JUNE 1941· 169

MAGNETIC OIL FILTERS

by J. A. HARINGX.

Filters in which the oil is made to fl~w through a magnetic field .have been constructedfor the remoV:alof particles of iron from circulating lubricating oil. In this articie a descrip-tion is given of such a filter which lias in addition to a good efficiency in capturing ironparticles many other. advantages also, such as very low pressure loss in the oil circuit, nochance of the filter becoming clogged, ease in cleaning the filte~when it is saturated. Theaction of this type of filter is studied. In ordinary use, where new particles of iron are con-tinually entering the oil due to wear on the parts of the machine lubricated, an equilibriumis found to occur which is characterized by a certain residual concentratien of iron in theoil. As the filter gradually fills up, the equilibrium is shifted to higher concentrations. If a .certain limit is fixed for the residual concentration, the time can be calculated which mayelapse before the filter must be cleaned. By means of a series of tests the necessary datahave been obtained. The results are elaborated into a graph In which the influence of theviscosity of t~e oil is taken into account.

, ,

In the pressure lubrication of bearings; gears,etc., it is very important to keep the circulatingoil as clean as possible. If the oil is contaminated,for instance by fine particles of iron from the 'wearon machine parts which rub against each .other,the formation of a continuous film of oil is not onlyprevented, so that the lubricating power decreases,·but in addition the particles cause extra wear bygrinding against the parts in question. Furthermorefinely diVided metals, especially ~on, by catalyticaction, cause acidification of the oil, which is alsoa disadvantage for, the lubricating properties.In' addition to the old text~e or copper gauze

filters which are used_for the purification of cir-culating lubricating oil, magnetic filters have re-cently become more and more common. A magneticfilter of simple construction was described several

. years ago in this periodical t}, whilé several pos-sibilities of application were also discussed at thattime. In the meantime a new filter of improved con-· struction (type No. 7715) has been developed 2) anda series of investigations ón the functioning of·these oil filters' has been carried out. In the fol-lowing article the filter and the tests referred to aredescribed.

............ ';

The construction

In order to be able to explain the essential fea-tures of the new construction, the simplest way isperhaps to consider o/st the construction describedpreviously. This is shown in jig. 1. In an iron housing,a permanent magnet is so fastened that a magneticcircuit with a' ring-shaped air gap is formed. Theoil to be purified flows -through this air gap. Theconstruction is such that the stream lines of the oil

1) L. H. de Langen, Philips .techn. Rev" 2, 295, 1937.2) J. B. Aninga, Polytechn. Wbl., 33, 86, 1939.

. 66.067.3 : 621.318

ah?- the lines' along which the particles of iron are,drawn to the magnet interseet each other at thesmallest possible angles; by this means a relativelysmall lateral acceleration of the iron particles in theoil is already'en:ough to take the~ out ofthe current .and 1;0 the surface. of the magnet .. Although this filter had a very sàtisfactory

, cleaning' power and was able to take up large quan-tities of iron filings, in practice 'various undesired

c

Fig. 1. Magnetic oil filter of the old type. The magnetic cir-cuit is formed by. the permanent magnet A, the iron cap Band the iron housing C. The oil entering at F flows throughthe air gap between A and C past the holes E in the coppercylinder D and leaves the filter at G. The stream lines of the'oil are indicated by broken lines, the magnetic lines of forceare drawn as full lines. .

170 PHILIPS TECHNICAL REVIEW Vol. 6, No. 6

phenomena were found to occur when the filterwas approaching the limit of its capacity of reten-tion. Due to the deposition of the iron the availableopening for the flowof oil, which in connection withsecuring a strong magnetic field could not be madeall too wide, became narrower and narrower, withthe result that on the one hand a considerable pres-sure loss occurred in the filter, while on the other

a

up the requirement the above-mentioned disad-vantages could be avoided.The new construction is shown in fig. 2a and b.

The magnetic circuit is huilt up of a cylindricalmagnet provided with two disc-shaped pole piecesand five soft iron rings placed around the magnetone above the other at small distances apart. Inthe intermediate spaces - the air gaps - strong

6

b

Fig. 2. Magnetic filter of new construction. In the cross section diagram (b) may be seenthe permanent magnet A which with the pole pieces B and the five rings R forms the mag-netic circuit. The oil entering at F flows via the holes E in the pole pieces past the insideand outside walls of the set of rings, and thus not through, but along the air gaps.The housing C and the cylinder D are of non-magnetic material. The oilleaves the filterat G. The rings are in two sections, as may clearly be seen in the photograph of the openedfilter (a), and are held in place by brass strips S welded to them. The whole magneticsystem also rests upon the ends of these strips. After unscrewing the cover of the housingthe magnet system can be drawn out of it by means of the handle provided.

hand a wisp of the deposited material was sometimestorn away by the oil and passed into circulationagain. Furthermore the cleaning of the filter whenfull was fairly difficult, since the magnet retainsthe deposit very firmly, a fact with which it maynot be reproached since that is its function.

In the new construction the idea that the streamlines of the oil and the lines of attraction of the ironparticles should interseet at only small angles hasbeen given up. Considering the high strengths of themagnetic fields which can be obtained with themagnet steels available at present this conditionis indeed no longer so important, and by giving

magnetic fields occur. The oil now, however, flowsnot t h r ou gh but along the inside and outsideof the air gaps, as illustrated in fig. 3. It may beseen that it is actually the sp rea d lines of force ofthe magnetic field which draw the iron particlesfloating in the oil out of their original paths andcause them to enter the air gaps. Due to the factthat the iron particles are now deposited 0uts i d ethe stream of oil, the tearing off of bits of the de-posit has been made impossible, while at the sametime no appreciable increase in loss in pressure canoccur as the filter gradually fills up. In the teststo he described below the pressure loss, with a flow

JUNE 1941 MAGNETIC OIL FILTERS 171

of 500 l/hr and a viscosity of the oil of 35 cp 3), wasfound to increase during use only from 0.05 atm.in empty condition to 0.10 atm. in practically fullcondition of the filter. A pressure loss of this smallmagnitude is of no practical significance; at least,

III II I+ +J8~/l

Fig. 3. Sketch of the stream lines (broken lines) and the mag-netic lines of force (fulllines) in the neighbourhood of one ofthe air gaps.

compared with the pl-essure losses at this rate offlow in textile filters, it may be neglected. The be-haviour of the pressure loss for other values of theviscosity may be seen in fig. 4 where curve a holdsfor the empty state and curve b for the practicallyfull sta te of the filter.

The cleaning of the filled filter could also be madevery much easier with the new construction thanwith the old. For this purpose each of the rings ismade in two sections which are held free in positionby brass strips. For cleaning, the whole magnetsystem is removed from the housing and the halvesof the rings can easily be drawn off the magnet.Due to the fact that the rings are now no longer

0,

2----_----/ +--- ---,........ ......---_V f..--"" r--V -----1

~ hf-" -

l% la Ilf 1 16 18 P 22 24 2fi ~

atm0,4

0,3

0.2

OE

~ ~ ~ w ro w ~ ~ ~ ~ ~~J85/B

Fig. 4. Pressure loss in the magnetic filter with a rate of flowof 500 l/hr as a function of the viscosity of the oil in cp andDE, respectively. Curve: a holds for the clean filter, b for thefilter filled with 7 g of deposit.

3) A liquid has a viscosity of 1 poise (p) or 100 centipoise (cp),when a shearing stress of 1 dyne/ern" is necessary to main-tain in the liquid a velocity gradient of 1(cmfsec)fcm, At20°C,for instance water has a viscosity of about 1 cp, rapeoil one of about 100 cp. Technically the viscosity is usuallygiven on an empirical scale (Engier degrees, OE) which isalso indicated on the abscissa of fig. 4,. The recalculationinto cP. is impossible with strict accuracy.

magnetic the deposit can be rinsed ~ff withoutdifficulty with some liquid or other such as petrol,trichlorethane or the like. To prevent particles ofiron being deposited on the permanent magnet itselfa non-magnetic cylinder is introduced between themagnet and the rings (D in fig. 2b).The quantity of deposit which the filter can take

up is proportional to the total volume of the airgaps. The larger this volume is, however, the lowerthe magnetic field strength which can be obtained.A suitable oompromise had therefore to be found.With the chosen dimensions of the air gaps (1.5 mmlong, 275 mm'' cross section, thus a total volumeof 6 X 415 = about 2 500 mm3) a total of about10 grams of iron filings can be taken up, while amagnetic field strength of 8 000 gauss is obtained.A second compromise was necessary as to the

cross section of the channel of flow for t.he oil. Thischannel must be made narrow so that the iron par-ticles to be captured pass as close as possible alongthe air gaps. On the other hand less oil per hour canbe dealt with when the channel is p.arrow, if thevelocity of flow of the oil is not to. be increased,

Fig. 5. Part of the lubrication system of an 800 h.p. Dieselengine. In the foreground six magnetic filters may be seenwhich are connected in parallel in the oil line. In the cylin-drical tank on the left in the background there is a coppergauze filter which serves to remove coarse impurities fromthe oil.

172 PHILIPS TECHNICAL REVIEW Vol. 6, No. 6,

thereby increasing the chance that, some iron par-ticles escape being captured by the .filter. The di-~ensions .finally chosen. (cross section 550 mmê)make it possible to deai ~th 500 litres of oil perho~. In special cases where greater flow capacitiesare necessary several filters 'may he+eonnected in'parallel in the oil line. An example of such an in-stallation with six filters is reproduced in jig. 5..

Further consideration of the action of the mter. / '

, When the magnetic :filter is installed in a closedsystem in which oil circulates which is very muchcontaminated- by iron, the oil is seen to become'clear only gradually, for instance within fifteenminutes 'or a half hour" This shows that all the ironparticles are not caught by the filter at once asthey pass it, but a certain fraction only each time:the oil must pass repeatedly through the filterbefore the concentration of. iron falls, practicallyto zero. If a certain quantity of new: iron is con-tinually formed in the system, the concentration'is found never to fall below a certain value.

As we shall see this Iimrting value of the coneen-tration is indeed very much lower than the .concen-tration which would be present in the oil withoutthe use .of the filter; but the residual ooncentrationscannot in 'ge~eral be entirely neglected .. It. wastherefore desirable to investigate the action of thefilter more closely... ; . . '

The residual eoncentratien

" Fig. 6 shows diagrammatically the situation ID

which the :filteris used. The lubricating oil, which iskept in eirculation by a pump, comes from a reser-'voir, passes first through the filter, then along thepart of the machine to be lubricated and finallyreturns to the reservoir.'. Suppose that the oil in the reservöir has an ironconcentration of Cl' If V is the amount of oil flowingthrough the filter per unit of time, the :filter must

RJ8$19

Fig. 6. Diagram of a lubricating system with oil circulation.R reservoir, P oil pump, F magnetic filter, 'M part of. themachine to be lubricated.. The oil has an iron concentrationof Cl preceding the filter and iron concentration of C2 afterpassing it.

" ,

deal with an' amount of iron particles equal to ClVper unit of time. As a first approximation we maynow assume that the same fraction of this quantitybrought to the filter is always retained by it. Theamount of deposit already caught by the filter,which we call G, thus grows per unit of time by theamount

LIG =' y clV, (1)

where y is a factor smaller than unity which weshall in the future designate as the retentioncoefficent.If Q is the amount of new sediment formed per

unit of time, Q may be larger or smaller than LIG"or equal to LIG. When Q>LlG more sediment isintroduced into the circulating oil than is removedfrom it. The concentration Cl of the iron will there-fore gradually increase, at the same time, however,according to (1) the amount captured per nnit oftime, LIG, also becomes larger until the additionof iron to and its removal from the oil are justbalanced, i.e. until Q = L1G. In the same way w:henQ<LlG the concentration Cl will gradually fall until'LIG has also fallen so far that Q = LIG. It is thereforeclear that in any case after some time a condition,of equilibrium will be established .where

From this we can calculate the residual coneen-.tration' Cl which may be expected. We must,how~ver,. keep in mind that the concerrtrations ofthe iron preceding and following the filter are notequal. Of the amount of iron entering the filter ClVa part YCl V is held by it. Thus only the amount(f-y)cl V leaves the filter, i.e. the concentratien é2of the iron in the oil conducted to the part of themachine to be lubricated amounts to

C2 = (l-y) Cl' •

From this with (2) it follows that

1-y' Q'c2 = ---

Y V

(3)

(4)

When the. filter has functioned during a timein the equilibrium state, with a constant addition(Q) of new sediment to the oil; it has taken up atotal quantity of iron '

G= Q' t • (5)

To give a numerical example: suppose that in themachine Q = 5 mg/hr of new iron filings areformedand that the retention coefficient has the apparentlyvery low' valu~ of r = 0.01.' If ~e substitute for' V.the above-mentioned value of 500 l/hr, C2 becomes'oilly 1 mg/I.

JUNE 1941 MAGNETIC OIL FILTERS ' 1:73

Wi~h a retaining capacity of Gmax' = 7. grams,according to (5) the filter must be cleaned after, 1400 hours of use. Without the filt:~r after 1400~our~, when ä total of 50 litres ~f oil take -part inthe circulation there would be an iron coneen-/ration of 1~.0mgfl4).

Further consideration of the ~et~ntj,0Ttcoefficient

How large is the retention coefficient y of thefilter described? It is impossible to give an im-mediate answer. to this question since the above.assumption of a constant value of y is by no meansjustified. The retention' coefficient depends upon'a number of factors, the chief of ~hich are: thedegree G to which the filter is filled, the iron con-c_eD;tratión Cl in the oil, th~ viscosity o.f'the oil,the rate of flow and the, size' of the iron particles.The last three factors remain approximately ~-.cha~ged during use of the filter; or at leas; theymay he kept' constant in the experiments to be de-scribed in the following. This is; however, not trueof the first two, so that we must extend our con- ,sideration of. this point somewhat. We' shall first'discuss the' influ~nce of the degree G to which thefilter is filled:

When the filter is entirely filled with the dep~sit. y must in any case fall to zero. This transition to

; J zero does not take place abruptly, whi~h is under- .standabie when it is kept in mind that the magnetic :field itself is .influenced by the iron particles taken'up., The field strength will become greater due tothe decrease in the magnetic resistance' of the airgaps. At the .same time, however, the -spreading,w,hich is just what we must have, will become less.This last effect is found to dominate and for the

G'J8520

Fig. 7.,Approximate' form of.the curve showing the variationof the retention coefficient y with the' degree of filling G ofthe filter.

4) In general before such high concentrations are reachedthe oil w~llhave to be renewed. - ,.

).

behaviour of y with G a curve of the form sketched 'in fig. 7 is obtained. ,

The result of this variation of y 'with G is that,. strictly speaking,' we may no longer speak of adefinite equilibrium condition: during use, as· thefilter slowly fills up, 'y will 'decrease slowly and the-iron concentration in the oil ~l gradually increase.Nevertheless upo:q..closer consideration it is foundthat when' y does not, vary to~ much ~th G, acondition which is always satisfied in 'use, the suc-cessive states 'of the system may still be conceivedof as a gradually shifting equilibrium. -This. meansthat we may continue to use equation (4) for thecalculation ofc2, with a different value of y at everymoment corresponding to, the degree' of filling G,of the filter at that moment.If the requirem'ent' is made that the residual

cóncentration C2 of the iron may not exceed a eer- ."rain value, it includes the requirement that thefilter must be cleaned, not whtin: if / is full, butSOOIl,er,namely when the :'retention c~~fficient hasfallen to that value of y which a'ècofding to (4) 'cor-responds to the permissible value '~f r.2~:~From agraph like that' of fig. 7, one may tlÎen re~'d -off atwhat degree of filling G the filter' should be cleaned,and from equation (5) the time' çan he caiculat~dwhich.is necessary to reach this degree: Equation (5)is no longer exactly valid, since due to the slow'shift of the equilibrium a certain 'amount of the .newly formed sediment is used in increasing theiron concentratien ofthe oil. Nevertheless, éoiisider-ing the lo~ concentration and the small quantities . ' 'of oil which are usually used for the circulation, thedeviation from equation (5) is onlyslight," ":." The' second factor which varies during use .andwhich affects the retention coefficient is the concen-'

"tration Cl itself. lts effect is. easily understood'qualitatively. Thé diagram, of the lines of forcesketched in fig. 3 will be somewh'at altered when themagnetic resistance is' decreased on both sides of- .~ "the air gap proper by the presenct? of a large n~mber'of iron particles. The "spreading" increases as it ''were with the iron concentration of the oil' and at,. .. ,the same time the attractive effect of the field also .

'. increases. This. effeçt is found to ;he:quite consider-able in magnitude.

The fact that y depends upon Cl can be taken intoac~ount i11;quite the ;ame way B:sabove. For everystate characterized by a definite value of G and ofCl there is a certain value of the retention coeffi-cient y. If we represent this. relation graphically'by plotting y as a function of G for a series of values 'of cl (jig. 8)'-::wè can. again ~elld off. the degree offilling G of the filter at which it must be cleaned,

,"

.1

".

174 _PHILIPS TECHNICAL REVIEW Vol. 6, No. 6

'when a given maximum residual concentratien c2is prescribed. y can be calculated from (;2 according

'1'

Fig. 8. Approximate form of the family of curves which in-dicates the retention coefficient y as a function of the degreeof fillingGwith as parameter the iron concentration Cl at theinlet to the filter. If a.givenreeidual concentration C2 i~permis-sible at the outlet of the filter, with the aid of equations (4)and (3) y' and Cl can he calculated, and the permissible degreeof filling can then be read off in this graph. '

• • # .,

to (4) and cl according to (3). We .shall see belowhow this -graph can be drawn somewhat moresimply; the principle, however, remain~ the sam~.

Experimental determinatien of the retention coef-. , ficient

It is now a question of studying more closelythe relation J' (G; Cl) sk~tched iufig. 8 for our filter,~hich can only he done experimentally. The 'neces-

. sary experiments are, preferably related as muchas possible to practical conditions. One of the con-diti~ns ,should therefore be a constant and regularlydistrihuted addition Q of iron particles to the. oil,

<and in order .to 'include the whole range of degrees .', of filling the experiment. for each value of 'Q (with.the corresponding series of concentrations Cl which'are traversed) should be contiIi~ed for a long time,as is apparent from the nu~nerical example describedabove. In order to avoid this we have arranged theexperiments somewhat differently. A certain 'quan- .tity (H) of.iron filings was added once only to the~i~cttIating oil (total amount v). I~ contrast to the'usual case ii:t which the iron concentratien increases·'very slowly as the' filter becomes full, we here havea fairly rapid decrease of the concentration, and theexperiment is concluded in.a few hours. The' va-riation of the concentration Cl with the time isdetermined by measuring c~ chemic~riy. at specifiedintervals. The retention coefficient y is then foundas follows. At each momen.t }he following 'is true:

. '.

The; amount of iron G retained III the filter thus .grows per unit of time by

dG , dCI-=-v-·dt dr

On the other hand, according to the definition of theretention coefficient '(equation (1)) this growthwas given by' YCIV. Therefore'

, dcl•Y clV=-v-,dt

V d In-c.y=---_.V dt

(7)

Thus if we plot In ci against t, y follows from theslope of the curve obtained.

In all èxperiments the maximum rate of How ofV= 500 l/hr was used, while the viscosity of the oil,which varies very much with the temperature, waskept constant by placing the reservoir in a ther~o-stat. The experiments were carried outwith a veryfine carbonyl iron powder which satisfactorily re-sembles the sediment' .most commonly formed inpractice as far as shape and size or' the particles(about 2 microns) are concerned. After the additionofthe powder (G = 1 to 10 g in v = 5 litres of oil),the oil was first pumped through the system forseveral hours with no magnet in the filter, in orderto distrihute 'the iron evenly throughout the 'oil.. Special care had to be taken that all the oil tookpart in the circulation, i.e. that there was no direct

, c:

(6)

,]8522

Fig. 9. Arrangement.for the experiments for the determiIiationof the retention coefficient. R reservoir, T thermostat, P tooth-wheel pump, F magnetic filter, ~, k2 taps. In order to avoid.a direct current in the reservoir between inlet and outlet,which would hinder the uniform distribution of the ironthroughout the oil, the inlet ends in a vessel B over the edgeof which the oil flows in all directions, There is a small hole inthe bottom of the vessel to prevent any iron which has' settledfrom being removed from the circulation. At specified intervals. a small quantity of oil is tapped off through the line L and itsiron concentratien is determined. At 1nl, m2 a differentialmanometer. was connected which indicated the pressure lossin the filter. . .

,JUNE 1941

"

MAGNETIC. OIL FILTERS 175

Table I.. Elaboration of the. series of measurements 2 and 5 (as ex~mples). The rows with large numbers are' derived directly fro~the' curves 2 and 5, respectively in 'fig. 10, while the rows with small numbers are found-by interpolation.

-'

current in the r~servoir from inlet to outlet: 'Fig. 9.sho~s how this was accomplished,

. The apparently so simple experiments still contained manystumbling blocks. While the circulationpump is getting underway, for example, there is great danger of the formationof small air bubles in the oil, which later have no opportunityof escaping fro'm the rapidly flowing oil and which uponpassing through the. filter may afrect the deposition of ironparticles considerably by their surface tension. By starting, the pump as gradually as possible the formation of air bubblesis almost entirely avoided. Furthermore a small' amountof iron powder is found to cling to the walls of the circulationsystem at various points, which results in the fact 'that theiron concentratien 'is' slightly lowêr than corresponding to'equation (6). This can he taken into account by using a cor-rected value for v as well as for H. Further, the concentration

• Cl in equation (6) is actually an average value over the totalquantity of oil, while the action of the filter is determined bythe concentration at the inlet to the filter. A correction must'also be applied for this by correlating the slope (y) found atpoint t in the In cl-t curve with the average value of Cl'

over an adjacent time intervnl e; • is here the time necessaryfor the total amount of.oil to be pumped once around the sys-tem, in our ,case about Bû sec. Nevertheless, after all these,corrections, the results 'still show considerable divergence,"caused by the limited. accuracy of the measurement of the .concentratien. TIlls measurement is made by tapping from theoilline in fig,' 9 a sample of 25 cc of oil, "igniting" the oil anddetermining the iron content of the residue.

Rèsults of measurements.. .

In the manner described three series of measure-ments were carried out' in the first instance, ..in,which 13/4, 5 and 10 g of iron powder, respectively,. were added to the oil:lnfig. 10 th~ variation ofthe .measured concentration Cl is given as function ofthe time elapsed after the insertion' of the. magnetfor these three casés (Î, 2, 3). If we now consider'for instance curve 2, for every point èl, t_ of thecurve we can 'determine the values of'y, C2 arid G

'corresponding to that value of Cl according to equä- .tions (7), (3) and (6). In table I the result of thisdetermination is given for a series of' values of ë~for curve 2.

{l tt.000

600400

1\.,,'

·1'" \\ .60

\ \' I\.""40

2() J"~"".......tla

_i~'.5 ~ "" ...

6 ... ~4 -, ~ ..., t'S"

[).....~r-.... - ~ r---2" r--

I r-= I-t

my

o 20 40 60 80 100 120 140 160 180 200 mmJ8S2J

Fig. 10. Variation of the iron concentration Cl (in mgjl) withthe time t (in min) in different experiments. At the beginningof each experiment quantities of iron were added to the oil:H == 1.75 g (curve 1), 5 g. (curve 2) and 10 g (curve 3). In Cl

is plotted against t in order to find the retention coefficient y by .graphical differentiation according to equation (7). The brokenline curves 4 and 5 with H = 5 and 4 g. respectively, werereeorded with a higher viscosity of the oil; namely with 1} = 85cp, while in curves 1,2. and 3 the viscosity was 35 cp.

,According to the discussion abo~e, for every valueof the para~eter Cl we would have to plot the J;e-.tention coefficient y as a function of G. Th~n onthe graph obtained, beginning ;w-ith·r and Cl' thepermissible value .of :G could be read off. In the,practical application, however, we must assumethát the values of Q [the quantity of new sedimentcontinuously formed in the part of the .machinebeing lubricated) and of C2 (the permissible residualconcentration] 'are given, and that r and Cl mustthen be calculated with the help of (4) and (3). It .

Experiment 2 , Experiment 5

6.16

y C.

mijlGDig ; mg

Clmg!l

50

0.0311

0.0084

G '-, C~

mgj1y

786 - 0.260'0.237 77.6 41.0 100

580500

321100

4620

18.61053

2

o830 ,

24103853

41874343

4350440244314443

400 0.196o.uo0.0860.068

0.0680.0520.0360.0250

0.0172

50"'0.730

20105

0.274"0~0931

0,0384 '

0.01742 4449 3

753

400

0.15~0.141

0.126. 0.081

0.05850.0415

0.04150.0263

0.0260.,0.0123

0.0056

638500

350100

4720

19.210

9.755

3

20

o1070

18683180

34503590

35943640

36413665

3675

4.40 10,8

0.434 1.065. " ~

0.135, 0.331

100.0763 "

0.0206

176 PHILIPS TECHNICAL REVIEW Vol.,6, No, 6 "

is therefore simpler to make the graph in sueh a For each curve for one value of C2~ each of theway that the given quantities, in addition to the three series of measurements 1,2, 3 gives one pointrequired value of G, occur in it directly. In order (a pair of values of Q and G), so that we wouldto do this we choose C2 as parameter with the values have to draw each 'curve with the aid of only three. of 11/2, 2, 3, 5, 10 ... mg/l, for example. In table 'I measured points. This would be quite 'difficult con-several of these values are :filled in by intrapolation. sidering the fairly great divergences. Two other(by means o~ auxiliary graphs) and in each row the 'series of measurements (see below) performed by aquantity slightly different method, however, provided two

more points for every curve, so that the shape o,fthecurves was somewhat better determined. The familyof cm.:ves obtained is reproduced in fig. 11.

The influence of the viscosity

While in the above in the theoretical as well asthe experimental considerations we have only in-vestigated the influence of the' degree of :filling Gand the iron concentration cIon the retention co-efficient, we shall in conclusion also study the in-fluence of the viscosity of the oil. The fact that theviscosity will influence the action of the magnetic

, filter is easily understood: the more .viscous theliquidf the' greater the force necessary to removea particle from it. At a given rate of flow the mag- .netic 'field will capture the iron. particles from a,thick oil less well than from a thin oil.'

As in our experiments, we mayalso assume thatfor normal use in a given installation, in which theworking te~perature and type of oil chosen ate

'50 fixed, the oil will ha~e' a constant viscosity. ·Agraph like fig. 11 can therefore be used, with a sep-arate 'curve for each viscosity. Actually, however,there is a still simpler method. .

The three series of measurements 1, 2, 3 in theexperiments with the arrangement, described wereobtai~ed with a single definite viscosity of the oil,'namely 1] = approx. 35 cp. Two additional experi-ments (4 and 5) were performed with a viscosity,of 1]' = approx. 85 cp. The variation ofthe measured, concentration with th~ time for these latter cases,'~here IJ '-- 5 and 4 g, respectively, is given as a'·' .broken line in fig. 10. It may immediately be seenthat the retaining action of the :filter is sm~ller'since the iron concentràtion decreases more slo:wly.

o If, however, the retention coeffièient y as well as the.>8524 values of C2, G and YC2 V/(l':_y) " Q' (see tabl~ I'

Fig. Ll. The quantity yc2V/(1__;'),'whlch is equal to Q i~ the; where this has .been done for curve 5), are againequilibrium state of normal functioning, is here plotted as a I I i d fr th d if . . bfunction of G with C

2as a parameter. For every curve C2 there ca cu a e om e curves, an I ag am as a ove

are three measureq points av~ilable from the series of measu.re- '. the points of-a family' of curves as in :fig. 11 are .ments 1, 2, 3, while two senes of measurements 4, 5 carried 'd '. d fr h I h .' ~ d-out with a different viscosity of the oil provided two more etermme om tese va ues, tese points are roun •points for every curve. (For,the broken line curves the quan- to:fit very well in the old family of curves. when thetity YC2Vj(1-y) here plotted may not be set equal to Q, since I f thei din' t Q' ulti r d b theat such high concentrations the change in the quantity of va ues 0 ell. or l:l es . ,are mu lp ,IC Yiron present in -the oil may no longer be neglected. The curves ratio of the viscosities 1//1J __:. 85/35. The agreementare, however, of scarcely any importance, and are here given " f b t d -ith th fai I I ;only for the sake of fixingmore completely the shape of the' ,~as ar as may e expec e WI ne arr y arge

, 'curves of thè whole family). ' 'divergences already' mentioned ~ is so good that

'"

,,' ~." ,.,

yc2V· .'-=Q ..l-y ,

is calculated and indicated in the :fifthcolumn. (Forthe rate of flow Vwe always use the maximum valueof 500 l/hr,"with which the experiments were alsoperformed). If we now,plot this quantity as 'a func-tion of Gwith C2 as a parameter, the required valueof G can be read off directly from the given valuesof Q and c2•

'Q(g/h)

.' ..

(8)

. , .~

JU~E 1941 M'AGNETIG OIL FILTERS' 177

,we used the, two' new points obtained in' this wayfor every C2 curve to draw a smooth curve throughthe points, '.

The result is therefore that' the same fa:prily'of curves is obtained if for experiments with the

"viscosity 1/ the quantity 'r/Q' is' plotted and forexperiments with the viscosity 1] the quantity Q'YJ.We have therefore redrawn the graph of fig. 11withthis ordinate, see fig. 12, so that it may be :used for any desired viscosity. At the- same timea number of interpolated curves are given in thisfigure., '

'A numerical example will serve to illustràte the use of thegraph: A large gear box in which about 25 mg of sediment is,formed per hour (this apparently very high' figure actuallydoes occur in practice) is lubricated with 1 000 litres of oil perhour. Since the maximurn citpacity of flow of the magneticfilter here described is 500 l/hr, two filters in parallel musthe installated in the oil circuit. At the working temperaturetheoil has a viscosity of 50 cp. The problem is to deterniine~hen ih~ filters 'IÎ:lustbe cleaned if the iron' concentratien may'not exceed 2 mg/l (in: general a ,con~ide~ablygreater concen-tration will he permissible in a gear box). •, Per filter the.amount of new sediment introducèd is Q = 12.5

mg/hr. If in fig; 12 we jrrtersect the horizontalline correspond-ing to 0.0125 X 50;';: 0.625 by the curve corresponding to'C2 = 2 mg/l, then: at the abscissa of the point of intérsectionit may be read offthat each filtermay become filled to a weightof G = 3 grams of iron particles. The time necessary for this ,is 3/0.0125 =:= 240 working hours. After that time the filters"must be clèaned. (If one waits for instance 600 instead of24,0 hours Gbecomes 0.5 g and according to fig. 12 the residualconcentratien ha~ still only increased to 3 mg/I). .If it were desired in the above case to keep thè iron concen-.

tratien below 2 mgfl without the use of filters, thèn perworking hour 251'2 = 12.5 litres of fresh oil woûld have to beadded. This illustrates very clearly the advantage of theuse. of such magnetic filters. '

In conciusion a few words may 'be said about,thc permissible val~e ,of the iron concentration c2 ~

(residual concentration). This quantity, like' theamount- of sediment Q, will vary very much 'ac-.oording to the nature of the part of the machine,lubricated; with accurately finished axles and.bearirigs .P2 must' be small, for instance not morethan several mg/I. If at the same tinie a high' value

ofQ'and/or 1]occurs it might be' that the correspond-ing horizorrtal Tine in fig. 12 lies in its entirety,higher th~n, curve c2• This would mean that the..retention coefficient of the 'magnetic filter is frómthe very'first unable to keep the residual concen-:tration below the desired, value, In such a caserecourse may be had to two or morè magnetic ~filters connected iii se rie s in the oilline - solution'~'ihichis possible because of the very slight pressureloss of these filters already mentioned. Experiencegained urrtil jnow, however, indicates that for byfar' the majority of cases the retention 'coefficientof one filter is more than adequate.

"

Q~ (cp.gjh)100

t:;rr;;: ~~

t-. :-... rvt-.t-. .... . .... r;

201'-1'- t-.t-.

r-.:-...r-. r-.. r-K1

101'- I'- r- r-.

5i"t- r-

1"-, r-.. :-...:-... :-"'r-.2 I"-r-.. ..........r-.. r-..t-. .....r-. :--r-.

1 r-.. r-.IC'> ' - ,

~5r-. .

t-- i ,:-... r-I r-..r-. "-,

"tr-. r-DJ r- It:;

5

:2

0,0

0,0 0 1 2 3 4-, 5 6' 7 8 9 10 G(g)

Fig. 12. Graph for use in the practical application of the mag- ' ,netic filter described. With -a given viscosity 'I of the óil (incp), a given quantity of-new sediment Q (in g/hr) and a givenpermissible residual concentration C2 of the iron (in mg'll),,~itcan beread- off on this graph to what degree G {in grams ofiron) the filter may become :filledbefore it must be cleaned.G/Q is then the corresponding number of working hours. Thegraphis valid for a rate offlow of 500 l/hr and for iron particles2 microns in size. ..

, '

"