Documente Academic
Documente Profesional
Documente Cultură
Engineers
http://pme.sagepub.com/
Published by:
http://www.sagepublications.com
On behalf of:
Additional services and information for Proceedings of the Institution of Mechanical Engineers can be found at:
Subscriptions: http://pme.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://pme.sagepub.com/content/186/1/755.refs.html
What is This?
The dynamic performance of a hydrostatic transmission with a variable delivery pump used for speed control
is predicted using a simple mathematical model. The analysis is then extended to take into account the trans-
mission slip and torque losses, and the prime mover droop by employing signal flow techniques. The results
were compared with those from an extensive test programme employing electro-hydraulic test techniques
carried out on a typical hydrostatic drive.
The signal flow analysis improved the prediction of the transmission dynamic performance, but it was found
that errors of up to 40 per cent were occurring, as many of the loads employed could not be accurately repre-
sented by simple mathematical models. A vector approach was adopted, using measured load loci that provided
a correlation of within 4 per cent of the experimentally measured response, by using a value of oil bulk modulus
midway between the isothermal and isentropic tangent values.
Mean return line pressure, restricted boost system and flexible pipeline effects were investigated.
Torque available to drive pump. pipe resilience, and most experimenters have allowed for
Torque required to accelerate motor and load. this by using an arbitrary value of effective bulk modulus
Supply line pipe wall thickness. which gave good agreement between predicted and experi-
Volume of oil in supply line including the dead mentally obtained values.
volume of pump and motor exposed to supply The determination of the bulk modulus of hydraulic
line. oil has been the subject of papers by Hayward (6)(7)
Fraction of motor displacement. and other writers, and because of the non-linear change
Fraction of pump displacement. of volume with pressure and the effects of heat inter-
Engine governor setting. change, it is necessary to consider, on the one hand,
Damping ratio. whether to take the secant value or the tangent value and
Time constants. on the other whether to assume that the compression or
Engine and pump rotational speed. expansion is isothermal or isentropic. In addition, in
Motor and load rotational speed. practical systems, the effect of wall elasticity has to be
Frequency of sinusoidal swash plate variations. taken into account, and the effective bulk modulus is
Transmission and load natural frequency. given by the well-known relation
Lower case letters throughout the text refer to small varia- -1_--+-
1 1 . . .
tions about a steady state operating condition, unless Be B o B ,
otherwise stated. 1
where - = d/tE
BP
2 THEORETICAL C O N S I D E R A T I O N S
For example, Hansen (8) used an effective bulk modulus
2.1 Bulk modulus Be of 1 1 . 6 ~ 1 0bar
~ ( 1 . 7 ~lo5 lbf/in2) to predict the
In the reported experimental work on hydraulic systems, dynamic response of a valve-controlled hydraulic motor
there have been many attempts to obtain a reasonable system. This gave an overestimate of natural frequency.
correlation between the experimental and predicted values Lambert and Davies (9) and Renner (10) used values of
obtained for system natural frequency. The main un- 7.75 x lo3 bar (1.14 x lo5 lbf/in2) and 6.8 x lo3 bar (1.00
known has been the effect of the oil compressibility and the x lo5 lbf/in2) respectively, and both obtained reasonable
1
1
2' xM -KF M
agreement between their predicted and experimental teristics of the various transmission elements. For example,
values. Aeration and mounting compliances were suggested the typical speed-torque characteristics of an engine for
as possible causes of the low values of bulk modulus. various governor settings are shown, together with the
Healey and Stringer (11) used a value of 1 8 . 4 lo3
~ bar linearized equation relating small changes in speed with
(2-7x lo5 lbf/in2) and obtained good agreement by taking the engine torque variations and governor settings. The
into account the effect of oil inertia in the pipelines. quantity SSZJST, (= KE), the engine speed droop, shows
More recently Keating and Martin (12)used a value of how the engine speed changes with torque and the
6.80 x lo3 bar (1.00 x lo5 lbf/in2) and obtained reasonable quantity 8QJS Y (= KG), the governor speed control con-
agreement between their theoretical approach and their stant, shows how the engine speed varies with governor
experimentally obtained values for natural frequency. setting. Delays in development of engine developed torque
Many workers in the field including Walters (13)~Gille after a change in governor setting have been neglected as
(5) and Shearer (2), suggest a general value for B to be relatively insignificant.
taken as 13.6 x lo3 bar (2 x los lbf/in2), despite the fact The hydrostatic pump is dealt with in two parts. First,
that this value is considerably lower than that determined the linearized relation between flow, speed, swash plate
by static tests. Shearer stated ‘The logical basis of the angle, and pressure is obtained. The quantity SQp/SPH
practice is open to question, but it seems to work’. (= Kp)is the pump slip coefficient. SQp/S Y p(= &Q,) is
the change in flowlunit change in swash angle, and
SQp/SQ, (= DIXp) is the flow change/unit change in
2.2 Transmission transfer function pump speed. I n a similar manner, the change in pump
Simple linearized analysis can be applied to the transmis- torque for changes in speed, swash plate angle, and pres-
sion shown diagrammatically in Fig. la. Assuming it has a sure are obtained. STp/SP, (= DIXp+Kfp)is the pump
constant speed prime mover, no losses, and constant re- displacement plus the pressure dependent torque co-
turn line pressure, the following transfer function is ob- efficient, STp/SXp(= D1PH) is the change in pump
tained : torque/unit change of swash plate angle. STp/SQl (= Fp)
(DZXM)2Be is the pump viscous torque coefficient. A constant return
line pressure has been assumed.
An overall transfer function can now be obtained by the
use of the signal flow techniques developed by Mason (14)
and applied to oil hydraulics by Bowns (15)and Korn (16).
a second order equation where The right-hand column on Fig. 2 gives the signal flow
graph appropriate to each of the individual equations.
The signal flow graphs can be combined to obtain a
complete signal flow graph for the transmission system;
using Mason’s Rule, the transfer function relating any
two system parameters can be readily obtained. The signal
flow graph and the transfer function relating output speed
Shearer (2) used Newton’s linearized analysis (I) and to input swash plate position is also shown in Fig. 1. The
carefully discussed the assumptions made. He pointed out engine droop increases the order of the equation and is
that there are lower limits that can occur in line pressures, also responsible for the additional terms in the numerator.
and that these should be avoided for linearity. Newton’s A lumping of parameters in the equation of Fig. 1 gives
analysis assumed a constant speed prime mover, and took a transfer function of the following form:
into account the losses within the transmission. The return
line pressure was considered to be constant. The natural
frequency W , and damping ratio 5 obtained is as follows :
The effect of varying any one system parameter on the
overall system response depends on the values of the
other relevant parameters, and cannot be expressed in a
general form. However, for a particular system, the effect
of varying any one parameter can be obtained by substitu-
tion of values into the transfer function of Fig. 1. For
convenience a digital computer programme has been
written to substitute values of system parameter into the
above transfer function; it uses the Newton Raphson
2.3 Analysis using signal flow techniques technique to evaluate T~ and hence K, T ~ w,,
, and 5 (see
A fuller analysis of the basic transmission system shown in Appendix 1).
Fig. 1 can be obtained with the aid of signal flow theory.
This facilitates the use of a more complex model. A change 2.4 The effect of a change in system parameters on
in load, a change in engine speed, or a change in swash the natural frequency and damping ratio
plate position of either unit will give a resultant change in The hydrostatic system tested later has been taken as an
the running conditions of each of the units involved. The example and the effect of varying the supply line volume
static and dynamic characteristics of each element are to and the slope of the torque speed curve of the load (fL)
some extent non-linear, and small perturbation tech- has been investigated.
niques can be used if the changes are not too great. Fig. 3 shows that the natural frequency rises rapidly
Fig. 2 shows graphically the static and dynamic charac- with the reduction of supply line volume, and is therefore
Proc lnstn Mech Engrs 1972 Vo I 186 55/72
ENGINE 8i PUMP
INERTIA
PUMP-FLOW
RELATIONSHIP
3'
PI PRESSURE-Pa
PUMP-TORQUE
RELATIONSHIP
TRANSMISSION
tINB
Chrmmsmnm
OR
MOZOR-FLOW
OR
MOZOR-TORQUE
RELATIONSHIP
.a
engines or electric motors. T h e natural frequency and input frequency the load vector is t L / W Z terminating at the
damping ratio remain unchanged and the time constants point X.
T , and r 2 in equation (5) are close together, thus mak- We will now work backwards to the input movement of
ing little difference to the overall system response. For the swash plate, x,.
example with a 5 per cent droop (no load to full load) the Stage (I) The torque to accelerate the hydraulic motor
A
ratio of T 2 / r 1is 1.008 giving an attenuation in the output must be included as the vect .>I-t,/wz. This will be 90" in
at low frequencies of only 0.05 dB. advance of the speed vector w 2 .
Proc lnstn Mech Engrs 1972 Vol 186 55/72
Hence
Hence
P
\ 4P
= D,SLIL
W
- L
converted to a differential pressure vector p d / w Z .The re-
turn line pressure vector pL/wz can be added to this to
L
4 T E S T PROGRAMME
Frequency response tests were carried out with a nomin-
obtain the supply pressure ph/wZ.The analysis can then ally constant engine speed, to determine the effect of pump
be carried forward as above. The case of Fig. 6d with cavi- swash plate angle, motor swash plate angle, mean load
tation could be analysed in a similar manner using the torque and the slope of the torque-speed curve, the slope
describing function technique if thought necessary. of the engine torque-speed curve, load inertia, system
Extensive cavitation may, however, reduce the pump per- volume, mean pressure in the low pressure line, back up
formance, and the mean motor speed may drop. pump capacity, and the low pressure line relief valve
characteristic. Effects of flexible pipes and of pressure
The work presented throughout this paper is based gauges in the supply line were investigated.
upon the use of lumped parameters. The justification for Oil temperature and oil properties were kept as con-
this assumption lies in the fact that all the transmission stant as possible. The oil used was Shell Tellus 27, at a
lines used are very short in relation to the wavelength of temperature of 5O0Cf2 K.
the pressure signals being transmitted. However, the The input variable for the above tests was pump swash
assumption must be checked before the analysis presented plate angle. Frequency response tests were carried out
is applied to any system. with different amplitude sinusoidal inputs.
electro-hydraulic inputs is to carry out a closed loop fre- However, using the digital transfer function analyser, it
quency response test of the electro-hydraulic input servo, is possible to record the response of the electro-hydraulic
and subsequently allow for this after obtaining the overall servo, and the overall transmission, including the servo
frequency response of the system. This method is time system, consecutively a t any frequency, and then by sub-
consuming, and it is not always possible to ensure that the traction immediately obtain the net transmission response.
servo dynamic performance has not been affected by a It is also a simple matter to continuously monitor the
change in system variables, such as oil temperature. servo response, to ensure a constant amplitude of pump
3-
Boost pump
Tank
swash plate angular movement. I n addition to recording load damping coefficient has been increased. Therefore,
the output speed response of the transmission, line pres- the set of results, although valid for the particular engine,
sure variations and torque variations could be recorded transmission and load configuration, lacks generality, and
simultaneously. cannot readily be compared with the theoretical work.
I n conjunction with the digital readout from the trans- The remainder of the tests have been carried out in what
fer function analyser, the supply and return line pressures, is apparently a more artificial manner, to allow comparison
differential line pressure, transmission output speed, and with the theoretical analysis.
pump swash plate angle were recorded, using an ultra-
violet recorder. 5.2 Investigation of load effects
Variation of load damping coefficient was carried out by
5 DISCUSSION OF RESULTS altering the loading system characteristics and holding the
mean swash plate angles, mean pump and motor speeds
5.1 Test with typical load and load torque constant (tests 5-8). The rest results are
Consider first tests 1-4 (see Table l), in which the torque- shown on Fig. 9, and the effects are shown on Fig. 10. The
speed characteristic of the load was set up to represent a apparent natural frequency, as determined from the 90'
typical windage type load. Frequency response tests were phase lag point, alters with the load damping coefficient
carried out at different pump swash plate positions. The from 4.8 Hz at 0.037 Nm/rad s to 6.5 Hz at 0.587 Nm/
load in this case was applied by screwing up a needle rad s. This suggests a considerable increase in natural
valve in the loading pump delivery line. The amplitude of frequency with increase in load damping coefficient. This
the pump swash plate oscillation was kept constant. The is not in agreement with the theoretical predictions, but
test results are shown on Fig. 9. the reasons for the apparent discrepancy are given below.
However, a short consideration of the set-up shows that The damping ratio can be seen to increase rapidly with
many of the system parameters have been altered. For increase in load damping.
example, the system pressure and the motor speed have Tests 9-12 were carried out at reduced amplitude. The
both altered, with corresponding change in the slip and results are shown on Fig. 11 and are similar in all respects
torque losses of the units. By increasing the motor speed to those shown on Fig. 10, indicating that there were no
and moving up the load torque-speed characteristic, the significant non-linearities present.
1 I
3.76 4.75 7.42 0.0167 0.037 0.957 constant
14 I 4.75
4.75 I 7-42
7.42
0.190
0.190
13.6
13.6
0.035
0.037
0.957
0.957 I) Large load inertia at two
input amplitudes
0.0167 6-1 0.415
0.0167 20.4 0.419 line pressure
1 1 1
0-0167 25.0 0.422 0.957
18 1.25 4.75 7.42 0.0167 13.6 05% Testswith
19 4-75 7.42 0.540 restricted
20 0.0167 boost system
1 I
21 7.42 0,0167 13.6 0.234 0.957
3-76
i!i
7.42 0.0167
iz
0.037
i:g
0.957
r z b hose
flexible y
Test
No.
Simple
theory of
Natural frequency--om, Hz
Experimental Simple
theory of
equation (2)
1 signal
flow analysis
1
Damping ratio-[
From Vector
analysis
1 Experimental
therefore ---o;.)= 1 + - p .
tL
WZfL
JL
fL'
. . . (7)
If this is plotted on a vector locus, it gives the vertical
line shown on Fig. 12. The actual values obtained during
tests 5-8, which departed significantly from this line, are
also shown on Fig. 12. This shows that the load was not by
any means the inertial and viscous load first assumed.
As can be seen from Table 2, the vector locus technique
is very much more accurate, and would appear to be the
most valuable tool for prediction of dynamic performance.
Fig. 10. Effect of load damping coefficent on transmission It slightly overestimates the natural frequency and slightly
dynamic response, ( a ) loci, ( b ) logarithmic plot underestimates the damping ratio.
HZ
Hz
Fig. 13. The effect of an increased inertia load on overall transmission
response
of equation (2) and 1-75Hz using the signal flow analysis. The return line pressure variations PJwZ may be added
The experimentally measured value was 1.70 Hz. vectorially to the differential line pressure variations
d L
.z 1000-
z 800-
3
600-
W
% 400-
Pressure
200-
0
5: 0- 05 10 15 0 0.2 0.4 06
TIME s TIME s
TEST No. 21
80
.ClOOO-
s
$ 800- 60
", 600- 40
W
% 400-
a 20
200-
0
= 0- ' 0 0.5 10 1.5
TIME s TIME s TIME s
Fig. 14. Pressure, speed, and swash traces for system w i t h restricted boost upply, tests 18 and 21
40
w
Ol
80 TI
W
U
-1
w
120
I
k
160
I I I I I
0.4 0 6 0% 10 2.0 4.0 60 8.0 $0 20
Hz
Fig. 16. The effect o f using a flexible hose supply line on overall
transmission response
The return line in most conventional systems contri- differentials across the pump and motor units, but there is
butes to the stiffness of the transmission. no reason to assume that the method should not apply
For the transmission considered the hydraulic pipeline when the pressure differential is much higher, nor when
was relatively short, 1.80 m (6 ft), which is approximately the oil temperatures are higher.
1
-5-, of the wavelength of pressure waves at 15 Hz; under
these conditions a lumped parameter approach is per-
fectly satisfactory. For longer pipelines, it may be neces- APPENDIX 1
sary to take into account the wave nature of the pressure
D I G I T A L C O M P U T E R P R O G R A M M E FOR
signals.
EVALUATING OVERALL TRANSFER F U N C T I O N
Oil inertia effects were found to add less than 1 per cent
OF T R A N S M I S S I O N F R O M S I G N A L F L O W
to the effective inertia of the motor.
ANALYSIS
Use of the isentropic tangent bulk modulus of the oil in
conjunction with carefully determined system parameters Subroutine 'solve'-solves transfer function for T ~ T, ~ 5,,
enables estimates of natural frequency to be made within w , and K.
4 per cent. Subroutine 'resp'-evaluates frequency response, ampli-
The tests were carried out at relatively low pressure tude ratio and phase angle.
P R 0 (i R A I*!T R I) R E S
C TR A N S!1 IS S 1014 n Y f l A M IC .P E S PON S E
C O l l f ! f I N A , B , C , E , G I H, AMP
77 FORMAT(I3)
RCAD(Sr77) NUllRER
I T Ef?=Q
DO 99 K n l l F I T a I I NUMOER
READ(Sr20O)KODE
200 F O R F l A T ( I 1 )
R E A D ( 5 , 3 ) A I U T l , D P , SP, P D T C P , V T C P , 8 , D M , S M , P D T C M , V T C M ,
1 AINTZ, ESLD, THETAP, V O L , THETAM, DFL, BOMGP, SPRESS
3 FORVAT (8EIOm3)
SP=-SP
s Fl=- s 11
ESLDn-ESLD
WRITE ( 6 1 5 ) ESLD, THETAP, VOLtTIIETAM, DFL, BOMGPtSPRESS
5 F O R M A T ( 1 H I I 1O X , 4 H E S L D , 1 3 X e 6 H T H E T A P t 1 4 X , 3 H V O L , 1 3 X , 6 H T H E T A M , 1 4 X t
9 S H D F L , 1 4 X ~ 5 H B O M G P , 1 2 % ~ 6 H S P R E S S / / lHO,lP7El8mS////)
AMP = DP*BOLIGP/ ( D M * T H E T A M )
A =-AINTl*ESLD
H=7=VTCP*ESLD+DP*SPRESS*ESLD*THETAP/ROMGP
D E N (1 I4 1 = D f.1* T H E T A bl* ( D M * 1 H E f A M - P b T C M )
G = l-VfCP*ESLD-(DFL+VTCM)*((SP*SM)*(l~VTCP*ESLD~+ESL~*~P*THETA
1 (DP*TIIETAP+PDTCP) )/flFNOf!1
E =-~ZNT~*ESLD-(~FL+VTC~)*(DVOL*(~-VTCP*ESLD))/(~EN~~~~*~)=AINT
1 SP~Sl4)*(l~VTCP*ESl.D)+ESLD*DP*THEfAP*(I)P*THETAP+PDTCP))/DENOI
1 + ( D F L + V T C 11 1 * ( S P + S M 1 * A 1 N T 1 E S L 0 / D L N 0 M 1
C = - ( 0 F L + V T C M 1 * V O L * A Ir4T1 * E S L D / ( B * D E N O M l ) + A 1 N t 2 * V O L * ( 1 - V T C P * E S L D
1 ( D E N o t I I * R 1 + A I N T 1 * A IN T Z * E S L D * ( S P + S M ) / D E N O M I
R X - A I r J T l * A I N T Z * E S L D * V O L / ( D E N O M I *R)'
W R I T E ( 6 ~ 1 )5A M P , 4 , G I E C t 8 , H
, , , ,
1 5 F 0 R M A T ( 1 H 0 t 1 1 X 3 11 A M P 1 6 X I 1 H A 1 7 X IH G ,A7 X 1H E I7 X 1 H C 1 1 7 X 1H B e l 7 ;
.
1 1t l t l / / l H ~ , l P 7 E 1 8 , 5 / / / / )
IF(KOD€-3)I0,20,10
70 C A L L PESP(KOD€,R99,&20)
20 CALL S O L V E ( 1 T E M )
99 C O N T I N I J E
W R I T E ( 6,2 2 2 1 N UM 0 E R I T E M
222 F O R M A T ( ' 1 * * * ' , I b , ' S E T S OF D A T A A R E P R O C E S S E D t ' t 1 3 , ' A R E T E R M
=TED AFTER 25 l l ' E R A T I O N S * * * I )
STOP
END
Proc lnstn Mech Engrs 1972 Vol 186 55/72
SIIDROIITINE SOLVE( I T E M )
C r ) f l M O N A , P , C p E ,G, H , Ar4P
x = o
ITN = 0
I P I TE (6 I 5O
4 I
Sr) FORt,AAT(11i% / / , 4 X , 4 H X N F W / )
35 W = O * X + * 3 + C * X * * Z + E + X + G
Y = 3*R*X++2+2+C+XtE
XNEW = X - W / Y
L J R I T E ( 0 , 4 5 ) XNEW
4 5 F O R V A T ( ' ' t 1P 1 E l 1 . 4 )
I F ( ARS((X-XNEW)/XNEW)-l.OE-03) 48, 3 9 , 39
39 IT14 f ITN+1
I F (ITN-25) 41, 47, 47
41 x = XNEW
GO TO 35
47 W R I T E ( h r Z 2 2 )
222 F O R M A T ( / / ' +** T E R M I N A T E D AFTER 25 I T E R A T I O N S +**')
I T EY 3I T E M+1
R E T 1) R N
48 XtlEW = - X N E W
R 1 = C-R*XrJEV
S 2 = E-(C-R+XNEW)*XNEW
k'CJ S Q = R 2 / R
G A I N = AMP*H/(B*WNSQ*XNEW)
T l = A/H
T2 =l/XNEW
DAI4P = R l / l J
N W = SQRT(WNS(J)
2E TA= D A M P / ( 2+ W N )
WRITE ( 6 , 58) G A I N , T I , T Z , Z E T A , WN
58 FORMAT ( 1 H O t / / I I Z X 4 H G A I N I 17X 2HTI , 18X 2HT2 , 18X ,
1 4 H 7 E T 9 t 1 7 X t 2HWN I / / 1PSE20.4 / / )
RETURN
END
SUBROIJTINE RESP(KODE,*,+)
C O f l M 0 f . l A , A , C, E , G I H , A M P
WRITE(6,l)
1 FORMAT(1~~@,~3X,9~FREQUENCY~3lX,~~~AMPLITUD~,Z3X,5HPHASE//9X,lOH
lES/SEC,11X,~1iRADS/SEC,I2X,6HABSVAL~I7X,ZHDB,lSX,7HDEGREES // 1
R E A D ( 5 , 2 4 1 O M G F S T r OCIGLST, O M G I N C
2 4 FORMAT (3E10.31)
OClEGA OFIFFST
26 nI,IGSQ = OMEGfi*OfMEGA
DENT)r.l(?= ~ C ~ C * O M G S Q ~ + + Z + ~ E + O M E G A I B + O M E G A * O M E G A * O M G S Q ~ * ~ Z
TREAL = AMP+(W+(G~C*OMGSQ)+A*~E*~MGSQ*+2~)~/DENOM2
TIblAG = A ~ l P + ( A * O ~ E G A * ( G ~ C * O M G S ~ ~ ~ H * O M E G A * ( E ~ B ~ O M G S Q ~ ) / ~ E ~ O ~ Z
AOSVAL = SORT(TREAL**Z+TIMAGE**?)
DR = ?O*ALOGlO(ABSVAL)
PHI\SE = A T A N 2 ( T I M A G , T R E A L ) * S 7 . 2 9 5 7 8
CYCLES = OWtGA/(3.142*2)
3 5 F OR14 A T ( 5 E 2 0 5 ) .
N R I T E (6,35) C Y C L E S , O M E G A , A B S V A L , D B , P H A S E
O M E G A =) O M E G A + O M G I N C
I F (OMEGA-OMGLST) 26,26,38
38 RETURN 'KODE
E N0
m I s the orime
Steady state conditions of test 7:
Load torque, T,
Load speed, Q,
Pump swash, X ,
= 33-4Nm (24.6lbf ft)
= 61.0 rad/s
= 4-42"
I
Motor swash, X , = 7.42"
Determine low Assume constant Supply line pressure, P,I = 87-0bar (1280lbf/in2)
frequency output speed prime mover
speed attenuation Return line pressure, P, = 13.6 bar (200lbf/in2)
using signal flow n,=constant
Additional information required to evaluate system dynamic
performance
be represented
by viscous and Dead volume in pump (ex-
i n e r t i a l terms
posed to supply line) = 86.2 cm3 (5.27in3)
Dead volume in motor (ex-
Consider load to be Determine load posed to supply line) = 86.2 cm3 (5.27in3)
a f i r s t order lead locus
7,
Supply pipe length = 1.76 m (5.75 ft)
TL - f +J s (P) Supply pipe internal dia-
q - L L
meter = 2.54 cm (1.00in)
Supply pipe volume = 924.0cm3 (56.5in3)
Total supply line volume,
V = 1096.4 cm3 (67.04in3)
If the load were removed completely, this would only (g) Compressed flow loss
2
w2 Be w2
I n order to have carried out the signal flow analysis, it
14clw2l (l/s)/(fev/min)x 10’ 0.835 6.19
would be necessary to determine the driving inertia of the
prime mover plus pump, .I1. /gc/w2 Degrees + 141.8 + 164
(2) Can the load be represented by a viscous and inertial
load ? (h) Total flow loss
_. -~ -.
Type of load: positive displacement pump and associated -
- -+-
4c 4 s
loading valves. w2 w2 w2
t2 tm
-*
22
- HAYWARD, A. T. J. ‘Compressibility measurements on hy-
- = -+-+fm draulic fluids’, N.E.L. Report No. 173 (Pt I), 1964 (Dec.).
w2 w2 w2 HARSEN,J. T. ‘How to use readily available specification
I tz/w21 Nm/(rev/min) 0.0350 0.0869 data to predict dynamic properties of hydraulic motors’,
/ t 2 / w 2 Degrees +51.8 + 74
Machine design, Jan. 1961 33 (No. 2), 132-136.
LAMBERT, T. H. and DAVIES, R. M. ‘Investigation of the re-
sponse of hydraulic servomechanisms with inertial load’, 3.
(e) Supply line pressure variation mech. Engng Sci. 1963 5 (No. 3), 281.
RENNER, H. ‘Fluid motors’, Design Engineering Conf. 1965,
A.S.M.E. Paper 65-MD-28, 1966 (March).
HEALEY, A. J. and STRINGER, J. D. ‘Dynamic characteristics
1PhiW2 I bar/(rev/min) 0.099 0.226 of an oil hydraulic constant speed drive’, Proc. Insrn mech.
lPh/WZ Degrees +514 + 74 Engrs 1968-69 183 (Pt l), 683.
KEATING,T. and MARTIN,A. R. ‘Response testing of a
(f) Slip flow loss hydraulic servomechanism with mechanical feedback’, 2nd
Fluid Power Symp., Jan. 1971, A2-25 to 44 (B.H.R.A.,
Cranfield).
WALTERS, R. ‘Hydraulic servo systems analysis and syn-
thesis’, Sperry Industrial Group, pub. I G 100/A,1964.
MASON,S. J. ‘Feedback theory. Further properties of signal
flow graphs’, Proc. Inst. Radio Engrs 1956 44, 920-926.
(15) BOWNS,D. E. ‘Signal flow diagrams as an aid to the analysis tem’, First Fluid Power Symposium, 1969 (Jan.) (B.H.R.A.,
of hydraulic systems’, University of Bath, School of Engin- Cranfield).
eering Report No. 126, 1968 (June). (IS) BOWNS,D. E. ‘The dynamic transfer characteristics of
(16)KORN, J. Hydrostatic transmission systems 1969 (Intertext reciprocating engines’, Proc. Znstn mech. Engrs 1970-71
Books, London). 185, 185-201.
(17)MCCALLION, H.,DUDLEY, B. R. and KNIGHT, G. C. ‘Analy- (19) ANON. ‘Technical data on Shell Tellus oils’, Shell Inter-
sis of a dynamically loaded hydrostatic transmission sys- national Petroleum Company Limited, 1963.
Discussion
G. A. Broadhurst Wolverhampton
I note that the boost circuit is incorrectly designed for a
closed circuit, reversible flow, transmission circuit such as
this.
I n such a circuit, there is no return line as such, both
9 1
Fig. 19. Significance of the periodic variability parameter by determination of the dynamic
characteristics of hydrostatic systems
(1) reduction in performance, and consequently result in widely differing lags at the
(2) energization in adjoining systems with the driving frequency w,,.
chain indicated, and thereby, Whereas such behaviour is not unusual in this kind of
(3) inherent noise whilst under hydrostatic control. system, this suggests that there are too many unknowns for
This appears remarkable when most operating conditions making any accurate deductions concerning values of bulk
are static. modulus-in this connection the assumption that the stiff-
Considering the periodical parametric variability in ness contribution of the fluid is solely provided by one
connection with the determination of dynamical behaviour hydraulic line appears to be invalidated by the results
we employ the linearized equation of motion-a differen- shown in Fig. 14, where peak values of return line pressure
tial equation with variable coefficients of the Hills type, reach about 600 lb/in2, which is a high proportion of the
which are solvable only in special cases. For the dynamic transmission line pressure. This leads one to the conclu-
sion that the return line must play a significant part in the
model, in order to describe the parametric vibrations in
hydrostatic equipment we borrow control engineering stiffness of the system, meaning that the effective value of
methods in the calculation of a scanning system based on B, is no longer 240 000 lb/in2, but is somewhere between
formal impulses and so employ 2-transforms to obtain a 120 000 and 240 000 lb/in2.
good approximate solution. This relates to the diagram, The N.E.L.report (7) is really concerned with condi-
tions of pressure in excess of 5000 lb/in2 and where care
and the necessary modification in the drive of the dynamic
has been taken to ensure the total absence of free air. Since
model. On this model the magnitude of the noise level can
it is known that transmissions such as the one used are
be calculated, the likelihood of achievement of designed
subjected to cavitation locally at the ‘kidney’ ports and at
displacement seen, and the steps necessary for improve-
other discontinuities, it is very likely that some of the air
ment made.
dissolved comes out of solution: since it takes a finite time
BIBLIOGRAPHY
for it to be re-dissolved into the liquid, I suggest that one
MARTIN,C. ‘Der Einfluss drehwinkelabhhgiger Motorpara- is concerned not only with lower pressures than those of
meter auf das stationare Betriebs-Verhaltenvon Hydromotoren’,
Muschinenbautechnik 1972 21 (Heft 5), 224-232. (7), but also with free air being present, and consequently
MARTIN,C. ‘The effect of motor parameter as function of angle it may be more appropriate to consider the system as being
of rotation on steady state performance of hydromotors’, double-acting and to allow for the free air in the manner
Hydruul. Pneum. Pwr 1972 18 (No.212), 356 and (No.213), 400. suggested by Rendel and Allen (zo), where they show that
with a 10 per cent free-air content at 500 lb/in2, the com-
G . Orloff Fellow bined bulk modulus has a value of 50 per cent of the
Referring to section 6.3, the accuracy requirements of nominal value of B,.
performance prediction can vary in industry, depending on This certainly seems to be valid for normal modes of
the purpose of the calculations. For tendering, the simple operation of positioning systems, and it would also seem
equation (2)leads to rapid results of sufficient accuracy for desirable to use similar assumptions in the case of velocity
wn and errs in the direction of safety with respect to 5, and transmissions like the one under discussion, particularly if
this is usually corrected for from experience-conse- they are intended to behave normally when occasionally
quently the formula serves a very useful purpose. filled with a fluid containing a high proportion of air, as
I n the case of the more detailed design work which may sometimes happens in industry.
follow, perturbation theory gives adequately good accuracy
REFERENCE
in most conditions, and even in the absence of computers,
an experienced performance analyst will choose perturbed (20) RENDEL, D and ALLEN,G. R. ‘Air in hydraulic transmis-
sions’, Aircr. Engng 1951 23, 337-338.
coefficients sufficiently well to give confidence in the hard-
ware manufactured. I n the presence of gross load non-
linearities it can be more desirable to avail oneself of a A. H. Richards Member
digital computer for dealing with a multiplicity of coeffi- With regard to the evaluation of bulk modulus, which
cients derived from an empirical load curve, replacing the remains the only unknown in the theoretical treatment, I
more simple speed-dependent relationship in Fig. 2-it is should like to ask the authors whether they have con-
not clear to me, however, why changing methods to ‘vector sidered the effect of rate of change of pressure, d(dP,)/dt.
locus’ should result in greater accuracy. As a frequency response test proceeds from low to high
Signal-flow techniques are understood to be suited frequencies, the acceleration of the load will be directly
particularly for handling the perturbed equations of multi- proportional to the product of signal frequency and half
variable systems and others of high complexity-what is amplitude. The increasing acceleration will entail greater
the justification for doing so in the present case which is pressure oscillation amplitude, and hence greater rate of
relatively simple ? change of pressure.
It would have been interesting to know the contents of It is usually assumed that the isothermal bulk modulus
the significant parameters in equation (5) and to compare is associated with slow changes of pressure and the isen-
them with equations (3) and (4). tropic value with rapid changes. I n consequence, I am
It would have been interesting also to see a set of Bode interested to know if a trend in the bulk modulus could be
curves for tests 5-8 permitting a comparison with Fig. 11, detected over the test, and was any attempt made to fit
particularly since values of fL taken as tangents to the rele- equation ( 5 ) to the measured results over the whole range
vant curve in Fig. 9 lead one to suspect the presence of of one test ?
large-amplitude cyclic non-linearities. Alternatively, for a particular system configuration, did
Some non-linear behaviour is already apparent in the bulk modulus as derived from the vector locus method
Fig. 11, where the phase curves intersect at about -50” show any trend as the input amplitude was varied?
Proc lnstn Mech Engrs 1972 Vol 186 55/72
prevent negative replenishment flow, as shown in Fig. 6 4 for this. For several of the tests reported in the paper, an
and, as a consequence, cavitation could exist for larger artificially large boost system was used to keep return-line
portions of the pressure oscillation. The high-pressure pressure variations to a minimum, but for tests 18-21 a
peaks would normally only be limited by the pressure- restricted boost system was used and the effective value of
limiting valves connected across the motor ports. bulk modulus which allowed prediction of the experi-
It follows, therefore, that the hydraulic stiffness of the mental dynamic response using the vector locus approach
return line would be more dependent on the oil charac- was still 240 000 lbf/in2 within 4 per cent. Indeed, one of
teristics than the impedance of the relief valve, and could the most important points of the paper is to show that there
result in worthwhile increase in natural frequency. Could are not too many unknowns in a hydrostatic transmission
the authors comment on this aspect ? system. An accurate assessment of natural frequency and
Finally, could the authors give their views on employing damping ratio can be made without too much difficulty.
much higher boost pressures with a view to stiffening the I t might be opportune to point out that the transmission
system and preventing cavitation ? tested contained a relatively small oil volume and had a
load of low inertia. Predictions would be more accurate
D. E. Bowns Member and J. Worton-Griffiths still with larger volumes and high inertias.
(Graduate) (Authors) The effect of air on transmission performance has in-
G. Orloff discusses the use of equation (2) and we agree terested the authors and G. Orloff is referred to (25). Air in
with him that, as a rapid check on natural frequency, it is the fluid is not the bogey it has hitherto been thought and
extremely valuable. However, it gives consistently high re- work on the transmission dynamics with air present has
sults if the correct value of bulk modulus is used and substantiated the methods of analysis used in the present
therefore is of dubious value for tendering purposes, par- paper.
ticularly as customers become more aware of the necessity I n reply to A. H. Richards, the effect of the rate of
for dynamic specification of performance. If the load is change of pressure has been taken into account in the
frequency dependent, as was the load in the experimental overall transmission analysis but not in considering whether
rig, no computerized method is possible which would take to use isentropic or isothermal bulk moduli.
this into account, except a completely step-by-step ap- The results obtained by analysing the experimentally
proach such as that adopted by Knight et al. (24), which determined frequency response using the vector-analysis
is considerably more complex than the methods discussed technique did not indicate any trends in the bulk modulus
here. The vector-locus technique is familiar in concept to values as the frequency or input amplitude were varied.
most control engineers and does give a rapid and accurate Values obtained for a typical test are given in Table 3.
check on the dynamic response of this type of mechanical It should be pointed out that at low frequencies where
system. It can cope with many non-linearities and does not the flow losses were small and where slip losses made up
necessitate the use of a computer. Examples of its use are the significant part of the total loss, this method for deter-
discussed below. mining the effective bulk modulus of the oil was not
The signal-flow technique assists greatly in the analysis satisfactory. This made an evaluation of bulk modulus at
of systems such as the one under discussion; the block low rates of change of pressure impossible.
diagram is complex and extremely difficult to disentangle C. R. Burrows's comments on the use of the measured
and a straightforward mathematical approach involves the load loci are very pertinent and so are his algebraic correc-
solution in the simplest case of eight simultaneous differen- tions. They have been incorporated in the corrigenda.
tial equations. The problem is, in fact, multivariable and A. B. Goodwin's points about Fig. 1 and Fig. 2 are both
one advantage of the signal-flow approach is that with very correct, and we are in agreement with both him and C. R.
little additional work the relation between any other input
and output (for example, the effect of the change in load Table 3. Effective bulk modulusx l o 5 Ibf/in2
torque on motor speed) can be obtained very easily. How-
ever, in spite of all the care taken, poor correlation was Hz Tat No.
obtained between the theoretical Bode plot developed
from equation ( 5 ) and the experimentally determined 1.01 1-05 2.03
values. It was for this reason that the vector-analysis 2.0 2.70 2.33 2.45
technique was adopted. 3-0 2.28 2.23 2.29
G. Orloff's request for Bode curves for tests 5-8 is easily 4.0 2-18 2-11 2.18
4.5 2-24 2.10 2.17
fulfilled, as they are all included on Fig. 10, whilst his 5.0 2.30 2.07 2.15
worry about the intersection of the phase lag curves at 5-5 2.34 2-11 2-18
6.0 2.18 2.14 2.21
about -50" is not relevant as the transfer function of the 6.5 2.20 2.14 2.22
system is much more complex than the second order form 7-0 2.18 2.14 2.21
on which he appears to be basing his arguments. Curve 9 80 2.24 2.18 2.25
9.0 2.20 2-18 2.27
was in fact carried out with a load of different torque speed 10.0 2.26 2.19 2.27
characteristics and when the load static and dynamic 12.0 2.29 2-22 2-30
characteristics were taken into account the transmission Mtan 2.24 2.15 2.23
coefficientsobtained were in close agreement with those of
the other tests. Standard
G. Orloff goes on to state that the stiffness of the return deviation 2.5 2.3 2.2
pcr cent
line makes some contribution to the system stiffness-in-
deed it does, and one of the most important advantages (Three tests at different mean return line pressures and load
of the vector locus approach is that it can be used to account damping coefficients.)
Burrows that the line-differential pressure p D rather than I. Watts is incorrect in assuming that cavitation would
the high-pressure line variation p H should be used in the occur for longer periods with non-return valves installed
torque equations. Indeed when variations in return-line in the system. The cavitation described in Fig. 6d occurs
pressure were significant, the former value was used in the when the relevant non-return valve is fully open and the
analysis presented in the paper. boost system has too little capacity to cope with the large
With the relatively short lines used it was considered demands made on it. He is, however, correct in stating
justifiable to neglect pipe losses, but for systems where that much higher boost pressures can stiffen up the system
they are not negligible it might be better to incorporate and prevent cavitation. Unfortunately higher boost pres-
them as a deduction from the motor torque coefficient. sures would reduce the maximum pressure differential
A. B. Goodwin’s point about minimum damping has available for any given design of pump or motor unit and
academic interest. However, low damping is to be avoided also lead to increased power losses in the boost system. In
in hydrostatic transmission systems and unfortunately the a design where such considerations are unimportant, high
designer usually has no means of controlling it. boost pressures would certainly be advantageous.
In reply to I. Watts’s point about load frequency re- We thank J. Thoma for his contribution on Bondgraphs
sponse, we do not underestimate the difficulties in obtain- and consider that this method should be more widely
ing this. Indeed, determination of the load characteristics known.
is a major part of the difficulty in all hydraulic design. C. Martin’s contribution is also interesting. He suggests
However, it seems to us important to pinpoint the difficul- the use of sampled data techniques for determining the
ties, rather than attribute subsequent inaccuracy to such cyclic flow and pressure fluctuations and corresponding
effects as that of air in the transmission fluid or other un- resonances. Such methods have already been used by one
knowns. of the present authors for analysis of the dynamic be-
We agree with G. A. Broadhurst that a closed-circuit haviour of reciprocating engines (18) but during the test
reversible flow transmission system should be designed as work carried out on the transmission there was no experi-
in Fig. 18 and indeed the test system was initially arranged mental evidence of behaviour which could be attributed to
in this manner. However, the system was being used as a sampling phenomena.
speed-control circuit and the pump swash-plate angles
were never reversed. Hence the large transient pressures REFERENCES
mentioned by G. A. Broadhurst were not experienced in (24) KNIGHT,
G. C., MCCALLION,
H. and DUDLEY,
B. R. ‘Con-
the experimental system. The reason for removing the nection capacitance effects in hydrostatic transmission
check valves was to remove some of the system unknowns systems and their prediction by mathematical model’, Proc.
Instn mech. Engrs 1972 186,661-670.
and enable reasonable control of the experiments. There
(25) BOWNS,D. E., EDGE, K. and WORTON-GRIFFITHS, J. ‘The
are, however, still transient pressures as can be seen from effect of air on the operating characteristics of a hydrostatic
Fig. 16. These would not be reduced by the inclusion of transmission’, Proc. Third Int. Fluid Pwr Symp., Turin,
non-return valves as the line pressure differential is not 1973 (to be published).
reversed.
Corrigenda