]\gt ; Distribution of Velocity in a Viscous Fluid .
the value is more accurate , this being the of the values obtained in the most satisfactory experiments .
The experiments were all carried out in the neeri Laboratory of Cambridge University , and the author is indebted to Professor Ewing , F.RS .
, for much kindly advice and ement given during the course of the research .
On the Distribution of Velocity in Viscous Fluid orer the Cross-Section of a Pipe , on the Action at the Critical Velocity .
By JOHN MORROW , M.Sc .
( Vict .
) , M.Eng .
( Liverpool ) , in ineering , University College , Bristol .
( Communicated by Professor H. S. Hels-Shaw , F.R.S. Received March 3 , \mdash ; Read bIarch 30 , 1905 .
) 1 .
Introduction.\mdash ; An expression for the velocity at which stream-line motion breaks down in cylindrical pipes has been obtained by Osborne Iieynolds , ether with many others he has measured the fall of pressure occurring in different lengths .
Lord Kelvin , Lord and BeynoldsS have investigated the stability of different kinds of flow for viscous and non-viscous fluids , but how nearly the theoretical conditions of velocity distribution actually occur has not hitherto formed the subject of research .
Experiments have often been made on the variation of mean linear velocities in the state , but when the motion is irrotational , these are too low to admit of accurate measurement by any method so far employed .
The objects of the present paper are to determine these velocities , to investigate the change which takes place at the critical velocity , and to find the relation between the velocity and the pressure to which it ) ives rise in \ldquo ; Pitot\ldquo ; gauge of the form used .
The research has become possible chiefly through the aid of an exceedingly delicate pressure gauge recently designed by my Professor A. P. Chattock and Mr. J. D. Fry .
I am indebted to both these entlemen for much valuable advice .
' Phil. Trans 1883 , Part III , p. 935 .
'Phil .
Mag August , 1887 .
Phil. lIag July , 1892 , etc. S 'Phil .
Trans , 1895 , vol. 186 , p. 123 .
206 Mr. J. Morrow .
On the Distribution of Velocity [ Mar. 3 , : Osborne Reynolds has shown that in parallel channels there are two limits to stream-line flow .
The lower one is the velocity at which motion , inally eddying , will , when left to itself , settle down into stream-line motion , and the upper limit is that at which the motion of a fluid , originally at rest , traversing a pipe in au unstable stream-line condition , suddenly breaks down into the eddying or turbulent state .
The variation of the lower limit with change of temperature has been further investigated by Coker and Clement , similarly that of the upper limit by Barnes and Coker .
The present paper deals with the lower limit of stream-line flow , and the term " " critical velocity\ldquo ; is used in that sense throughout .
2 .
Description of Apparatus.\mdash ; In an earlier apparatus the pipe was about cm .
in diameter and the Pitot tube 20 cm .
long .
There was danger of the Pitot tube bendin , the velocity distribution was not symmetrical , and the temperature effects were considerable .
In the new apparatus the pipe was of glass , cm .
in diameter .
It was with a bell-mouth in the entrance chamber , so that the water should enter under circumstances conducing to steadiness of flow , and by means of a fine tube a stream of colour could be admitted to indicate the nature of the motion .
The water was from the hydraulic installation at University College , Bristol , and there was no other discharge from the mains while the experiments were in progress .
The Pitot and static head tubes are shown in fig. 1 , being a brass pipe in continuation of A. To insure accuracy in the motion of the Pitot tube , , it was carried by a rigid ular framework composed of a tube rod , and two cross-pieces F. The uides in which this frame slides are two short tubes and and the longer one , and to these are attached the cross-piece and scale S. Thus , by turning the screw , the Pitot may be traversed across pipe , its exact position being indicated by the reading on the scale opposite an index on F. The static head tube is at , a hole being drilled in the side of the glass pipe .
A connection to would , therefore , give the pressure due to the velocity and static head combined , whilst one to would the static head only .
The two passages from the Pitot tube pass to one end of a reversing device , the other end of which is in communication with the static head tube .
The two sides of this circuit are connected to the limbs of the pressure 'Phil .
Trans , vol. 201 , pp. 45\mdash ; 61 .
'Roy .
Soc. Proc 1905 .
] in viscous fluid over the Cross-Section of a Pipe .
207 gauge .
By this means ( 1 ) the static head can be brought to one side of the gauge and the Pitot tube to the other ; ( 2 ) the arrangement may be reversed ; or ( 3 ) the gauge may be short circuited on itself .
It was foumd useful to be able to damp the slight oscillations occurring FIQ .
1 .
at some velocities in the pressure gauge .
For this purpose a small air chaml ) and two constricted passages were provided .
They could be cut out of the circuit , and were only used for verifying previous observations .
is a general view .
It will be seen that the discharge took place Mr. J. Morrow .
On the Distribution of Velocity into a tank at the end of the pipe , and from there the water passed to waste by an overflow .
The whole apparatus is carried on a stout timber base , the pipe and its attachments at each end being very carefully centred so as to ensure a maximum and symmetrical flow .
3 .
Measurement of pressure , discharge !
, and \mdash ; The pressure gauge is of the Chattock-Fry differential type , and in construction is very similar to that in use at the National Physical Laboratory .
* As employed in this research , however , it depended for its action on the difference in density of two liquids .
From fig. 3 it will be seen to consist essentially of a -tube : of which bb are the arms and aa the horizontal connection .
The two parts of aa meet in the chamber , and the fluid pressure is applied at .
The tubes aa and the lower portions of bb contain carbon-tetra-chloride , the upper portions of bb and the chamber are filled with water .
A menisous surface of separation is shown at , , and the manipulation of the gauge consists in keeping the surface , viewed the microscope , absolutely unaltered in position .
This is done by turning the graduated disc of the screw , and so the the small angle necessitated by the variation of pressure diffel.ence .
There is in this way no motion of the liquids relatively to the glass tubes which contain them , and hence errors due to surface-tension , refraction , or viscosity are entirely avoided .
The pitch of the screw was 1 mm. , and on the circumference of the disc were 200 divisions .
These could be further divided by a vernier , but its use was abandoned as being within the limit of sensitiveness of the meniscus surface and miscroscope .
* See Stanton , ' Proc. Inst. C. vol. 156 . ?
05 .
] in Viscous Fluid over the Cross-Section of a Pipe .
209 A and the dimensions shown in fig. 3 , and the reading of the screw , the pressure difference is given by FIG. 3 .
was found to be and ) was taken from measurement on a Westphalia Balance to be at C. To obtain the mean velocity of the water , it was necessary to measure the discharge .
This was done by collecting the water passing through the apparatus in a given time , and weighing it on a carefully-calibrated spring balance to the nearest 10 grammes .
The temperature was obtained by a thermometer placed in the end tank close to the outlet of the pipe .
4 .
Method of making the Experiments.\mdash ; In the velocity distribution , experiments were made with the Pitot tube in various positions , commencing near one side of the pipe and at successive points across the diameter , the water running continuously while a complete series of observations was made .
The temperature was noted , and the discharge taken as the mean of six independent measurements .
The following was the general system adopted for reading the pressures:\mdash ; For any position of the Pitot tube , connection with the pressure gauge was made by opening the taps , the was adjusted , and the reading taken .
The connections were then reversed , and the reading taken .
To elirninate errors due to friction or surface-tension , the observation was then repeated by reading , reversing the connections , and then re-opening S. This was con- tinued until there was no apparent motion of the meniscus surface .
The difference between the two readings was then double that due to the velocity of the stream at the point .
A check on the accuracy of the gauge applied during each series of readings .
The strument being adjusted when shortcircuited , its zero reading was taken .
The pressure due to the velocity was then read on each side of this zero .
In the earlier experiments errors were discovered in this way , and found to be due to a very slight eccentricity FIG. of the point of the screw .
Tests were also made for displacement of the surface when no water was flowing , and occasionally a very small correction had to be allowed .
This was attributed to slight differences of temperature in the passages leading to the gauge . .
Determination of the Pitot Constant.\mdash ; If a tube be placed in a stream , as shown in fig. 4 , and it be assumed , as in the Bernoulli Theorem , that the flow is steady , and that the velocities of consecutive stream-lines are not widely differe1lt , we may treat the fluid immediately at the mouth of the tube as at rest , and apply the theorem as though unimpaired by viscosity .
The pressure in the stationary fluid is then given by for this is the pressure in the adjoining stream-tube , where the velocity is indefinitely small .
The formula requires experimental vel'ification .
Since the pressure must be proportional to we may in all cases assume If is the height above that due to the static head , to which the fluid will rise , the velocity of the stream approaching is where is the Pitot constant to be determined .
This may be done by a comparison of the gauge with the discharge from the pipe .
The total flux is 1905 .
] in a Viscous Ftuid over the Cross-Section of Pipe .
211 where is the velocity distant from the centre , and the radius of the pipe .
And , since where is the gauge reading and a constant , we may write To determine the , curves were plotted having as abscissa and as ordinates .
The areas enclosed by the curves were then ated by the method of ordinates , and the values obtained used for finding , and hence the Pitot constant .
An example is given in Table I , in which readings were taken at 19 different points in the diameter of the pipe .
Table I. The second line in the table contains the square roots of the means of corresponding readings on each side of the centre .
In fig. 5 the curve shows the relative flux through different elemental tubes , into which we may suppose the stream to be divided .
Its area was obtained from the mean of 83 ordinates .
The curve is proportional to that of velocity distribution .
The mean value of obtained from the six experiments described in the next section is , and in calculating velocities I have adopted this constant .
The number may be compared with those recently found for air , namely , , by Dr. T. E. Stanton*and by Mr. R. Threlfall , F.R.S. 6 .
Velocity Curves for Different Rates ofDischarge.\mdash ; Experiments were made at six rates of discharge , the curve of distribution of velocity being obtained in each case , and the result checked by the mean velocity from the Pitot readings with that obtained from the actual measurement of the discharge .
The critical velocity , as determined by the point at which the law of velocity distribution changed , occurred when the mean was about 4 per second ( this being , of course , the velocity at which stream-line motion ceased to be the stable flow ) , and above this there was a decided change in the way in which the velocities increased with an increasing discharge .
'Proc .
Inst. C.E. , ' vol. 166 .
' Proc. Inst. Mech. Eng 1904 .
Mr. J. Morrow .
On the Distribution of Velocity The observed readings were reduced to velocities as already described , and these are given in Tables II and III .
The curves were plotted , and the values of determined .
FIG. 5 .
The temperature varied between and C. Table II.\mdash ; Experiments below Critical Velocity .
1905 .
] in a Viscous Fluid over the Cross-Section of a Pipe .
Table III.\mdash ; Experiments above Critical Velocity .
The curves in fig. 6 are obtained by plotting the numbers in the last two tables .
They show , as ordinates , the actual velocity of the stream at FIG. 6 .
VOL. LXXVI.\mdash ; A. Mr. J. Morrow .
On the of Velocity different distances from the centre of the pipe , the figures on the curves denoting the mean velocity as calculated from the discharge .
It will be seen that below the critical velocity , the curve approximates to a parabola , and that afterwards there is a gradual change in form in the direction of a uniform velocity in the greater part of the pipe with a sudden fall to zero at t'ne walls .
These results are contrary to the frequent statement that , when the conditions of flow change from stream-line to eddying or rotational motion there is an equally sudden change in the law of distribu- tion of linear velocity from the parabolic to nearly uniform velocity over the cross-section .
There appears to be no slipping at the walls of the tube .
The nature of the change occurring at the critical velocity is exhibited in fig. 7 , in which abscissae are mean and ordinates corresponding actual velocities at the different radii marked on the curves .
From these it appears that , below the critical , the velocity at any fixed point is proportional to the total flux , and may be expressed approximately by the formula being the distance from the axis of the pipe , , the mean velocity , A the ratio of maximum to mean , and and constants .
In this case 1905 .
] in a Viscous Fluid over the Cross-Section of a Pipe .
215 According to the mathematical theory of viscosity*the surface graphically representing distribution of velocity is a paraboloid , and the equation is of the above form , but having the constants The two cases are , however , not strictly comparable for , whilst the theory takes no account of terminal conditions , in the experiments these are of importance .
There would probably be accelerations and retardations over the cross-section , tending to equalise the velocities and the tubes of flow would not then be strictly parallel to the axis .
It is , I think , evident that the variation of the curves from the parabola is in part due to these causes , and in part due to the fact that any resistance or obstruction across the pipe also tends to equalise the velocities .
Beyond the critical velocity and within the very narrow limits of these experiments the curves in fig. 7 are approximately parallel when not near the sides of the pipe .
At these low velocities , too , the distribution is represented roughly by the equation In practical caiculations in Hydraulics it is often assumed that the tangential resistance per unit area in a pipe is where is a numerical constant depending on the nature of the surface .
Accepting this as an expression of observed facts , Professor Lamb deduces an conclusion .
If denote the mean velocity at any point , we have at the surface Now let be the radial distance between two surfaces moving with relative velocity in the regular " " laminar\ldquo ; flow which would give the same tangential stress , and we find In the experiments , the last is the only one to which this formula can be applied .
Thus taking , and , we have cm .
The point at which the velocity is equal to the mean is found from fig. 6 to be almost exactly 0.5 cm .
, an agreement which is remarkably close for an approximate method .
Summary and Conclusion.\mdash ; The results may be briefly summarised as follows : * Lamb 's ' Hydrodynamics , ' p. 621 .
Ibid. , p. 574 .
Distribution of Vlocity in Viscous Fluid .
1 .
The experiments provide a partial confirmation of the theoretically obtained law of velocity distribution , but show that this distribution can only be obtained under very special conditions , of which absolute freedom from obstructions and end effects are important ; and hence 2 .
When the flow is direct and stream-lines exist , the velocity distribution is not necessarily exactly that ich may be described as characteristic of " " normal\ldquo ; flow .
3 .
At the critical velocity the irrotational straight line motion ceases and is followed by one in which the paths of the particles of fluid are eddying and turbulent .
The law of distribution of mean linear velocity parallel to the axis simultaneously changes from the parabolic ( or approximately parabolic ) to that typical of eddying motion .
4 .
The critical velocity in question ( being that at which eddying motion ceases to be transformed into direct motion , and not that at which a highly unstable stream-line motion is suddenly disturbed ) , is not accompanied by a sudden change in the velocity parallel to the axis at any point in the cross- section .
On the other hand , as the total flux increases , the experiments show a gradunl transition from one state to the other , due to the change which has occurred in the law of velocity distribution .
5 .
The observations have little bearing on the upper limit of stream-line flow , as observed by colour hands .
They indicate , however , that the unstable direct motion would follow an approximately parabolic law of velocity distribution ( as represented by the equation obtained for stream-line motion ) and that at the higher critical velocity this distribution would suddenly change to that represented by the equation iven for eddying motion .
In this case then , instead of a gradual } of velocity , there would actually be sudden and large changes in the velocity parallel to the axis at different points in the cross-section of the pipe .
6 .
The " " Pitot Law\ldquo ; , is at least approximately true at exceedingly low velocities .
In conclusion I must mention my indebtedness to Professor R. M. Ferrier for his uniform kindness during the ress of the work , and for allowing portions of the apparatus to be constructed in the workshops of the College .

