d, kimi t (k = d., a, b, q.) .... 17
et = Rtt~
t
+ Lit dit
dt + r mn,
'''''n
t~ t . . . . . . . . . . . . . . . . 18
1.) r.:. 1
, = -2 aL
t k
mk
ax n = one half of equ. 15 ........... 20
J
2.) r; n k
.
= -2 aL
nk
ax"'
= one half of equ. 15 ........... 21
tk Rmkt
'm
+ Lmk dim
- + r(mt),
dt
k
0m.t
t t .......•..••. 23
126 GABRIEL KRON
where
as - L rd sin (J.
b.) Hence the following identity exists between the flux-linkages set
up at axis d r by currents at the new and by currents at the old coordi-
nate axes
idrL rd iaL rd cos (J - ibL rd sin (J
The relations between the differentials of the new variables dx>: and
NON-RIEMANNIAN DYNA:\IICS 127
the old variables dx"' are found from the two current-equations by
cancelling dt
ds dr ql' q8
d 1 0 0 0 0
m
8
-- -- --
a cos 8 sin 6 0 0
-0- -- --
C:, = b 0 - sin 6 cos e 0 0
-- ----
q. 0 0 0 1 0
-- -- --
L-.
0 0 0 0 1
........ 28
As an example the current along the new axis dr is found from its values
along the old axes
·dr _
L -
cdr ·d8
ds I
+ Cdra I·a +C drb L·b + cdr q 8 I
·q8
COS
8 I·a _ .
SIn
6·b
I.
The values of the other vectors and polyadics along the new coordinate
axes are also found with the aid of the transformation tensor C:, from
their values along the old coordinate axes, but their formulae of trans-
formation are different from that of i" in equ. 30.
aj"' aj"' b
dx dr = -
axa d.ra + -axh dx . . . . . . . . . . . . . . • . . . 31
If this last equation is assumed then
.'
FIG. 18
1.) the unit vectors and indices d and q are interchanged, also
a and b
2.) the instantaneous position of the coordinate axes used in the
previous paper are given in Fig. 18, while their position as used in this
paper is given in Fig. 1.]
ea -R
- afll~+~+~1
de "'la1 • •.•.••••• , .•••••• 34
134 GABRIEL KRON
new polyadics der.ived with the aid of the rectangular matrices are called
"induced" polyadics
p:'r~~-L-:~c;C:--I ..................... 38
d. dr ~ qr q.
dB Lsd Md 0 0 0
"d Md L rd 0 0 0
r
Ln = qr 0 0 ~ ~ 0
q. 0 0 ~~ 0
0 0 0 0 L tt ........ 39
~ON-RIE:\IANNIAN DYN.nnCS 137
d. qrdr q.
d. r.d 0 0 0
-- -- -- --
0
dT 0 rr -0- 0 0
-- -- --
R". = qr 0 0 fj' 0 0
-- -- -- --
q. 0 0 0 ~ 0
-- -- -- --
0 0 0 0 r I ........... 40
c.) Among the space vectors so far all but two were found to be
tensors. It will be shown now that the remaining two vectors, L a{1difJ jdt
which includes the induced voltage and r aB, 'Yiai{1 which includes the gene-
rated voltage are not tensors, neither are di a, di ajdt or r a{1, 'Y and their
law of transformation does not follow the simple formula of equ. 36.
In other words the space vector, say di a has different lengths and different
positions when measured from different sets of coordinate axes, although
the vector i a is an invariant,
. = d'1, 1r cm
d lm + . ac,; dX
1T" - t I ••••••••.••••••• 42
ax'
11'"
The formula shows that the Ilew vector consists of the old vector dirC":
plus an additional vector which is a function of the instantaneous velocity
etc. If both old and new coordinate axes are stationary the additional
vector disappears and dim becomes a tensor.
b.) A physical representation is given in fig. 19. Let the current
vector in the rotor change from OA to OB while the axis of the sliprings
changes from 0 to 0 +dO.
The observers on the stationary axes measure the scalar value of the
current before the change along d r as OC and after the change as OD.
Along axis q they measure the change in current as EF. Hence they
construct the di" vector as AB.
The observers on the moving axes measure the scalar value of the
current before the change along axis a as OG and after the change as
138 GABRIEL KRON
01 = OL instead of OQ, since the axis itself from which the measure-
ment is made has also moved and the projection of OB on the changed
axis is 01. It is true that the observers also measure a change of dis-
placement de, but that does not appear to them as a change of current
as measured by an ammeter (it will be shown below that that appears
to them as an additional voltage). Similarly along axis b the change
of current is OH - 01 = HK. Hence the resultant dim as measured
from the moving axes is constructed as AM.
dr
o 8
. ;,'
/ :". .
:/.
.
". .
~
. .. f"': ,.
O~E :: CI.
FIG. 19
d;m di"
~ = -
dt dt"
cm + zO'. ~-
aC'; t'. . . . . . . . . . . . . . . . . ,
ax'
'3
.':1:
NON-RIEMANNIAN DYNAMICS 139
Also
dim ~ a dt" m
Lmk -d = LnC kCm -d C.
t t
+ L ..aC kCmt't' ac,:
.. a"
--
ax'
diE C ...k
L ". -d + L "a c~ca""
I.
aC'~
mt t - - .............. .
44
t ax'
since C~nC': 0; unit matrix.
ek = Rmkl
'm
de + r mn t'm'n
+ L m" dim 1 .l.
r EO', 'lr
r mn, k Cmanc + Lac,: c
f 0'
k
11" ml..- - -
k
1f'
I ••••• •••••• • 46
ax"
This formula like every other transformation formula is true for any
two coordinate systems and is one of the fundamental formulae of the
Absolute Calculus. The first part leaves the operator unchanged, while
the second part again is a function of the velocity and it disappears if
both sets of coordinate axes are stationary.
q. -0- 0 0 0 0
--
0 0 0 0 0 ...... 48
ds dr qr qs
ds 0 0 0 0 0
-- -- --
dr 0 0 - Lrq - ill q 0
-- -- --
aC"', e l.. qr -Md L rd
-- - -
0 0 0
Lin!. ax t ~ qs --
0 0 0 0 0
-- --
0 0 0 0 0 .... .49
ds a r qr q.
I~'I
ds I 0 0 r
I - .'IJ"
.
I
r 0 0 Lrq - L rd
I'mt. k cmGk d
qr - Jfd Lrq - L rd 0
~
qs 0 JIq 0 .. 50
,..
ds dr qr q.
• ds -0- 0 - I1Id 0
--
r d r -0- 0 - L rd 0
ft,T."
qr -0- Lrq 0
--
0
q. 0 Jlq 0 0 .......... 51
b.) If r.t. ~ is assumed as an operator in the old coordinate system,
the same value of r mt. k given in equ. 15 is found by the use of equ. 47
that is by
r rn t.k = rd. ~ C:, C~ + Lo ac~ C~ .. ............ 52
ax
The recalculation of r mt. k serves as a check on the correctness of
r .t.~. That is
d. a b
m rl. - 0 - Md sin 0 - Md cos 0 0
-
n 0 (L rq - L rd ) sin 0 COl" 0 - (L rd cos 2 0 - Lrq sin 2 0)
-
0
-
b 0 Lrq cm;20 + L rd sin 2 0 (LTd - Lrq) sin 0 cos 8 0
- -
q. 0 Mqcos8 - Mq sin 0 0 ...53
k
a b
m d. 0 0 0 0
a - Md sin 8(Lfq - L rd ) sin 8 cos 8 L rd sin 2 8 + Lrq cos 2 8 Mq cos 8
b - Md cosO - (L rd cos 2 0 + 1-, q sin 2 8)
--
(LTd - L Tq ) sin 0 cos 8 - Mq .~inO
NON-RIEMANNIAN DYNAMICS
q, 0 0 0 0
~-- ---- --------------~--
,i'54
141
142 GABRIEL KRON
the sum of these two matrices gives the matrix of equ. 15.
c.) The value of r.~. I is found from
= rmn,l C'; C: .................... 55
r H • 1
qT -Md/ 2 (L Tq - L rd )/2 0
--
0
q. 0 Mq/2 0 0 ... 58 ------
150 GABRIEL KRON
b.) Since the applied steady-state voltages along the direct and quad-
rature axes of the armature are constants eJr = e sin 0 and e qr = e cos 0,
the steady-state impedance tensor of the group is found by making p = o.
Hence (since pO = Vw = wand wL = X)
d; d. dr qr q. q;
d; r8d - 0
I
- X;"a sin 0 X;"d cos 0 0 0
--
d. 0 r.d 0 - Xmd 0 0
-- - - - --
dr 0 0
rr + r;
- sin o. - X rd - X;a cos2 0
0 0
cos 0 (X;q - X;a) - X;q sin2 0
= - - -- --
rr + r; + sin o.
ZaP
X rq + X;asin20 +
qr 0 0 0 0
X;q cos2 15 cos 0 (X;q - X;a)
- - -- --
q. -0- 0 X mq 0 r8 q 0
--
q; 0 0 - X;"q cos 15 - X;"q sin 15 0 r,o
I
The sum of the two matrices 57 and 58 is the matrix of r.t, as 'If
given in equ. 51. Hence both definitions give the same voltages.
That is r(.t) , "i' = r(t,) , "i' = r,t, "i'.
3.) r: u, t is identical with r ,u, t given in equ. 56.
It should be noted that r: t . " is not equal to r tu , .. or in other words
r ,u, " is asymmetrical in the indices E and (j in either definition of r mn, k.
e.) With the definition of section VIe matrix 51 would have been
denoted as r I,. " instead of r,t, ".
XVII. The Coriolis Voltage
a.) An interesting physical interpretation can be given to the trans-
formation formula of r <>/3. 'Y in equs. 46 or 47 and 55.
Let the Equation of Voltage (equ. 17) be set up for the represent-
ative machine with stationary coordinate axes
_',
e" - R",t + L", di'
dt + rEI. "t t
','t
. . . • . • . . . . . . • . • . 59
b.) The observers on the stationary axes measure two voltages due to the
presence of flux lines:
1.) An induced voltage Lndi'/dt = d<p .. /dt in all stator and rotor
axes assuming:
a.) the rotor conductors stationary,
b.) the currents varying.
2.) A generated voltage r,l. "i'i f = itt"i t in all rotor axes assuming:
a.) the currents unvarying,
b.) the rotor conductors cutting the resultant rotor flux-density waves
with a velocity it.
c.) In changing over to any machine in which one rotor layer has
moving coordinate axes, not the whole length of the previous vector
L ..,di'/dt appears as' an induced voltage, only a part of it since the
moving observers measure a smaller d1'm (section XIV). The decrease
of the induced voltage vector appears however as an additional gene-
rated voltage vector. That is the mOV1'ng observers measure three
voltages:
1.) An induced voltage in all stator and rotor 'axes, assuming:
a.) the rotor conductors stationary,
b.) the coordinate axes stationary,
c.) the currents varying.
This voltage is smaller than the corresponding voltage measured by
the stationary observers.
144 GABRIEL KRON
The first term on the right hand side is the induced voltage measured
by the stationary observers, but expressed along the moving coordinates
by C~. The second term is the decrease of the induced voltage.
2.) By equ. 52 the total generated voltage is
r mt,k~.....I / r ./."C'C"'m'I+L
m k~ [
aC:' Crk~'m['l. . . . . . . . . . . 61
"'--ax;
NON-RIEMANNIAN DYNAMICS 145
The first term on the right hand side is the generated voltage vector
measured by stationary observers but expressed along the moving
coordinates by C~. The second term is the Coriolis voltage, the
increase in the generated voltage.
e.) The decrease in the induced voltage is equal to the increase in
the ~enerated voltage, that is
L C"C' aC:
... " "' ax l
','1
1 I
+ L...C.. aC,~
ax 1
k l
'm I
1
= L ...C,,7
.. ·'·I7 (ac:
ax t C·m + aC:'
ax cm) =
l '
L ...C"','I
k1 1
a(C:C:')
ax = 0t
since C:C:, is the idemfactor and its derivative is zero. Hence from
equ. 56
dim
Lmk({i +r 'm'l
ml . k 7 1 = L"'dt(di' + .'1) C . CI.,..t'Z k · •••••••. 63
CONSTANT SPEED
ea = R afJ7'fJ ;t + rpI.aillil
+ LafJ d'fJ
= (RafJ + LafJp + rill. ail) i B
, e-:~nZ-::;ii ,. . . . . . . . . . . . . . . . . . . .... 65
from which the current is found by calculating the inverse of Za~ = yap
fJ
1i = eayafJ 1·······················66
yafJ may be called the "admittance matrix." (See section XXXId.)
b.) For the representative machine with stationary coordinate axes
the impedance matrix is
d. dr qr q.
d. r.d + L.dP MdP - MdP(J 0 I
ali -- Z C"C~
a 'fJ + LH fiX! a 1 I• • . . • . . • . • • • • • • • • 68
ac~ C~·/
IZ U
as follows from equs. 64 and 46. If no slip-rings exist the last term
disappears and ZafJ is a tensor. The torque is byequ. 19
d.) When any number and any types of rotating machines are con-
nected in any manner whatever the transient impedance of the group is
equal to the sum of the transient impedances of the individual units,
transformed by the connection tensor of the group C~. That is
(c jd) = ac bd
_ e-2X2m sm
0 a (r r0sm a - X r cos a )
r orque -
(rr + rsd + Xm COS at> - XsdXr + rrXsd + rsdXr + Xmsm a, cos av).-
2 2 )2 ( 2 0 0)
0 0 -M~ (cos ope' + sin op) -M~ (sin ope' - cos op) 0 r;q+L;qp
149
f q.
2 L; sin ap -
(r; L; p)
+ (r.
+ L. p) n
+ + L; cos ap +
f M sin ap
2Mn cos ap - Mn sin ape L; sin apO +Mnp L; cos o'pa - Mnp
Mnp L; (cos ap - sin ape)
+ r~ + L~p - L~pO 0
L; (sin ap cos apO)
+ L~pO r~ + L~p Mnp
Mn (sin ap cos apO)
+ MnpO Mnp (r.+ L.p) n 2
NON-RIEMANNIAN DYNAMICS
151
152 GABRIEL KRON
f a
ds f a
d s rs + jX. jsXm 0
Za~ = f jX m (rr + jsXr) + (1'; + jX;) (r; + jX;) ~,a
a 0 (1'; + jX;) ~,a r; + jX;
NON-HOLONOMIC DYNAMICAL SYSTEMS
_ R 1'(j
ea - a13
+ dt
d<{'a _
- R 'Il
a13 1
+ --dt-
d(Lalli(j) -
-
R
a(jl
'13 + L a/3 dill
Iii + dL ail
dt '131
1 .. 72
154 GABRIEL KRON
is valid. This equation is not valid for any other axes since it gives
only the induced and Coriolis voltages, but it does not give the generated
voltage due to the motion of rotor conductors. Any attempt of routine
substitution fails, which immediately can be seen in case of machines
with stationary coordinate axes where dL,,~/dt = ilaL,,~/axl = 0 as
mentioned at the beginning of this section.
This fact has not been recognized before and this is the reason why
numerous treatises and papers on rotating machinery start with
Hamilton's, Lagrange's or Maxwell's equation of the usual form, but
use them only for the alternator with moving coordinate axes and for
no other machine. They all are compelled to analyze the complicated
physical phenomena inside of each type of machine and even in the
alternator if its moving axes are changed to stationary axes along the
field pole and the interpolar space, since the routine substitution, which
after all is the whole purpose of all generalized equations, fails.
b.) In the following sections r ,,~, "'( will be defined in terms of the
metric tensor for all coordinate systems. The three matrices of r I~, "'(
r "I, "'( and r ,,~, I will be found to be defined slightly differently from the
matrices defined previously, but again the voltages and torques calcu-
lated from them are identical with those found by previous defini-
tions of r ,,~, "'(.
Also in the next three sections all rotating machines will be con-
sidered derived from the representative machine with moving coordinate
axes, that is their connection tensor C'; will be the product of the con-
nection tensor C: given in figs. 2-17 with that of the representative
machine with stationary coordinate axes C';. In other words CO; will
be equal to C:C';.
fk = R mkX·m + amk
dxm
-d + (aa
-
mk 1 aa
- 2- -k
mn ) ' 'm 'n I
X X .... , ...
73
t ax" ax
_!aa"8cac8·,,cm·'Yc"
2 axk m nX "x 'Y
_!2• aaa aaxk
C a
mCa."cm.'Yc n
nX "x 'Y
- 21 aa8 ac!
axk ca.
mX"C"X'Y
m • C"
'Y .................. . 74
But
idt (aT)
axv
_ 0'1'V + 0'1' (act _ ac!) CkCnj;'Y
ax or ax" ax" v 'Y
+ axv
of = f,
v
I.... 76
This is the modified Equation of Motion of Lagrange valid for non-
holonomic dynamical systems as given by Whittaker in his "Analytical
Dynamics," with the notation slightly changed and aF jax v added.
It should be noted that in equ. 75 all aafJ and are functions of the C:
old coordinates and in the n dynamical equations there are 2n variables.
They are the n old coordinates and the n new differentials or velocities.
The other n equations are dx v = C'kdx k •
156 GABRIEL KRON
f = R XE· + dx,
-d + (r mn,k cmcnc" +
a amI. -ac~u ck) XO'"XE.
..
t
11" E1r a1l'"E E 11" 11'"
Bx
Even though in the general case of non-holonomic systems the expres-
sion in parenthesis is not a function of the n new coordinates but of
the n old coordinates and the n new differentials, still a geometry can be
defined, the so called "non-holonomic" geometry in u'h1'ch the expression
in the parenthesis plays the role of the "coeffidents of connection."
Schouten defines
di'
a"., dt + A,u. ".XUX' + R",x' = f". I ...•....••.... 78
I
XXIV. The Equation of Motion of Quasi-holonomic Systems
a.) Due to the special character of the coordinates of electrical machinery,
equ. 75 and 78 assume special forms. All L", and C~ are functions of
one variable xt which is both an old and a new coordinate and so in the
n dynamical equations there are only n unknowns, the n new coordinates.
Hence equ. 75 and 78 must be identical with the Equation of Motion
as derived in equ. 10 for any coordinate system, that is it must be
identical with
It 1'S important to note that T aoyu is skew-symmetric in l' and (]' that is
I T-yau = - Tua-y I....................... 82
Hence the generalized Christoffel symbol can be eJ:pressed as the sum
of the ordinary Christoffel symbol and a tensor of rank three,
PHYSICAL INTERPRETATIONS
TIM = La! aC l
ax
t C~ . .......... , .......... 86
158 GABRIEL KRON
101_1I
dr 8 qr
acz C k =d T
ax! .. qrl~ 0
NON-RIEMANNIAN DYNAMICS 159
I ~a = Ra~t.~ + dipa
dt
+?t I. a
it I
•••••••••.•.••••• 95
I ~a - Ra~t.~ + dIPa
- (It
+ 1/;la i t I................ .96
160 GABRIEL KRON
et Rltil :t
+ Lit d' + T -ya1i-yi" ....••..•....•.. 99
l
That is
r-ya. I T-yal ....................... 100
CTO~q~~ = ififJti
fJ
I·················· .101
~ON-RIE:\IANNIAN DYNA:\-IICS 161
eaLa. = R"
aaLaL" + L a" di- a t". + [ta, (J 1'1"
l lal" + T laallala
'1" . .... 102
at
The time rate of change of the stored magnetic energy for machines
with moving coordinate axes is (if a and (3 are stationary, m and n
moving axes)
dT _ 1 d(La8C:C~imi") = L dim in
mn "t
+ ~2 dLa8
rtt
iai8 + La8 aC: iminit .. 103.
ax t
-
dt - 2
- rJ
Ht "
The last term is the power due to the Coriolis voltage, the previous
term is zero and
Tt
dT = L aa Tt
a
di la. + [ta, (J
1'1"
l l"la . .............. 104
e"l". = R"
a"lal" + at
dT ................... 108
162 GABRIEL KRON
a.) It should be noted that there are several definitions of r afJ,1' each
definition giving different matrices for Pill, 1', rOIl, l' and r a,3, / but still the
same final voltages and torques.
In defining r afJ, l' from the standard Lagrangian equation (sections V
and VI) there were four different definitions of rap, l' according to the
definition of the Christoffel symbol with two (equ. 11) or with three
terms (equ. 8) and ac~ording to the order of the indices in defining
them as r mn, k or r nm, k.
In defining r afJ,1' from the generalized equation oj Lagrange (section
XXIV) the same arbitrary definitions can be selected but even so the
three matrices of r afJ.1' are different from those above.
l.) The rotor generated voltage may be defined as r lll ,1'iiJ (equ. 91) or
r a', l' (equ. 51) with matrix 51 or as the sum of two matn'ces 57 or 58 when
defined as rUIl>, l' in section Yrd.
2.) The torque may be defined with matriJ: 51 as given in equ. 98 by
T 1'01/ or with £ts symmetrical part, matrix 56, since the torque is a homo-
geneous quadratic form and the skew-symmetrical matrix gives zero
torque
3.) The Coriolis voltage may be defined as [tl3,')'] or [at,,),] or [(t13),')']
by equ. 8 or 11.
b.) Also it should be noted that the transformation formula 46 could
have been defined also as
d (a!)
dt ax"
a!, act
+ aX" axn
c:c~x'Y + a~ =
ax"
e... ....... , .. 114
where the first term on the left-hand member represents now only the
induced and Coriolis voltages while the second member represents the
rotor generated voltages. The extra term appears however only in
the equations of the axes stationary on the rotor. For all stator axes
and for all axes moving with the rotor the generalized equation reduces
to the standard Lagrangian form.
The Equation of Torque is
d~t (a~)
ax
- a~ - aa~
ax
ac~ C~x'Y + ax
x ax
a~ = ft ... ....... 115
The second term on the left-hand side is identically zero (equ. 97) and
the machine torque is given now by the third term. It is interesting
to note that although axis t remaines unchanged still the original equation
is not valid for it only the generalized one.
Equ. 107 shows that the power due to the additional terms in the
generalized equation is zero. Hence the additional terms may be called
"gyroscopic forces."
c.) These interpretations are in conformity with the rule given by
164 GABRIEL KRON
r~,3 = n.C"
aC'C1'
,3 (f
+ 8Xfi
aC: C1'
(f I.............. 117
I
e" = R~,3ifJ + di"
dt1'
+ r~ ifJi1' I
... ............ 118
ds a " b ~
",c!, 0 M4si"e/2L
M. cose/2L 0
t a ~si"e/2L Lt. -L,,)(cos"e-si,,20)/2L -M,coselzL
jL,.-L.,).siYle CGse/L
rll"'''"b M4cos 6 12L (L"I-L~coS2e-5i\'l2e)12L (L!.!.- L,ils'llI I) coselL M\ si." 912L
Ojs 0 -M, cos6/ZL M,si"e/2L 0
THE CONTRAVARIANT OPERATORS THE REPRESENTATIVE MACHINE or
WITH MOVING COORDINATE AXES
ds d, 0 q. q, t J d, f1 'I, qs t
ds LtoI tA ~.. IA 0 0 0 ds rsdl,./A -',M4/A 0 0 0
It d, LM~1l Lsd /A 0 0 0 d. -r.dM./A r.lsd/A
'If 0 0 0
LII'·. Cj. i 0 0 L,~/B -Mq/f 0 R~IT= II, 0 0 r. L~!./B -!i9.~1B 0
0 0 M,/B L,~/B 0 lis 0 0 -.!..~B !!\L~8 0
'Is
t 0 0 0 0 IlL i 0 0 0 0 tIL
'If
ds eI, 'If 9' II· d, dy q. ~
d) 0 0 M.L•• IB MdM~/e d, 0 £ Md/2L 0 0
rt
f
r",,= 9,d. 0 0 Lr~~ ~L.vS d, 0 0 (L,rl~I2L ~..!:
,. M.iL,.tA .•,l,. /A 0 0
0
r"= II, ~ZL L,rL,,)IZJ
0
0
-M,/2L
0
0 0
"
N,.MjtA. M,LsdtA. 0
TABLE IV
166
NON-RIEMANNIAN DYNAMICS 167
impressed while all axes are short-circuited. Also while the metric
tensor is the measure of the permeances of the various magnetic circuits
with the windings open-circuited, the inverse metric tensor in the measure
of the reluctances of the magnetic circuits with the windings short-circuited.
In the first case the magnetic lines follow paths with the maximum
possible permeances, in the second case they follow the paths with the
minimum possible permeances.
d.) Each term of the inverse of a dyadic as Lap is found as follows:
1.) Replace each term by the determinant formed when the corre-
sponding row and column are removed.
2.) Multiply it by plus or minus one according as the term occupies
an even or odd element counted from the upper left hand corner.
3.) Interchange the two indices.
4.) Divide it by the determinant of the dyadic.
dia dia
Oi"
ot _ + r f3-y zf3. dx-Y
dt
a
dt -- dt + r il'Y··
a
i f3 ;-y •••....... 125
168 GABRIEL KROK
L a {3Oi{3/ ot also aIP,,/ ot represent all the voltages due to the presence of
flux-lines, that is the sum of the induced, generated and Coriolis voltages.
c.) The absolute derivative of the metric tensor is zero. That is by
equs. 124, 82 and 8
circle) or any other quantity. But viewed from the outside space the
difference between them is something else, dA, a definite quantity.
The relation between the two differences is
oA" = dA" + r~'YAIl dx'Y ................. 131
where r~'Y is any arbitrary number system of rank three (see equs. 15
and 16). Each coefficient is a function of the variables dx" and is
called the "coefficient of connection."
c.) If r~'Y 1'S defined in terms of the metric tensor as
which shows that the above definition of r~/l is equivalent to the follow-
ing definition of the inverse connection tensor
C-; = nLI',Lh . ..................... 134
C-; can be taken as the inverse of C; since
C~C: = C:n LI"£O"( = L"oLh = o~ ............ ]35
where 0; is a unit matrix with m unit coefficients in the diagonal.
xxx-v . .von-holonomic Sub-spaces
In many cases in which the number of new coordinates is less than
the number of old coordinates the transformation tensor C: can not
be written as ax" / ax". Such a case occurs for instance in the single-
phase alternator fig. 3 where axis b should be removed. (In general
moving coordinate axes form non-integrable transformation tensors.)
Lately the expression "non-holonomic sub-space" has been introduced
to denote the restricted positions of the particle.
For both types of sub-spaces with each point of the n-dimensional
space there is associated an m-dimensional infinitesimal surface element
and the particle must leave that point along the surface element. But
while in a true sub-space the surface elements may be combined to form a
family of surfaces so that the particle can never leave the particular surface
it happens to be on, in a non-holonomic sub-space the surface elements
can not be combined into a surface and the particle may pass from anyone
point of the n-dimensional space to any of its other points.
To visualize the problem, in a repulsion motor, fig. 6 which is a true
two-dimensional sub-space in a three- (or four-) dimensional non-
Riemannian space no current (or rather no charges) can flow along
NON -RIE:\IANNIAN DYN.nnCS 173
a b c
a ra + Laap Mohp MacP
Z"f3 = b MbaP 1'b + Lbbp Mbcp
c MoaP Mcbp rc + Leep
NON-RIEMANNIAN DYNAMICS 175
HUNTING
e" = R ".
.~fj
,I'
+ di"".,
-dt r.,,-Yzfjt-Y
changes to
de" + ae" dx
~ - R"·ri'fj
fj - ,f1U~ + dR" + d(di") + rex d'fj' + rexfj-yt'fjd' + dr"fj.,l'/3'
,fjl'fj ~ B., 1 ~-y 1" ~., ,137
In covariant form it is
ll
de", + ae" dxll = R.fjdill + L"fj d(ddi ) + dL"fj ddifJ + r/3.""dill i-y + rfj-y,,,illdi" + drB-y,,,illir
axil t t [.. 138
Equ, 137 is equivalent to
or to
de a = d(R af3 i (3
) + de:t ) ................. 140
a
dea + ::; dxfJ = RafJdifJ + d d~a + d1/;t. ait + 1/;t. a dit . ..... 142
det = R tt di t + L tt d(:t) + r p1'. t diP i1' + r p1'. t i 13 di1' + drfJ1'. t ifJ i1' .. 143
In terms of space vectors the equation is
2.) dl/;." ti" = r Il." tdilli" + dr Il." tilli" due to change of 1/;." t and to
original current.
Among the fluxes:
A.) The change in the flux-linkage vector dcp" = d(L"llill) can be
divided into illdL"1l and L"lldill
B.) the change in the rotor flux-density vector dl/;." t = d(rll." till) can
be divided into drll." till and rll." tdill
C.) the change in the Coriolis flux-density vector is d(~, t" - ~'Y' t).
In the equations occurs only dif;t,,, = d(r llt , "ill) = drllt, "ill + rllt, "dill.
To each of the six flux changes corresponds a change of voltage or
torque. The three changes due to dL"Il, dr i3'Y, t and dr itt,,, can be
measured only by the moving observers. Hence for a machine with
stationary coordinate axes the Equation of Small Oscillations reduces to
de" + ::~ dx ll = R"lldill + L"1l ~: + rll'Y, " dill i'Y + rll'Y, " ill di'Y .. 145
A more detailed physical discussion is given in the previous papers.
XL. The Motional Impedance
a.) Let any machine with stationary coordinate axes and without
sliprings be considered. If at the instant of disturbance the machine
is in equilibrium, that is if all its currents are known, the Equation of
Small Oscillation, equ. 103 can be written as
de" = (R"fj + L,,{3p + rll'Y, "i'Y + r'Yll, "i'Y)dill.
The expression in parenthesis is a tensor of rank two Z"1l called the
"motional impedance tensor" representing the opposition of a machine
to a suddenly applied infinitesimal terminal voltage or shaft torque,
allowing its speed also to vary, that is
I Z"r; = R,,1l + L"IlP + rll'Y, "i v + r'YB. "i'Y I.......... 146
Comparing it with the "transient impedance tensor" Z"it equ. 64, it
is found that Z,,1l is only a special case of Zall, that is
Z"1l = Z"fl + rll'Y, ti'Y + r'Yt. "i'Y + r'Yll, ti'Y ........ ... 147
Z"r; differs from Z"1l only by an extra row and column corresponding to the
additional axis t.
The Equation of Small Oscillations becomes
1 de" = z"lldilll· ....... " .... , ...... 148
from which dill is found as de"f""r; by calculating the inverse of Zall'
For the representative machine
178
d.
d. rsd + L.dP Mdp - MdPf) 0 i qr Md
a
dr Mdp rr + LrdP , - LrdPf) 0 i qr (Lrd - Lrq) - i q• M q
Z.,~ = qr 0 Lrqpf) rr + Lrqp Mqp i dr (L rd - Lrq) + id'Md
qs 0 Mqpf) Mqp r.q + Lsqp - i dr Mq
GABRIEL KRON
dr a a+ aC~ ca)if3CEi'YCX
a
• xi'{1>. = d(r f3 'Y Cf3C'YC
X a
E ax a f3 'Y
= [(dr fa3 'Y ) Cf3C'YCa
• X a
+ ac:
ax
dC!] if3C~i'YC~ . ..................... 155
d.) It is necessary to have on the right hand side Oi" (or rather dx a )
as the independent variable. Hence it is necessary to change
1.) the tensor a(Oi"/at) into the sum of two tensors so that one of the
tensors should be (a/at)Oi"
2.) the tensor a(R'!~~) into the sum of two tensors so that one of the
tensors should be R'!~i~.
XLIII. The Generalized Riemann-Christoffel Curvature Tensor
a.) In order to find the difference between a(ai"/at) and (a/at)ai" let
each be developed in its component parts.
a (Oi,,)
at
= d(ai")
at
+ r~... "1
Oi~
at
dx'Y = d di" + dr:! ifJi'Y + r~ dilii'Y
dt ... "1 ... "1
d'fJ
+ rp'YifJdi'Y + rp'Y ;t dx'Y + rp'Yr~ ,ixi'dx'Y
fJ
~.
-a u~" ~. + r"{J u~
-d u~"
= dt •....Q.~'Y = -~"
d d' + drp'Y 'Qd + r "
--~. . x'Y ~ -
di dx'Y
at "I dt dt "I dt
·· ,,-
K hfJ - ar
--
p•
ax'Y - ar;;'Y
------:::t:'
a;x,-
+r" rio.
h-y~. - ~'Y
I
r"U rX ....... . 157
Hence
R~~a == alSaxRM1-r>'~RxfJ
ax"
- R"'xr~" .1 . . . . . . . . . . . 159
It can be seen that the equation in invariant form consists of the previous
equation 137 with additional terms added and the same terms subtracted.
In equ. 137 the terms could not be grouped to form tensors but in the
presence of additional terms representing hypothetical voltages and
torques groups can be formed so that each group of terms is a tensor.
b.) In calculating all the possible coefficients K8~~a the following
should be noted
1.) K~~~a = - K~8fJa. This is the only skew-symmetry that occurs
in the generalized curvature tensor.
2.) a is always t. That is the curvature tensor is zero in the voltage
°
equation.
3.) Either or 'Y is t, that is only K i ~ ~ I and K ~ i ~ I exist where also
K ~ i ~I = - K i ~ fJI. Hence all possible coefficients can be represented in
a matrix form Kj~fJ' where only 'Y and f3 vary. All the other (n4 - n2)
coefficients are zero.
Hence the Equation of Small r oltages in invariant form is
a a
oe a + ae dX fJ
axfJ -
-
Ra.fJOI'fJ + o(oi
----u-) + R'"Yi3'a'fJd
~ X "Y
•••.•..••
163
• t --
ue R .t luI
.'1
-u + K ~"Y..fJ I'~1 ~'fJd'X + R
+ oCOit) "Y
"Y fJ'/'fJd
I X "Y •••••.•• 164
hand side of the above equations does not behave as a tensor, that is the
transformation formula of K i ~~ t is
t
axt = (r k n rkm t _ ar!n)
rt"" r'""t _ ar;" axt cmcn _ rtmn ac~
axt C" •• 167
m
""
or
K · .. t -- K'tnm
.. t - r mn ac n
t - t" Cm 168
t(J7i"
ax 11"0 • • • • • • • • • • • • • •
The coefficients of K ;,;.~t are the same as those of K tmnt each divided by L.
~
00
00
188
d. qr
- MiL8q/2B - M dL rd L. q/2B 0 0
- MdLrdL8 q/2B Lrq/2 - Lr/ L. q/2B 0 0
0 0 L rd /2 - Lr q2 L8 d/2A - MqLrqL8d/2A
0 0 - MqLrqL8a/2A - M q2 L8a/2A .. 170
d.
d. - Md 2L.q/2B - MaLrdL. q/2B 0 0
dr Md/2 - M dL rd L. q/2B Lrd/2 - Lr/ L. q/2B 0 0
K tud =
qr 0 0 Lrq/2 - Lr/ L. d/2A M q/2 - MqLrqL8d/'2A
q. 0 0 - M qLrqL8d/2A - M/L ed /2A .. 171
NON-RIEMANNIAN DYNAMICS 189
DIFFERENTIAL GEOMETRY
and the equation of the disturbed path is (replacing o(R~ fJifJ) by its two
components)
As long as the metric tensor of the surface L",{3 and the coefficients of
connection r "'fJ. 'Yare defined in any arbitrary manner, 1't is not necessary
to know the relations between the variables x"'. All (metric) properties of
the surface and of vectors on the surface in the neighborhood of a point
can be studied with the aid of the two sets of quantities LafJ and r afJ, 'Y
one having n2 the other n 3 terms.
NON-RIEMANNIAN DYNAMICS 191
- dx" - dx"
v"=~'" -Tt-dsdt- ds
- = " .... .......... . 182
Xv .
The projection of the velocity vector along the coordinate axes repre-
sents the instantaneous currents flowing in each winding. The magni-
tude of the velocity vector is
That is the square of the velocity vector is equal to twice the kinetic
energy stored in the machine at that instant.
d.) The acceleration vector e" can be divided into R"fJifJ which is in a
general direction and into 5i" /M
oi"
ot
= d(X"v)
dt
+ r~ "Y XiJX"Yv2 = (~X"
dt
+ V2dX")
ds
+ v2r~ "Y XiJX"Y = <!:!!.X"+v2hp.a
dt
where
d)." ax"
hp." = ds + r~"yxllx"Y = 58··············· .184
hp." is the "first curvature vector," p." is the "principal normal" and h
is the "first curvature." Hence the voltage vector ai"/at (acceleration
vector) due to the presence of flux lines lies in the plane of X" and p.".
e.) The power is represented by the product of the acceleration vectors
and their projection along the velocity vector X". Since the power
represented by T"fJ"yi"ii3i"Y is zero (equ. 107) the generated voltage vector
due to the motion of rotor conductors and the torque vector must lie
in a plane perpendicular to the direction of motion that is to X".
Hence the product of ai" lot and its projection on X" gives dT / dt while
the product of e" and its projection on X" gives the power input.
192 GABRIEL KRON