You are on page 1of 8

WEAR

ELSEVIER wear i% (1996)64-71

Comparison of an external gear pump wear model with test data


R . H . Frith, W . S c o t t
QueenslandUniversityof Technology.Schoolof MechanicalandManufacturingEngineering,2 GeorgeStreet, OPO Box2434, Brisbane,QLD4001,
Australia
Received3 August1994;accepted26 October1995

Abstract

Power hydraulic pump wear is difficult to quantify. Measuring the actual material lost due to wear is impossible in a practical sense. An
available simulatedprocedureis the contaminatedsensitivityperformancetest which measuresflowdegradation with increasing amounts of
contamination in the fluid. It is shown that wear can be related to flow degradation and knowing the mass of contamination causing wear,
models for pump wear can be explored.This paper consolidatesa simple modelfurther and shows that a good correlation exists betweenthe
model and availabIetest data.

Keywords: Hydraulics;Pumps;Wear;Model

1. Introduction wears, leakage increases to a point where the required flow


can no longer be met. The pump has then degraded.
Practical measurement of flog' degradation in hydraolic It has been demonstrated [3] that pump wear occurs in
pumps due to solid contamination has been undertaken for regions of close contact. Contaminant borne with the leakage
many years [ 1]. Despite an unresolved dispute over how to fluid abrades the close-coupled surfaces as it passes through,
interpret test results, the basic test has gone unchallenged and resulting in the generation of wear debris. With the loss of
is now nearing international standardisation [2]. Very little material both the gap between the surfaces and the leakage
fundamental understanding of the pump wear process has flow through the gap ate increased. Clearly, the increase in
been advanced, with the result that no model exists to describe pump leakage or alternatively the pump flow degradation is
it. Reported here is a method to relate flow degradation test related to the mass of wear debris generated. Moreover, the
results to the generation of wear debris. This will form the wear debris mass is related to the mass of damaging contam-
basis of an attempt to compare pump flow degradation results inant passing through the gap. It is argued that not all contam-
with a model for the wear process in pumps. inant causes wear, only that greater than a certain critical size,
x~t [3], which in turn will be related to the gap size.
If a relationship can be established between pump flow
2. Pump wear mechanism degradation and weardebris mass, and the damaging contam-
inant mass causing the wear can be determined, results from
pump flow degradation may become useful in exploring
With the pursuit for high efficiency, hydraulic pumps must
operate with minimal clearances. Any increase in clearance pump wear models.
is seen as a loss in flow due to an increase in leakage. It is
this reduction in flow or' 'flow degradation" that is measured
during a pump test. Although knowledge of how the pump 3. Pump test data
flow degrades due to solid contamination is important in
selecting wear-tolerant pumps, it provides no direct measure- Pump flow degradation testing has been in use since the
merit of wear. mid 1970s and is now recognised world wide, although some
Pumps are typically configured in a system to operate at variants are being implemented. There exists therefore a large
constant pressure with the excess flow passing over a relief database of results. Under an ongoing program of research,
valve or, in the case of pumps having a variable displacement, the Fluid Power Research Centre [4] published results for
flow is provided only on demand. In either case, as the pump 172 hydraulic pumps tested over aseven-yearperiod of which

0043-1648/96/$15.00© 1996Elsev~ ScienceS.A.All righ~~,rved


55DI0043-1648( 95)06845-7
K H. Frith, t?/.Scott/Wear 196 (I996) 64-71 65

Table 1
Pumpdegradationresultstakenfrom [4]. All pumpsweretestedwith 15.2cSt fluidwiththe excel~ionof 137widthwas testedwith22.2 cSt aad 132,133
and 171whichweretestedwith 18,8cSt

ID Dirt Flow Pressuvv Parliclesize "cut" (ttm)


(mgt-t ) (Irain-') (bar)
<5 < 10 <20 <30 <40 <50 <60 <70 <80

27 3OO 26.t 162 I 0.739 0.154


40 302 71.5 151.7 I 1 0.952 0.868 0.799 0.751 0.683 0.614
56 300 90 137.9 0.996 0.992 0.962 0.912 0.866 0.803 0.745 0.7
69 299 81.4 137.9 0.995 0.94 0.879 0.805 0.674
70 99 80.6 137.9 1 1 0,995 0.967 0.944 0.906 0.793 0.742
72 197.1 82.5 !37.9 1 0.991 0.945 0.899 0.821 0.693 0.628
73 46.1 86.7 137.9 1 0.996 0.991 0.961 0.926 0.882 0.852 0.790
84 300 179 206.9 ! 0.979 0.889 0.628
105 300 102.2 158.6 I 1 0.988 0.959 0.930 0.896 0.859 0.830
106 299 105.7 158.6 I 0.992 0.952 0.916 0.8"/5 0.8II 0.751 0.679
107 75 70 137.9 i I I I 0.995 0.989 0.973 0.962
122 302 86.1 137.9 0.995 0.981 0.92t 0.823 0.774 0.685
132 298 53.9 172.4 0.984 0.968 0.881 0,756 0.551
133 300 47.5 172.,* 0.9~2 0.955 0.902 0.715 0.159
t37 300 73.6 137.9 ! 0.990 0.949 0.859 0.720 0.635
168 300 92.3 206,9 I I I 0.987 0.908 0.881 0.768 0.615 0.486
171 299 89,6 172.4 I I 0.998 0.992 0.963 0.931 0.905 0.864 0.~27

121 were gear pumps. A selection of results is shown in mlmllNlm.


pu,mumaama~,.m{j,H
Table I and Fig. 1. Results were obtained by successively s o M i
increasing the contaminant distribution size range by sieving 1
at different levels (or "cuts"), but usually maintaining the "'"'.,gin
%%,
' "'" O.
" o ,
contaminant weight constant at 300 mg 1-1. However, in ,] t t
.................................................
o .....=.: ..o
some case.~ the contaminant weight was reduced to evaluate t as
its influence on pump degradation. After each test period of
30 min, the oil was filtered and the next size range added. ~ ~7
The test continued until the pump flow had degraded sub- OJI
. . . . . . . . . .
........................................
. . . . . . . . . . . . . . . . . . . . . :::":.l ° "1 . . . . . . . .
:,~;,;~.. c. :..:. ..................
stantially or the maximum contaminant size range was
reached, as " 7. :
m ~ 1N
Referring to Table 1 or Fig. 1, certain pumps have a greater 0
'ramo(namamdm}
tolerance, showing no evidence of wear until a larger contam-
inant size range is added. In the context of the model proposed
here, the range at which wear commences would be equiva- i
lent to x~t,
Knowing the unworn volumetric pump efficiency,
7/v= Qo/Qv, the leakage factor A, can be written in terms of o.7
flow degradation as follows, tl e.e

("t
A-- -- 0-~-
~v'Q Qo
I (1) u
°o~
0,3
where Q is the degraded pump flow, Qo is the unworn pump
0.1
flow and Q~ is the theoretical pump flow.
o
Using Eq. (t), values of A for several pumps are shown U ~ m U
in Fig. 2,
Fig.2.Leakagefncto¢,2t,derivedfromprop delpada[m resulls[4l.

4. Proposed model assumes that particles above a certain size passing through
the leakage gaps are either broken up or blunted so as to
Based on the reasonable assertion that wear takes place reduce further wear by that pmicle. An initial particle mass
only with the passage of particles through leakage paths in distribution, ~ x ) , is modified with the passage of time, by
the pump, a pump wear model was developed [ 3 ]. The model the combined effects of panicle size reduction or blunting
66 R,H.Frith,W.Scott/Wear196(1996)64-71

Letting

(1-a)A (3)
Y=(l+(1-a)A)
Eq. (2) can be solved to give
DI = D e--°v~' (4)
where D is the initial particle mass/volume which is assumed
uniform throughout the system.
Similarly, it can be shown that

1 dsi Q "1 +o~)D1


~=~yL (5)
F3g.3. Pumpwearmodel.
where a~ is the wear debris generation factor and sl is the
and the addition of wear debris. The model is illustrated in mass/unit fluid volume of non-damaging particles at the
Fig. 3 for the simplest hydraulic circuit. pump entrance,
Assumptions made were: Knowing the initial particle mass/unit fluid volume, s,
• only particles greater than a critical size x~, do damage
(wear);
• damage is dependent only on the damaging particle mass
(°,)
sl--s+D(l+o~) l - e - ~ (6)
or volume and not its shape or size;
s particles greater than x~t are reduced in size as they cause Inspecting Eq. (6), wear debris mass/unit fluid volume
damage; generated will be given by the difference betw~n ~ and the
• wear mass per unit volume distribution, ~ ( x ) , has no contribution due to the contaminant mass, ( s + D . ( l -
e - t~'/v)) ). Hence,
particles of size x,~, or greater;
• wear mass is directly related to the damaging particle mass. , q
Since the model assumes only two classes of particles, the S wffiDaw(1-e-~ ~') (7)
particle mass distribution, 4,(x) is represented by s (particle
mass per unit fluid volume causing no damage) and D (par- where the cumulative wear mass/unit fluid volume, s',,,=
ticle mass per unit fluid volume that causes damage). t

Taking the simple model (Fig. 3) and making the simpli-


f swdr.
lying assumption that particles break down or are blunted
o
before wear results in an appreciable increase in leakage flow,
the damaging particles, De, at the pump entrance change with 5. Validity of model
time as follows [3],

dDl Q (1-a),~ O Examining Table 2, increasing pump tolerance (increased


7 v(,T(~) ' (2) xcrit) should resuR in reduced wear as a consequence of a
reduction in the damaging portion of contamination. Wear
where Q is the flow, v is the system volume, a is the particle mass, Sw is related to the decrease in pump flow due to an
disintegration or blunting factor ( < 1.0), A is the leakage increase in internal leakage as the exposed surfaces in the
factor and the numeral subscript refers to a location in the leakage gap are eroded. Assuming leakage flow is rehted to
hydraulic system. the cube of the leakage gap height, it can be shown that [3]

Table2
Valuesforthedamagingcontaminantpotion,D in mgI-. ValuesshowninbracketsateD as a percentageofthetotalcontaminantmassof 300mgI- i

Cut (Ixm)
F(zc) 0-5 0-10 0-20 0-3~ 0-40 0-50 0-00
x~, (tim) 11.1 51,28 144.48 219.72 275.5 316.78 479,78

5 0. 235.06 (78) 276.95 (92) 284.84 (95) 287.91 (96) 289.5 (97) 293.06 (98)
t0 0. 0. 193.53 (65) 230.0 (77) 244.16 (81) 251.44 (84) 267.94 (89)
15 0. 0, 94,30 (31) 164.74 (55) 192.12 (64) 206,18 (67) 238.05 (79)
20 0, 0, 0, 102,73 (34) 142.67 (48) 163,17 (54) 209.66 (70)
25 0, 0. 0, 47.69 (16) 98,77 (33) 125.0 (42) 184.45 (62)
30 0. 0. 0. 0. 60.74 {20) 91.91 (31) 162.61 (54)
R.H. Frith, W. &ott l Wear196 (1996) 64--71 67

BFPP.-#62 values for x=, and 11,,were used, and noting that c as in
] ~ " ~ - ~ ....... : ..... ..:.-..,..++ ....... .......... .......... .......... i
O,~J . . . . . . . .". . . . . . . ~.. . . . . . . ~- . . . . . . ,'¢-. .~. . -. . .-. . ~. ' - : " ....... : ....... Eq. (9) and F_x1. (9a) isthe slope of a line relating ~ ( D ) j
.Ul+ . . . . . . . +. . . . . . . +. . . . . . . -]. . . . . . . J . - - ~ - - P - . . . . . . +. . . . . . . +
j=l
|o.,,t
+,"++ .......
....... !+.......
....... ~ ....... i ....... - ....... ! ...... i: ....... ~i.......
+....... +........
+ ++::::::::::::::::..+
....... +...... +....... .+
to 0, the coefficient of determination,
~,'~ :::::::::::::::::::::::::::::::: ....... ,~....... ~........... :::~
u.s3 ....... '., ....... +.'....... J. ....... ,.{........ ~ ....... '., ....... +...... J.
....... ....... , ....... ....... ::::::::::::::::::::::::::::
i=!
°~;t ....... i ..... :,:i....... ~....... . ....... ~ ..... . - - ~ #=l-
o I0 20 3o 40 5o 6o 70 ell

E e -E*'
Fig.4. Pumpdegradationresultsforpump62.
'°'L . J
where n is the number of data points and *m, the value o f ,
rl-- 0" -1113 as evaluated using the straight line fit parameters, was eval-
~,s,.=l~/ -t (~) uated and the trial values that maximised rz were assumed to
be the most likely values. A value of 1.0 for v2 indicates a
where Sw--V.s'+, V is the system volume, O~=Q/Qoat the perfect fit. Results are shown in Table 3.
end of the ith test period and cl is a constant related to pump An alternative scheme of estimating xmt is by observation
size (displacement). of the degradation curve and Caking it to be when the curve
Recognising that S~ as given in Eq. (8) is the total wear begins to depart from a degradation ratio of 1.0. It was nec-
mass after i test periods and assuming test time is sufficient essary at times to apply subjective judgment to curves typ-
to ensure the exponential term in Eq. (7) is negligible, Eq. ified by Fig. 4, in order to establish reasonable values. For
(7) can be introduced into Eq. (8) to give this scheme, a constant volumetric efficiency of 0.9 was
assumed. A summary of results is listed in Table 3. A sample
c.E W)j :'1-°'% -l (9) of data used in the analysis is shown in Table 4.
)=, L 1-~7~ J

where (D)~ refers to the damaging contaminant injected at


the start of the jth test period and c = V.ct.%. 6. Discussion o f m d ] ~
i
Recognising that ~ (D)j =D and recalling that D= Despite the simplifying assumptions made in arriving at
¢' j=l Eq. (9) and in papular the assumption of a uniform gap
f dp(x)dx [31, (9) can be written as size throughout the pump. there is a reasonable fit as dem-
x~lt
onstrated in Figs. 5-9. Some pumps do deviate markedly, but
only at extreme values of ~ as shown in Rg. 9.
cD=, (9a)
Variation in data about the straight line fit could be clue to
where the wear parameter problems associated with eusm'ing the pump is in the same
rl 0 -v/3 running state at the commencement of each contaminant
injection phase following clean up from the previous phase,
L]-~vJ in addition to the inevitable ener in measw'emenL A major
i.e. a linear relationship exists between D (which is dependent difficulty in analysis is the sensitivity of resolts to variations
on x~,) and ,. in both ~ and xmp Despite the inherent uncertainty in results,
In order to apply Eq. (9), xmt and vk must be known. In an attempt will be made to draw conclusions. It is clear,
the context of the model, it is reasonable to assume that xm, however, that for the model to be fully validated, x ~ needs
will correspond to the injection cut size at which wear appears to better understood and measurable, pm'ticularly in the con-
to commence. Normal experimental error makes it difficult text of variable gaps throughout the pump and during oper-
to select x,.m with certainty. Consider pump 62 by way of ation. Unworn volumetric effieiencies must also be known.
example. Referring to Fig. 4 which shows degradation results
against cut size, is it reasonable to take x,~t as tO Itm. the 6.1. Effectof varyingcontaminateconcentration
point at which wear appears to begin, or is experimental error
such that x~, could be taken as 30 pm or even 40 Inn? There is a concern that high contamination levels result in
While values for x~t can in principle be deduced from the a mode of wear which is not representative of that found in
pump degradation results, no values are given for ~,. Clearly, normal operation [ 1}. Among the data taken from [4], tests
relating Eq. (9) to published data is difficult. 69 to 73 would appear to be resolts from similar pumps tested
An attempt was made to use a scheme based on the "least at varying contamination levels. Observing estimates for c
squares" method to find both unknown panuneters. Here trial using both the least squares or observation methods, as shown
68 RH. Frith,W, Scott/ Wear 196 (1996)64-71

Table3
Analysisof pumpdatagivenin Table I

ID How Pressure Dirt Least squares O~rv~on Comments


(Imin-l) (bar) (mgl -t)
xmt ~v C X~t C

27 26.1 162 300 10 0.85 0,999 0.00187 10 0.0025


40 71.5 151.7 302 15 0.8 0,985 0.00029 20 0.0007
56 90.5 137.9 311 15 0.9 0,997 0.00045 15' 0.00045 °Couldbe
.g~t--20, c--
0,00065
73 86.7 137.9 46.1 25 0.8 0.996 0.002 15 0.002 r2 is highovera
rangeof results
70 80.6 137.9 99 30 0.9 0,986 0.003b 20 0.0014 t'Couldbe
x~, = 30, c =
0.0021 I,
~=0.85
72 82.5 137.5 197 20 0.8 0.991 0.00074 15 0.0009
69 81,4 137.9 299 10 0.85 0.995 0.00045 10 0.00063
84 179 207 300 25 0.8 0,990 0.00247 15 0.001 I
105 102 158.6 301 20 0.8 0.997 0,0002 20 0.0004 Goodfit
106 105 158.6 299 15 0.85 0.998 0,0003 15 0,00044 Goodfit
107 70 137.9 74.9 40 0.9 0.992 0.00157 40 0.00157
122 86.1 137.9 302 15 O.S 0.989 0.00037 I0 0.0005
132 53.9 172.4 298 10 0.8 0,980 0.{]0063 5 0.00063 Fluidviscosity
is higherthan
otben
133 47.5 172.4 300 15 0.85 0.972 0.00171 5 0.0008c ~Asfor 132 and
there is marked
devia~on from
a straightline
137 73.6 137.9 300 10 0,8 0.991 0.00039 5 0.00045
168 92.5 206.9 301 25 0.85 0,993 0.00081 20 0.00081
171 89.6 172.4 299 20 0,85 0,997 0.00027 20 0.0004 As for 132 and
good lit

Table4
The wearpmameter,~', using ~ - - 0,9 and the ~rresponding valuesfor the cumulativedamagingcontaminant.Du = ]~(D) for severalpumps

ID x~t Pm'sn~tcr Cut (p,m)


(ttm)
<5 <10 <20 <30 <40 <50 <60 <70 <80

56 ~ 0 0.0119 0.02M 0.1030 0,2146 0.3018 0.4049 0.5160 0.5941


15 U~ 0 0 94.29 259.03 451.15 657.34 871,87 1091,9 1315.62
62 ~ 0 0,0177 0.0177 0.0177 0.0177 0.0320 0.0804 0.1480 0,2327
40 D,~ 0 0 0 0 0 39.10 t01,44 178.93 266,90
105 ~: 0 0 0 0,0348 0.1104 0.1769 0,2463 0.3140 0.3626
20 D~ 0 0 0 10233 245.39 408.57 583.94 767.25 956.05

122 ~ o 0.0148 0.0540 0,1961 0,3738 0.4477 0.5653


!o o~ o 0 193.53 423~3 667.69 919.12 !!74,9

*where

L(I-~,)J

in Table 3, there is a decrease with increasing contaminate runs contrary to an expectation of wear increasing with
concentration. Of perhaps equal significance is the d e c r e ~ concentration.
inx,~,.Por identicalpumps, a decrease in cimpliesadecrease Assuming the pumps are identical, it is curious that x,~,
in the wear debris generation factor, o~ since c ffi V~o~. This decreases with increasing concemration. A lowerx~t implies
R,R Frith, W. S c o t t ~ W e a r 196 (1996)64-71 69

9FPR-#66 BFI~ -J 133


EIl-O,.q g.oJ
T ...... ~ ...... ~ ....... " ...... T ...... :....... : I ¢ ~ T . . . . . . ~. . . . . . . ~ . . . . . . . f . . . . . . V . . . . . . ".. . . . . . . :

...... 4....... ~....... ; ..... "1""':;:.r". ""'~ '---~--~-,Spm' "t ...... ....... ....... i ...... .... i ....... i t
m ...... ~-...... ;".... ' .... ~ ...... ~....... i]_.,~.,.,s., ~
loee . . . . . . *- . . . . : ....... . : , ,
® ......: .+. -!......~. .~.......i
so~ .... : "" . . . . . . ' ...... ' [ - - ¢. o l 4 S
L..._c.omm
| i
o o,1 o2 ~ &4 O5 0.6 o o.~ 04 u oll , I,~

~igt 5+Da'a for pump 56 plottedas wear p~rameteragainsttotal damaging Fig. 9. Damfor pump 133 pleUed as w e a r p a = m e t e r against fetal d a m a g i n g
co{.:a~r.inant for values ofx~,. contaminan~for valuesof x,,a.

9 F P R - I 105 end plates, adjusts m suit the increased concentration. As


El-&9
particles become worn, they adopt a rounded almost spherical
~ .......... 4. . . . . . . . . . . " .......... .~. . . . . . . . . i .-o-~,,,.,~ form and if sufficient quaatities of these worn panicles exist,
,4oe .......... ~. . . . . . . . . . . ~,. . . . . . . . . . ' ......... i -.,,- ~m-,o~m they could hold the end plates apart from the gear ends,
' --': ........ " ......... i
,~o ...........:.......... . ........ ~......... i .-.-,.-,~-a~ admitting other pa~cles through which would in turn create
. . . . . . . . . . -i. . . . . . . . . i. ....... ' ....... ! ~-~.mpm
m eeo
.o
~ .. I !........
~ "..... ; . . . . . i:!: wear debris at a reduced rate. 'Inis is speculative and needs
to be substantiated. Results for pump 107 would also tend to
add some support to the trend in decreasing c with increasing
o ~1 02 03 0.4 concentration. TSat is, c for 107 is substantially higher than
~ u t ~ that for pumps with the higher concentration (e.g. pump 40
Fig. 6. Data for pump 105 plottedas wearparameteragainsttotal damaging or 137).
contaminantfor valuesof x,~,.

BFPR-# 122 6.2. Effect of pump size


F,~ • Or9

mo ...... J ....... ~ ....... ~ ...... ~- ...... ~ . . . . . i: In discussing the physical significance of c, it was pointed
out that an inverse relatiouship to pump size may exist. Coln-
paring pump 27 with 106, it can be seen :hat the ratio of flow
is 4.02. The ratio of the respective values for c is 6.2 if the
least squares method is considered or 5.7 if the observation
i ,--o.- results are taken. An explanation for the difference may well
9 0,1 0.2, U 0.4 03 U be that displacement is not related directly to flow, since in
W e ~ this instance pump speeds may differ. In addition, it may be
Fig. 7. Data for pump 122 plottedas wear paran~teragainsttotal damaging that the relationship of pump size to c is more complex,
contaminantfor valuesof x~,,

6F'PR-I 137 6.3. Correc~ng rtSu|Ls


Bf.O,B
1,qOOiT..... ":...... : ......f ..... : ......
" ~ " ..........
~ f :.
If the mode! is valid, results can be corrected to a common
~w . . . . . ~. . . . . . i . . . . . J. . . . . . i - - . g ~ - ~ - - - - - ~ - - - - - - i "-"-~="0~ base and compared directly. Consider pumps 40, 56, 105,
122, 137 and 171. Applying the suggested correction for
WOO ..... ~ ..... -~ ..... .-- -: .... , ..... : ~xa~.ntm pump displacement and using 1001min -1 as a base, the
4M .... : , .... i .... :'" ~ ..... [ .... i ---~--..-~pm values for c corrected to the common base arc shown in
2 o o o ~ q .....~ ' L. ' ...... i ~ o . , ~
Table5
D 0.1 0.2 [3 0.4 U 06 B.7
Resultsof some pumpsmnmed to a commonbase of I001 mia" t
W ~ U I ~
Fig. 8. Datafor pump 137 plottedas wearparameteragainsttotal damaging
ID Flow (I ndn-I) c Cornered c
con,aminantfor values ofxmt.
0 71.5 0.00029 0,00021
a greater component of the contamination is damaging. Even 56 9O5 0.00024 0£0022
despite the reduced x~t, c still reduces with increasing con- lOS 102.0 0.00020 0.00020
86.1 0.00037 0.00032
tamination. There are no doubt several plausible uxplanations 122
73.6 O.O0039 0.00029
including the suggestion that the larger sized particles do not ]37 89.5 0.00027 0.00024
enter the gaps and hence cause no wear [6], but it is possible 171 Ikst fit = O.OOO23
that the dynamic behaviour of the gaps and in particular the
70 R.H. Frith,W. Scmt/Weav196(I996)64-71

1400 . . . . . . . . . ; ...... 'T, . . . . . . T ...... 'I,. . . . . . . ;'_" . . . . . ; Alternatively, the particles may simply be blunted, losing
......... i .......
i ~
...... i ....... i .......
[ ! • ' i
o" I their cutting edges. In any case, the concept propounded here
of a damaging and non-damaging component is not lost.
, m ......... :~. . . . . . ÷: . . . . . . ~~I. . . . . .a , [.' : . . ~ : s ,. . . . . . |![ 1 °'~ I
The simplifying assumption made in quantifying the dam-
0 ~ ; o ' :o ; ; aging component from cuts of ACFrD would be improved
by using the actual measured particle size distribution. Such
........ e .... ,"---i" ...... ~,....... ~....... "; i ' ,in I distributions have been measured in the past [7], ~ut their
mathematical representation lacks the relative simplicity of
0.~ 0.1 OlS 02 0.~ 0.3
raw ACFI'D [5]. However, it is doubtful that the effort
Weml~ expended will greatly improve the confidence in making a
Fig. 1O, Data 5"ore pumps 40, 56,105,122,137 and 171 connoted to a base correlation as demonstrated in Fig. 5, Fig. 6, Fig. 7, Fig. 8
flow of)001 rain -~ , and Fig. 9. The simplifying assumptions made in developing
the wear parameter and experimental errors involvedin meas-
Table 5. Corrected values for c are not substantially different
uring pump flow degradation could overshadow improve-
from each other.
ments gained in using a better model for the cut.
The corrected values are plotted in Fig. I0. The scatter of
The model suggests that if contamination is not replen-
results is not substantial and adds support to the proposed
ished, wear activity (see Eq. ( 7 ) ) decreases exponentially in
model as being reasonable. The value for c taken as the best
time. Although this has been consistently reported [1,3],
fit through the collective corrected data is shown in Table 5,
there is also recorded experience that indicates a rapid drop
in flow followed by a partial recovery period [ 1]. It has been
6.4. Effect of viscosity and pressure suggested by some [ 1] that this is due to rapid wear followed
by a settling in which reduces the large gaps worn during the
In developing Eq. (9), pressure and viscosity did not fea- first stage. It is believed there may be other explanations due
ture directly. The effect of both is felt in the value of ~v. to the dynamic behaviour of the gear pump end plates. During
Viscosity may provide a secondary effect on Otw.Pumps 132, the initial stages following contaminant injection, the plates
133 and 171 were tested with a higher viscosity and pressure may be forced away from the gear ends with the passage of
than the remaining pumps, but their respective values for c particles. The increased leakage would be seen as a rapid
are reasonably consistent with others, although 132 does pump flow degradation. As the contaminant breaks down or
appear to be lower than expected. This would tend to reinforce changes, the plates could regain their initial configuration,
the validity of the model. partially restoring the apparent flow degradation. This needs
to be experimentally verified.
6.5. Generaldiscussion It should be noted that the effects on wear of particle shape
and size greater than x,~ should be addressed if the model is
Although the discussion above adds support to the validity to be more realistic. In principle at least, wearmass generated,
of the model, other data would appear to be inconsistent. For sw, could be treated as a function of both particle size (x >
example, pump 84 with a flow of 1791 rain- i, has a relatively x~t) and shape. Particle shape could also be a function of
high value for c. Without details of the pump, it is difficult to x > x~,. Recourse to a distributed modelrather than the simple
establish a reason. In view of its hi~|t pressure rating, it is lumped model proposed here would be necessary. In view of
possibly of a substantially different design to the remaining the relatively good comparison of the simple model with test
pumps, although pump 168 is of a similar pressure and it data, it remains to he a~sesse~dwhether the increase in com-
appears to demonstrate consistency with the other pumps. If plexity and difficulty in determining suitable parameters
the model is valid, it can only be assumed that 84. is a partic- using test data for the distributions featured in a more detailed
ularly sensitive pump and by inference, ~ is uncharacterist- model would result in improvement.
ically high. Alternatively, the design is such that leakage is
confined to relatively narrow paths in comparison with the
pump size. This would have the effect of increasing c. If this
reasoning is valid, an inference is that pumps should be 7. Conclusions
designed to distribute leakage flow over a greater area, rather
than making the pump so tight as to confine leakage to narrow
paths. A model for pump wear has been developedwhich appears
Even though the model infers that damaging contaminant to be substantiated experimentally. Assuming it to be general,
is broken down to smaller non.damaging contaminant, it can pump wear behaviour in any operating system can be simu-
equally well be interpreted as changing from a damaging to lated. It is concluded that it is not the absolute levc~of con-
non-damaging state. Particles may attain a shape which con- ruminants that is of concern, rather, it is the level of damaging
forms to the real leakage surface profile, allowing it to readily contaminant that causes wear. It may be necessary to be able
roll or otherwise manoeuvre between the surface asperities. to distinguish between the two.
R.H, Frith, W. Scott/Wear 196 (1996~64-71 71

8. Nomenclature [21 ISO/DIS9632,Hydr~lic FluidPower- F~d DisplacemntPumps=


FlowDegradationdateto Classi~d AC Fi~ TestDust Contamina~-
Test Method, Draft s t ~ International Offanisafi~ for
c, c~ constants Standardisa~on,1990.
D initial particle mass/unit fluid volume that causes [3] R.H. FrithW. and Scott,Wear inexternalgearpum~- A
damage model,Wear. 172 0994) 121-126.
Dr particle mass/unit fluid volume that causes [4l R.InoeeandE.C.Rtch,'fhe~egapun~prafingsyster~BFPRJournal,
damage at the ith location in a hydraulic system 12f2) (1979) 131-139.
[5] R,H. Frith,A model of gearpump wear due to solidsco~amination.
(D)~ damaging particle mass/unit fluid volume Phi) Thesis,Queens~M UniversityofTechaololff,Australia,1994.
injected at the start of the jth test period [6] MJ. Day and RM. ~ ~ seadfivityof hydrmlic
f(x) function ofx pumps, Comaminati~ in Fluid Power Systems, Conference
Q pump flow Publication, I. Mech.E., Unive~y of Bath.UK.pp. 135-141.
Qo, unworn pump flow [7] BHRA,UK contaminant sensitivity project standard test methods -
Status report. No. BHRA/I/7, British HydromechanicsResearch
Q, theoretical volumetric pump flow Association.Cnnfield.UK, 1983.
r~ coefficient of determination
s particle mass/unit fluid volume that causes no
damage
s, wear mass rate/unit fluid volume
s'~ cumulative wear mass/unit fluid volume Biographies
( =Is.dO
S. total wear mass
V system volume W. Scott: is Professor of Tdb~logy at the Queensland Uni-
.~ particle size (~m) versity of Technology (QUT), sponsored by Fuchs Australia
xc cut particle size (l~m) Pry Ltd. Prof. Scott was educated in Mechanical Engineering
x~t critical particle size for wear to occur (itm) at the Royal College of Science and Technology (now Stratho
clyde University) by part-time study. He spent 17 yeats in
-/ parameter industry before taking a M ~ r of Science Degree in tribol-
a particle disintegration or blunting factor ( < 1.0) ogy and a Doctor of Philosophy on metal grinding at the
~ volumetric efficiency ( = Qo/Q~) University of Leeds. Prof Scott emigrated to Australia in
Or flow degradation ratio ( -- QJQo) 1973 to join the staffof the Mechanical Engineering Depe~-
A leakage factor ( = (Q~- Q) / Q) mcnt at QUT where he now leads the Tribology Research
,~ initial or unworn leakage factor Group.
( = (Q~- Qo)/Qo)
~(x) particle mass distribution/unit fluid volume R.H. Frith: Graduated in Mechanical Engineering from the
~w(x) wear mass distribution/unit fluid volume University of Queensland in 1973 and subsequently were on
X constant to complete a MEngSc degree. He worked for 10 yem-sas an
o wo. ,t investigations and maintenanceengineerwith the Queensland
Electricity Commission before taking up a position with the
Victoria University of Technology in 1995. His field of study
References being control of solid contamination in underground mining
[ ! ] R.H.FrithandW.Scott,Cont~lofconlaminatinnin hydraulicsystems hydraulic systems. He is presently employed by Pinnacle
- Anoverview,Wear, 165 (1993) 69--74. Engineering as the Chief Design Engineer.

You might also like