]\gt ; On the Flow of TWater throvgh Pipes Passages By Prof. A. H. GIBSON , D.Sc .
, University College , Dundee .
( CommuIlicated by Prof. J. E. Petavel , F.R.S. Received December 10 , 1909 , \mdash ; Read February 10 , 1910 .
) CONTEN TS .
PAGE 1 .
Introduction 366 2 .
Description of Apparatus\mdash ; Method of correcting Results for Friction 3 .
Loss of } in a Circular Pipe with Diverging Boundaries 370 4 .
Loss of Head in Pipes of Square Section with Diverging Boundaries 374 .
Loss of Head in a Pipe of gular Section with Diverging Boundaries 375 6 .
Loss of Head in a Circular Tube with Converging Boundaries : Critical Velocity .376 .
Summary and Conclusions 377 1 .
\mdash ; As is well known , boundaries in a passage conveying a fluid tend to unsteadiness of flow with consequent loss of head in eddy formation , the energy loss in such a pipe bein , greater than in a parallel pipe with the same mean velocity of flow .
boundaries , on the other hand , have an opposite tendency , and the present ation , which was commenced in laboratories of the Manchester University , and has been completed in the engineering department of !
University College , Dundee , was undertaken with a view of determining : : how this loss depends upon the angle of divergence of opposite sides of a circular pipe with uniformly diverging boundaries ; ( 2 ) the relative values of the losses in a circular and a square pipe having the same angle of ] divergence , or increasing in area between the same initial and final limits , in the same length ; ( 3 ) the effect of a divergence in the one pair of sides of a rectangular pipe , the second pair being parallel ; ( 4 ) the effect of a variable , as opposed to a ular , rate divergence sides ; ( 5 ) the loss of head 1. .
and , if possible , the critical velocity in a circular pipe with unifonuly converging boundaries .
Twenty-five pipes in all were constructed for use in the experiments .
These were very carefully made of wood , finished off with a coating of shellac varnish .
They may be divided into three sets .
A. Circular Pipes having a Straight Taper.\mdash ; These had a uniform smaller diameter of inches and a larger diameter of 3 inches .
the angle included by opposite generating lines of the cone of which the pipe forms a 1 On the flow of through Pipes ssages , etc. 367 frustrum , the following table shows the range of values of coyered in the experiments .
Length of pipe in inches ) Length of pipe in inches *Sudden enlargement .
In addition to these a circular trumpet-shaped pipe of the same length , inches , as the ) above , but designed so as to ) erive a uniform rate of retardation to the water , was prepared .
B. Pipes of a Straight Tapcr.\mdash ; Here is taken as the angle between opposite pairs of sides .
Ihree pipes were used with of inches , inches , and inches , and with equal respectively to , and .
The at the small and ends were eqnal to those of circles of inches and 3 inches diameter , squares of inches and inches sides .
C. Pipes of , Parallcl as to the ( ?
to the other Pair of Sidcs.\mdash ; Three pipes were prepared having equal to , and , and lengths of inches , inches , and inches .
In each case the smaller end formed a square of inches side , and larger end a rectangle inches by 1329 inches , the areas equal to those of circles of inches and 3 inches diameter .
As it was thought probable that a pipe uniform retardation , 01 one giving uniform change of velocity per unit length of pipe , would cause smaller loss of head than the taper pipe of the same , and of the same initial and final dimensions , it was decided to test this by comparison with the straight taper pipe having In the former of these pipes ( retardation uniform ) we have or constant ; thus ) , so that , ( 1 ) where is the width of the pipe at a distance from some datum , .
the ( known ) width at the entrance to the pipe .
If / 1 is the ( known ) width at exit , and if is the length of pipe , In the second pipe constant since , we have ( ) where and with flow taking place from to .
The orifices actually piercing the walls of the bushes were not more than inch in diameter , and great care was taken that no burr should be left on the interior surface of the bush after drilling these holes .
In the square pipes the pressures were measured at two points in a crosssection ; in one corner , and in the middle of one face .
The pressures at the two points were sensibly the same throughout .
Where any slight difference was noticed , as was the case in the pipe having the largest angle of gence , the pressure was rather greater in the corner , and the true pressure has been taken as the arithmetic mean of the two .
In the rectangular pipe the pressure was measured at three points in a cross-section , , at the centre of each face and in one corner .
Generally speaking , whenever any difference between these pressures was noticed , that in the centre of the shorter face was , and that in the centre of the longer face , i.e. of the parallel sides , was lowest .
This difference was small , however , in every case ( generally less than feet ) , and where it did exist the true pressure has been taken as the arithmetic mean of the greatest and least of these .
After the first few experiments with diverging flow it was found that , owing to the equalising of velocities in a cross-section , there was a gradual rise in the boundary pressure in larger bush in the direction of flow , further experiments showing that this rise in pressure extended over a distance of as much as 12 inches in the case of the circular pipe a sudden enlargement of section .
For this reason a much tYer bush was prepared for the discharge end of the pipe and a series of pressure were made , the pressure being taken at a sufficient distance from the ertd of the conical pipe to ensure of this action being collpleted .
This distance varied from 7 to 13 inches , ) eing greater with the greater angles of divergence .
The frictional loss per unit length of each bush was measured for a series of elocities of flow covering the whole , and from curves plotted from these results the friction losses due to those portions of tho bush between the ernls of the experimental pipe and the pressure have been determined for each experiment , and have been taken into accolmt in determining the effective difference of pressure at the two ends of the pipe .
In some cases the pressure was also measured at point midway along the pipe .
Loss in \mdash ; The friction loss in the conical pipe , as distinct from the loss essentially due to the ence of its ) OUlldaries , be determined on the assumption that it varies , as the power of the velocity as in a parallel pipe having a physically similar inner surface , to be given by feet of water , and assuming that the friction loss in a small length of a similar , but taper pipe , is given by , we get , for the same friction loss in a parallel pipe as in the taper pipe , velocity of flow when passing the same quantity of water as the taper pipe .
; so that for equal frictional losses , writing , we have base oengths , eight oiagram gives Vfromwhere Diameter oquivalent parallel phere VheIn tips under consideration , preliminary experiment showed n ; !
On plotting velocities corresponding to a iven flow in the taper pipe , which and so the diameter of the equivalent parallel pipe , giving the same friction loss as a taper pipe of the same length , can be determined .
In this case the pipe equivalent to the circular conical pipes is inches diameter , and such a pipe , 26 inches long , was prepared .
A series of experiments was carried out to determine the friction loss in this pipe for difl'erent flows , and from the curve showing these results the friction losses in the valious taper pipes have been determined .
11 3 .
Loss of Bead in Circular Pip ) with uniformty Diverguvg .\mdash ; Calli ( the mean velocity and the mean pressure in the smaller , and and those in the larger pipe , it llay be readily shown that , the effect of the differences of velocity at different points in the cross-section of the stream , in the case of a sudden enlargement of section if the pressure over the plane surface forming the junction of the pipes is the same as that in the ; smaller pipe ( an assumption which appears to be fairly well justified by experiment ) , the loss of head , is theoretically equal to As this expression depends only on the initial and final areas of the pipes , it has been thought advisable to express the loss of head in the taper pipes in each case as a percentage of this quantity .
On plotting the results of the whole series of experiments on circular !
SS pipes , it appears that for values of , greater than about ' feet per second and up to about 23 feet per second ( the highest velocity attained in the experiments ) , the percentage loss of head does not vary in any regular manner with the velocity , and appears to be sensibly constant .
Below about 5 feet second the percentage loss falls off somewhat rapidly with decreasing 1909 .
] th Pipes , etc. velocities , though the results reat i means of all observations above 5 feet per second have been plotted on a base the difference in the velocities of adjacent layers of the fluid , and , in consequence , the less the loss of energy in impact .
Where , owing to the cutting off of the corner at " " the free backward eddy formation is restricted , this will have the effect of increasing the resistance to the removal of the annulus immediately surrounding the stream , and consequently of the impact loss .
The critical value of , at which this increased loss becomes exactly balanced by the reduced loss due to the more dual divergence of the stream in succeeding portions of the tube , is apparently about .
The restriction of the backward flow to the face at " " \ldquo ; will be accompanied by a reduction of pressure over this face , and this effect is well marked in the experiments .
Calling the pressure at the centre of the length of the portion of the tube , and the pressures at entrance and exit , the following table indicates how varies with the of ence : 1909 .
] Water through Pipes and ssages , etc. 373 From to 1SOo the pressure over the faces is less than that in either the small or parallel pipes , while this pressure has its least value for a given velocity and a given pressure in the ] pipe when is about 1 Values of to and 18 have tely the value for so that the losses in such tubes should be equal .
From the curves of it pears that this is the case , the ]osses 103 per cent. when and per cent. when It is a fact that the semi-angle of ence ( of tube the maxilnum loss is , practically , identical that of obliquity of a square inclined plane from to which is accompanied by the maximum resultant pressure on the plane when exposed to the impact of a stream of water , or of air .
he only point , to be considered in this connection is to whether the loss in a circular pipe with uniforlnly diverging boundaries , expressed as a of the difference of kinetic energies per pound at the } ends of the pipe , is the same for a iven , for all ratios of initial and final areas , and for all values of these areas .
While the present series of expel.iments is incoml ) let in that it does not , in itself , settle this ) oint , considerations of similarity of flow would ppear to point to its being the case except for the proportionately reater effect of friction in smaller and rougher pipcs .
In any case , however , this ill not aensibly affect the result when is greater than about , owing to the comparatively small proportional effect of friction for reater vnlues , even in the case of a ) with a throat diameter as small as 1 inches .
Fnrthernlore , comparison with the results of such isolated experiments as are ilal ) shows that this appears to be sensibly the with atios of initial to final area between 1 to 3 and 1 to 9 and with a smaller diameter from inch to 22 inches .
The mean of the results obtained from the -shaped circular pipe are plotted at , and show the loss of head to have a of per cent. as against per cent. in the case of the straight pipe of the same .
The reason for this pccted result probably to ) found partly in the fact that the meau velocity ) is greater in the former pipe , thus giving a greater friction loss , but mainly in fact that towards the outlet of this pipe the of of the ides l up to about , so that the loss per unit of this latter ) ortion the pipe , expressed as a of the kinetic of flow , would ) * Dines , ' .
Soc. Proc vol. 48 ; Stanton , ' .
N. A. ' 1909 .
'Inst .
C. E. Proc vol. 159 , p. 222 ; vol. 169 , 323 ; ineeri , 1902 , p. 664 ; 'Amer .
Soc. .
E. Proc May , 1898 .
.\mdash ; Loss oi Head in Pipes of same Value of and of same Initial and Final Areas .
which the results compared with those from a circular pipe with the value of , and with the same initial and final areas , it appears that the loss is greater with the square pipe .
Since , however , the mean value of for a square pipe is greater than the angle between the centre lines of opposite faoes , a fairer comparison is obtained between pipes of the same initial and final area and of the same length .
Such pipes may be compared by means of the curves of , and from these it appears that loss in the square pipe is from 20 to 33 per cent. greater than in one of circular section , the relative difference being greater as the length is reduced .
From the form of the curve in fig. 4 it would also appear that the minimum loss in a square pipe is obtained when is approximately To test the effect of a change in the shape , as opposed to the area , of a 1909 .
] Water through Pipes and Passages , etc. cross-section , a pipe 994 inches long , having a uniform sectional area , and changing gradually in shape from a square of 2659 inches side to a rectangle of inches inches , was fitted to the outlet of the square pipe and the pressure difference at its two ends was measured .
Undel these conditions it was found that the loss in this pipe , due to its change of shape , was , or per cent. of the kinetic per in the pipe .
FIG. 5.\mdash ; Loss in Pipes of same Length and of same Initial and Finad 5 .
of Scclion Divcrging \mdash ; The results the experiments on pipes of ular section l one pair sides parallel are shown in and 5 .
As before , it appears that the percentage loss of head is sensibly independent of the velocity , elocities greater than 4 feet per second in the smaller pipc , frunl the minimum loss is obtained when is about thi , s curve it also appears that while values of between loss practically the same as iu a circular pipe with the , with or smaller values of the loss is much reate in the ipes of the same initial the same length , it is seen that for all moderate lengths the loss is with a rectangular than with either a square or circular , the additional eddy loss due to the reatcr d of the one pair of sidcs not being compensated the reduced loss due to the parallclistl 0 the Prof. A. H. Gibson .
On the Flow of second pair .
Two additional rectangular pipes , having the same length and initial and final areas as the straight taper pipe , but having the .
sides curved so as to give respectively in , uniform retardation Const and in uniform change of velocity per unit length of pipe Const were also experimented upon .
From the mean results , plotted at and in , it appears that the loss as compared with that in the straight taper pipe was reduced by per cent. in and by cent. in C. 6 .
Flouin a Circular Pipe jrging Boundaries.\mdash ; It was originally intended to carry out a series of experiments to determine to what extent the convergence of the boundaries of a circular tube affects the critical velocity of flow , but circumstances have caused bhese to be somewhat curtailed , and it has only been possible to examine four pipes , respectively equal to , and .
Sufficient evidence has , however , been obtained to show that the critical velocity in such a pipe is considerably higher than in a parallel pipe of the same mean radius .
The critical velocities were obtained in the usual way , by plotting the logarithmic of loss of head and of velocity at the contracted section , noting the point at which the index of suddenly changes .
As the .
motion at the entrance to the larger pipe was in every case unsteady , owing to the bends and irregularities in the supply pipe , these values mark in every case the lower critical velocity at which eddy motion settles down into non-sinuous or stream-line motion .
The following table gives the values obtained : \mdash ; Values of C. V. .
per second\mdash ; At large section ( 3-inch diam. ) 1.5 1.94 2.44 3.25 point where dinchesthroat ( inch d In each case the temperature of the water in these experiments was F. The lower critical velocity in a parallel pipe the same mean diameter ( inches ) at this temperature is feet per second , *while in a 3-inch and a inch pipe it is respectively and feet .
second , the higher critical velocities being respectively , and feet per second .
Should the ratio of higher to lower critical velocities have the same value in a conical pipe as in a parallel pipe , this would mean that in the case of a -inch jet discharged from a converging nozzle with steady flow in the supply pipe , the critical velocity would have the following values\mdash ; Value of 1909 .
Water through Pipes Passages , etc. 377 and would explain the glassy appearance of the high velocity jet the nozzle of a Pelton wheel .
Further experiments are , vever , needed to settle this point and also to determine how , with the same values of but with different inlet and diameters , the critical velocity depends on the latter factor .
7 .
lnsions .
following are the conclusions to be drawn from the experiments:\mdash ; In a circular pipe with uniformly diverging boundaries , the total loss of head attains its minimum value with a value of equal to about Owing to the comparatively large effect of friction in a pipe having a small value of , the value iving the minimum loss of head will be somewhat less in pipes .
( In pipes of the type used in Venturi meters , experiment shows that this value is about As is increased the loss of head , expressed as a of creases very rapidly from its minimum value of about per cent. to a maximum of about 121 per cent. when , afterwards diminishing to about 102 per cent. as is increased up to .
Owing to the comparatively small value of the friction loss , as opposed to impact loss with values of greater than about the latter values will presumably be the same for pipes of diameters largely from those used in the experiments .
( b ) By making the pipe trumpet shaped so as to make uniform throughout the pipe , the loss is in some cases increased , others diminished .
Only one such circular pipe was used , and here the loss was per cent. , as against per cent. in a straight tnper pipc of the same lengtb and having .
In the case of a rectangular pipe , however , pipes having boundaries curved so as to make respectively Const. and Const. showed that the loss as compared with that in the corresponding straight taper pipe was reduced respectively by and per cent. Further experiments are desirable to determine the precise form of curve giving least loss of head in each case .
( c ) The loss of head in a pipe of square section is greater\mdash ; at the least 20 per cent. greater\mdash ; than in a circular pipe of the same length and same initial and final areas , while the minimum loss is apparently obtained when the angle between opposite faces of the pipe is approximately ( d ) A change in the shape , as opposed to the area , of the cross-section of a pipe , leads to considerable loss of head .
Thus by the section of a pipe from that of a square of inches side to a rectangle inches by inches in a length of inches , a loss of head equal to was experienced .
second pair uniformly diverging , the loss of head is much greater than in a circular pipe having the same length , and the same initial and final areas .
The minimum loss is obtained with a value of in the neighbourhood of The critical velocity of flow in a circular pipe with uniformly conver boundaries is much greater than in a parallel pipe of the same mean diameter .
The critical velocity increases rapidly with the angle of ence , its lower value in the pipes considered ( from 3inches to diameter is inches:\mdash ; inches diameter ) .
at .
here the The critical velocity in a parallel pipe of the same mean diameter at this temperature is foot per second .
In conclusion , the author would acknowled , his indebtedness to the overnment Grant Committee of the Royal Society , by whose grant in aid the provision of the apparatus used in these experiments was made possible .

