]\gt ; The Mechanical Viscosity of Fluids .
By T. E. STANTON , D.Sc .
, M.Inst .
C.E. ( Communicated by R. T. Glazebrook , F.R.S. Received May 16 , \mdash ; Read June 29 , 1911 .
) ( the National Physical Laboratory .
) The experiments described in the following paper were undertaken in order to yate the relation between shearing stress and rate of distortion in fluids which are in eddying or sinuous motion , that is , motion in which the frictional resistance , at the boundaries of the solid over which they move , varies approximately as the square of the relative velocity , as distinguished from that steady or laminar motion in which the frictional resistance is proportional to the first power of the relative speed .
This shearing stress has been called by Osborne Reynolds* ' ' mechanical viscosity i.e. , " " viscosity arising from the molar motion of the fluid and which is not a property of the fluid independent of its motion as is its physical viscosity Thus , to quote Reynolds ' statement , in the eddying motion of a fluid in a parallel channel , " " although the meaIl motion at any point taken over a sufficient time is parallel to the axis of the pipe , it is made up of a succession of motions crossing the pipe in different directions In this case , the shearin stress at this point on a cylindrical surface coaxial with the pipe will include the momentum per second parallel to the pipe carried by the cross streams across the surface on which this shearing stress is measured On the other hand , ' the coefficient of physical viscosity is the coefficient of instantaneous resistance to distortion at a point with the fluid As far as the author is aware , no attempts have yet been made to measure the shearing stress in fluids which are in eddying motion , but Osborne Reynolds has assumed that this stress is proportional to the mean speed of flow , and is also a function of the dimensions of the channel in which the motion takes place .
For experimental purposes , it appeared obvious that the conditions under which the phenomenon could be most conveniently studied were those of the motion of air or water in parallel pipes of circular cross section , and as all the appliances used for the determination of the friction of air in pipes at the National Physical Laboratory were available , air was the fluid chosen for the " " Theory of Lubrication ' Phil. Trans 1886 , p. 165 .
viscosity of Fluids .
experiments .
Another reason for the choice of air was that in Threlfall's* observations on the distribution of mean axial velocity in air flowing through large pipes this distribution .
at any cross section appeared to be independent of the rate of flow .
Further , although in Threlfall 's experiments the general form of the velocity curve does not appear to have been investigated , yet the observed values of the ladial gradient of velocity some hope of the possibility of an accurate determination of the rate of distortion .
If this was found to be the case , a further determination of the variation of pressure along and across the pipe would furnish the data necessary for the estimation of the ratio of the stress to the rate of distortion .
Thus , if is the average shearing stress on the surface of any cylindrical portion of radius of the fluid coaxial with the pipe through which it flows , and the mean axial velocity of the fluid at this radius ; , writing , the object of the experiments was the determination of which may be called the coefficient of mechanical viscosity , as a function of and the dimensions of the pipe .
The Method of Estabtishing the Air Current and Estimating the Values of and The arrangement of one of the experimental pipes and the air fan used to set up the current is shown in fig. I. The air fan discharges into a horizontal pipe metres in length .
This pipe is connected by a bend to a vertical pipe metres high .
The experimental portion , 61 cm .
long , is at the upper extremity of the vertical pipe .
It was found by experiment that this length of outlet was sufficient to set up the limiting distribution of the velocity across the pipe as long as the diameter did not exceed cm .
The fan was driven by an electric motor , which was provided with a sensitive speed indicator , and during the observations at any given speed of flow the speed of the motor was kept constant by hand regulation of a carbon resistance in the armature circuit of the motor .
The values of were determined from the difference between the pressure in a small Pitot tube facing the current and that in a small orifice in the side of the pipe .
The Pitot tube was of rectangular section , the dimensions at the orifice being:\mdash ; In the direction of the radius of the pipe Perpendicular to the radius of the pipe mm " " Motion of Gases in Pipes ' Inst. Mech. gineers Proc 1904 .
Dr. T. E. Stanton .
[ May 16 , This tube was fixed to a micrometer , and could be moved across a diameter of the pipe .
The pressure difference was measured by a sensitive oil and water gauge , whose indications could be relied upon within an accuracy of mm. of water .
The final calibration of the Pitot tube was made by a comparison of its readings with the measured discharge of the pipe through a gas meter .
The determination of the shearing stress on any cylindrical portion of the fluid of radius was made by a direct measurement of the drop of pressure at the surface the pipe , together with the measurement of the variation of pressure along the radius .
The former was observed by taking the difference in two small tubes let into the sides of the pipe .
-As regards the latter , it has been shown by Osborne Reynolds*that although in steady motion , when the motion is sinuous , whexe 'Nature , ' 1898 , p. 468 .
1911 .
] The Viscosity of Fluids .
is the component of sinuous motion normal to the surface .
To detect such a variation , a small tube of radius mm. , with a hole in its side connected to a sensitive gauge , was moved across a diameter , but no variation of pressure could be found up to a distance of 20 mm. from the walls .
It was therefore assumed in the reductions that the pressure was constant across any section , and that the shearing stress was , therefore , proportional to the radius .
( A ) EXPERIMENTS 0N ARTIFICIALLY ROUGHENED PIPES .
As in ordinary smooth pipes it is known that the frictional resistance varies as a lower power of the velocity than the square , and that in consequence , as shown by Lord Rayleigh , friction is given by , where kinematical coefficient of viscosity , linear dimension of the pipe , and the velocity at the axis of the pipe .
In order to simplify the problem , an attempt was made to eliminate any effect due to a variation in by artificially roughening the pipes used until the friction was proportional to the square of the velocity .
This was done a double screw-thread on the inside surface of the pipes , and so the pitch and depth that the surfaces of pipes of different diameters were geometrically similar .
After several trials , two pipes were produced , of diameters and cm .
, in which the friction varied as the square of the velocity , and consequently the friction per unit area at the same velocities was found to be the same for each pipe , the numerical value being dynes per square centimetre .
The results of a large number of expeI'iments on these two pipes were as follows:\mdash ; Effect oVelocity of Flow on the Radial Distribution of Velocity .
To test the constancy of the velocity distribution in a radial direction , two sets of observations were made at the hest and lowest centre filament velocities which could be satisfactorily obtained .
These were 2220 and 546 cm .
per second respectively .
The two sets of reduced velocities for the cm .
diameter pipe are plotted in fig. II , and it will be seen that for this range of speed there is no definite change in the form of the curve , which , from the centre up to a distance of about 3 mm. from the wall of the pipe , is a parabola with its vertex in the axis of the pipe , the equation to which may be written Report of Advisory Committee for Aeronautics , ' 1909-10 , p. 38 ; also ' Phil. Mag fkic Dr. T. E. Stanton .
[ May 16 , Table I.\mdash ; Reduced Values of the .From the oregoing results the follow pipes are derived:\mdash ; ( a ) In such a pipe , through which critical value , it follows from the pa mechanical viscosity , as defined above , within a relatively small distance from 1 ( b ) Considering the variations in th surface coaxial with the pipe due to evident that , since the value of is centre filament , and the shearing force therefore is directly proportional to ] ( c ) In two pipes in which the velocity diameters the values of on two whose linear dimensions are proportion It follows , therefore , that , since the re squares of the corresponding velocities different pipes , be directly proportional The equation to the velocity curve being is known to be proportional to , it foUows tha The Viscosity Fluids .
It appears , therefore , that the expression for the coefficient of mechanical viscosity in the artificially roughened pipes considered may be written where is the velocity of flow at the axis of the pipe , the linear dimension of the pipe , and is a constant depending on the extent of the eddying motion , i.e. , the roughness of the surface .
( B ) EXPERIMENTS ON OBDINARY SMOOTH PIPES .
For these experiments two smooth brass pipes of approximately the same diameters as the artificially roughened ones were used .
( These were and cm .
diameter .
) As in the case of the rough pipes , two sets of observations , . .
VOL. LXXXV.\mdash ; A. Dr. T. E. Stanton .
[ May 16 , one at the highest speed of flow obtainable and the other at the lowest , were made on one of the pipes , and the distributions of axial velocity in the cases are shown in fig. .
It will be seen that , for a region extending the axis up to a value of the radius of approximately , the distributions are identical , but that beyond this radius the curves separate , indicating apparently a regiotl of viscous flow near the walls , in which the slope of the curves necessarily increases at a greater rate than the centre filament velocity , since the resistance varies as a of the speed greater than unity .
It may be mentioned that in the earlier experiments the corresponding curves plotted from the observations were not anywhere coincident , except at the axis , but , as a possible explanation of this was the insufficient length of the pipe between the fan and the portion under observation , further experiments were made with an increased length of pipe , and it was ultimately found that with a total length of 9 metres there was practical coincidence within the region mentioned .
It is of course possible that even then the coincidence was not exact , but the difference , if any , was too small to be detected by the .
It evident , therefore , that in the case of smooth pipes the region of viscous flow was of considerably greater area than in the case of the roughened pipes , in which it } ) eared to be nearly destroyed .
It was , however , found that the central portion of the velocity distribution curve in smooth pipes up to a radius of was parabolic in form , as in case of the roughened pipes , thus indicating that the value of was constant up to this limit .
As regards the effect of the size of the pipe , a precisely similar tion to that used for the hened pipes showed that , for two pipes in which the centre filament velocities were proportional to the respective diameters , the velocity distributions were strictly similar throughout the central portion of the flow referred to above .
* , therefore , this region , since the values of for arly given value of the radius are proportional to the velocity of flow , and since the friction per unit area at this radius is known to be proportional to , it is evident that in the case of the smooth pipe is proportional to .
Further , since for two smooth pipes in which the centre filament velocities are proportional to the respective diameters the values of at .
points in the central region are the it follows that is proportional to the external diameter of pipe , and the * The two velocity curves for this case have not been plotted as in case of the rough pipes ( fig. III ) , because the difference in the radii is too small to bring out their divergence in the region of viscous flow .
The values of the velocitieH determined are given in Table II .
1911 .
] The Viscosity of f'tuids .
complete expression for the mechanical viscosity for the region in question is .
where is a constant .
This is , of course , in agreement yith the value found for the roughened pipes , as in the latter case was a constant quantity for all values of Table II.\mdash ; Reduced Values of the Velocities in the two Smooth Pipes for Experiments in which is constant ( columns ( 1 ) and ( 2 ) ) and is constant ( colunns ( 1 ) and ( 3 ) ) .
Radius .
ecVelocity i Badius .
It appears , therefore , that the yeneral character of the flow of air in pipes consists of a central region in which the ratio of the shearing stress to the rate of change of distortion is constant for any particular value of the flow , and that in the neighbourhood of the walls this alters rapidly until it becomes equal to the physical viscosity at the boundary .
In such a composite state of motion as this , it is clear that exact similarity over the whole section of the motions of air in two smooth pipes of different diameters can only exist when the values of both and are the same for each .
From the value of here determined this condition can only be satisfied if the values of and are the same for the two pipes .
To illustrate this similarity , experiments were made on the distribution of velocity in the two smooth pipes used for this work and in the filament velocities were respectively 1625 cm .
per second in the cm .
pipe and 1017 cm .
per second in the cm .
pipe .
The expsrimental results for both pipes are given in Table II and are also plotted in fig. , and it will

