]\gt ; Mr. F. E. Rowett . .
in mass , of a constant cross-sectional area , and of a length of cm .
, while an ohm has a resistance of cm .
The agreement is most satisfactory .
The conclusion is that a resistance of 1 International ohm is equal to ohm , the probable error of approximately the sum of those involved in the resistance of the ohm and the International ohm .
It follows that a column of mercury , at C. , cm .
long , of constant cross-sectional area ( the same as that of the International ohm ) , has a resistance of 1 ohm .
The mass of this column will be I to express my thanks to the Drapers ' Company of London and to Sir Andrew Noble for their generous help , to Lord Rayleigh for his keen interest in the work , and to Dr. Glazebrook for his very valuable help advice throughout the investigation .
Elastic Hysteresis in Steel .
By F. E. ROWETT ( Whitworth Scholar ) , Research Student of the University of ( Communicated by Prof. B. inson , F.RS .
Received November 12 , \mdash ; Read December 11 , 1913 .
) rate of dissipation of energy by internal molecular friction , when steel undergoes variations of stress within what is regarded as the elastic limit , was the subject of a recent paper by Hopkinson and Trevor-YVilliam A bar of steel having elastic range of about 25 tons per square inch was subjected to direct ( axial ) push and pull , the limits of tension and compression being equal , and the frequency of the cycles about 120 per second .
The dissipation of by elastic hysteresis determined by the fall of temperature between the middle point of the bar and the ends .
It could be measured accurately in this way when the range of stress was near the elastic lilnit ( 25 tons per square inch ) and could be detected when the range was 10 tons per square inch .
One object of the research just referred to was to ascertain whether the dissipation of energy per cycle of stress increased with the speed of reyersal , in other words whether the internal friction to which this dissipation is * Roy .
Soc. Proc , vol. 87 , p. 502 .
Elastic Hysteresis ?
Steel .
due is of the nature of solid friction , or is more analogous to the viscosity of fluids .
For this purpose the bar was alternately compressed and extended , statically , in a testing machine , between the same limits of stress as in the high-speed tests , and the stress-strain curve was obtained by means of a specially designed extensometer .
The curve is of course a very loop , the area of which corresponds to the dissipation of energy in the slow-speed cycle .
It was not found possible to do more than determine the width of this loop at the origin , corresponding to the difference in len , th3 of the unstressed piece according as the last application of stress had been compression or tension .
The area of the loop is equal to the product of this difference into the of strain , multiplied by a constant which depends on the shape of the loop .
In the .
paper under consideration it was assumed that the loop was of lenticular form , beinero bounded by two arcs of circles of radius , and on that assumption the vsteresis in the static test was at all stress about times that found in the -speed cycle between the same limits of stress .
As pointed out in the paper , the uncertainty as to the shape of the loop precluded any very definite inference as to the effect of speed of reverHal on hysteresis except the difference , il any , was probably in the direction of reduced hysteresis at speeds , and that it was not more than 30 per cent. as between 120 cycles per second and quite slow sals .
The present earch is a further study of this problem .
In order to secure higher accuracy in the measurenlent of the static hysteresis , thinwalled tubes have been used in sion .
By the use of spirit levels in the manner described in greater detail below , it has been found possible to detect shearing strains of the order of 10 corresponding to a stress of .
per square inch , and to determine within about 5 per cent. the energy dissipated by elastic hysteresis in a cycle of stress well -ithin the elastic limit .
The corresponding -speed cycles were observed the method of torsional oscillations , first employed by Lord .
One end of the tube was fixed , and a fly-wheel was mted on the other end , of such inertia as to give a frequency of torsional vibration of about 70 per second , and the amplitude of vibration was recorded phically .
From the rate of decay , after .
for air-friction , the hysteresis is calculated .
Briefly , the result of these experiments is to show that the hysteresis is , robably within 5 per cent. , the same at high speeds as at low speeds .
Furthermore , the shape of the hysteresis loop , which wss left uncertain by Hopkinson and Trevor-Williams ' work , has been determined .
If the constant Mr. F. E. Rowett .
giving the arca of the loop from its principal dimensions , which has been determined by the present author for torsional cycles , be substituted for the conjectural constant used by them , the result is to bring the hysteresis of their static cycles into almost exact reement with that which they found at high speeds .
It also appears that in a hard drawn tube as supplied by the makers the hysteresis is for all stress-ranges only about one-eighth of the hysteresis in the same tube after it has been softened by annealing .
In the annealed tube the loss per cubic centimetre per cycle , with a stress-range in shear of dynes per square centimetre ( tons per square inch ) , is about 11,000 ergs , and it varies approximately as the cube of the stress-range .
Hopkinson and Trevor-Williams found at the corresponding range of stress a loss of about 4000 ergs in a steel bar which had not been annealed , and at higher stresses the loss in their experiments varied as the fourth power of the stress-range .
High-speed Cycles .
The lnaterial was in the form of tubes having a mean external diameter of inch , and a mean thickness , determined by weighing , of inch .
The steel had the following composition :Carbon , per cent. ; manganese , per cent. ; sulphur , per cent. ; phosphorus trace , the remainder iron .
The tubes , as they came from the makers , were hard , as the result of the drawing .
The elastic limit in torsion in this state was tons per square inch shear stress in either direction , iving an elastic range of tons in shear or 55 tons in tension and compression .
By annealing at so C. for fifteen minutes and cooling in the furnace the effect of the drawing was removed .
The tubes were then quite soft and ductile and had a well-marked yield point in torsion of tons per square inch shearing stress , the elasticity as determined by ordinary tests being nearly perfect up to that stress .
The elastic range in this condition is , therefore , roughly tons per square inch in shear or tons per square inch in tension and compression .
The specimen tube AB was firmly fixed at , fig. 1 ( thorough and smooth soldering was found to be efficient ) , into a steel block , which in turn was securely bolted down to a cast-iron frame of very massive proportions .
The upper end of the tube was fixed into a small flywheel , the whole with the tube vertical , as shown in fig. 1 .
Four very small stops were fitted into the flywheel , two utilised for applying a pure torque by means of stirrups and F. passed from the stirrups over pulleys fitted with ball-bearings , and were attached to scale pans and H. The two columns and were fitted with adjustable centres , about which Elastic Hysteresis Steel .
the ring PQ could be rotated .
This ring carried two stiff , fine-threaded screws which aged with the remaining two stops on the fiywheel .
The latter were fitted so that the upper edge of one and the lower edge the other coincided with each other and with the centre of the flywheel .
heights of the adjustable centres were set so that upon the the screws left the stops , the one at the upper and the other at the lower edge , at the same instant .
A mirror was attached to the tube at a definite distance from the fixed end , and records were taken of its motion upon motor-driven photographic films .
In making an experiment , a definite torque was applied to the ) by means of and stirrups , and the screws adjusted so as just to take the stress on turning the ring into position .
This could be conyeniently done by observing the spot 01 ' light , any nlovement of the spot on turning the ; indicating a faulty adjustment .
The weights and stirrups were now removed without the distribution of stress in the tube .
A record was finally obtained by the ring clear of the topb and thus releasing the flywheel , which was thereby set in torsional vibration .
A calibration was taken upon the same film by the same torque and obtaining the deflection in the of a continuous line on the , the ring meanwhile remaining free , so that the calil ) lation was independent of the setting or the adjustments .
It was found that 1 cm .
deflecCion corresponded to a stress of dynes per square centimetre , i.e. about tons per square incl ] .
The effect of air friction was ated by a small steel post Llbove Mr. F. E. Rowett .
the flywheel to which was attached a disc of thin sheet tin .
Films were obtained with the tin disc attached and also without the tin disc for the same specimen , and the increased obtained by difference .
The airfriction was found to be small by comparison when the steel tube used had been annealed .
The t.ubes , however , as supplied were in a hardened state due to the process of manufacture , and in this condition the dissipation of in the specimen was found to be considerably less than in the annealed state .
Advantage was taken of this fact to determine the energy lost by air friction .
The friction loss is due almost entirely to the vibration of the flywheel , the motion of the tube very small indeed .
The flywheel was a plain smoothly turned disc , and the friction loss psr cycle was assumed to be the same as of a disc of the same diameter of tin plate with a rim attached , the width of the rim equal to half the thickness of the flywheel .
The disc was attached at a distance ciently away from the flywheel to avoid any interaction , and the two had exactly the same motion .
The weight of the was so small in comparison to the flywheel that the frequency of vibrations was altered by only a very small amount ( about per cent The damping curves obtained are shown in fig. 2 .
The amount FIG. 2 .
of energy stored in the flywheel and tube per cubic centimetre of steel is equal to the ( stress-range ) ergs , where is the modulus of rigidity , the stressand being both expressed in dynes per squale centimetre .
Let amplitude , which is proportional to the of stress , and Hysteresis in Steel .
suppose the energy lost per cycle .
Then the loss of energy per second is given by where , number of cycles per second and A and are constants ; the inteo of which is .
where is the amplitude when If , corresponding to the law fiiction velocity2 , the law of variation of amplitude is iven by , where constant , that is the inverse of the stress-range plotted ainst the nmnber of cycles performed should yive a line .
The latter has been done for curves above and the results are otted in fig. 2 .
lines we ] obtained both for free ) also when the additional attached .
The equations of the lines were found to be I. vibratiou II .
Additional air friction where stress range in dynes per square centimetre , and nunlbe of cycles } ) erfornled .
The energy present per of metal modulus of rigidity .
From Equation I The energy lost per cycle per cubic centimetre dynes per squal.e celltinetre .
per inclI .
Therefore from I the energy lost per cubic centinnetre per cycle is and from II in per cubic centimetre .
cycle lost iction is therefore and by hysteresis , , Mr. F. E. Rowett .
being in dynes per square centimetre .
If be expressed in tons per square inch , Hysteresis loss ergs per cubic centimetre per cycle .
Air frictional loss , Fig. 3 shows the results plotted .
The frequency of the vibrations in the experiments was 67 cycles per second .
67 cycles per second .
FIG. 3 .
Experiments Annealed Some of the tubes employed were annealed by gradually heating them throughout their length to a temperature of in an electric furnace , maintaining this temperature for about 15 minutes , and then allowing them to cool in the furnace .
The cooling took about 4 hours .
For the same stress-range the damping was found to lave considerably increased and Table I gives the dissipation in ergs per cycle at the various stress-ranges , while fig. 4 shows results plotted .
The length of the tube employed was the same as in the previous experiments and the frequency was found to be the same .
Column 7 gives the ca]culated dissipation due to air friction and the hysteresis loss is obtained by difference .
It will be seen from fig. 4 that at a stress-range of ' dynes per square cm .
( tons per square inch ) hysteresis loss per cubic centi- etre per cycle is 11,000 ergs , whereas the air friction loss is 740 ergs or about per cent. Elastic Hysteresis in Steel .
Table I.\mdash ; Experiments on Annealed Tubes .
Annealed at 80 C. Cooled in Furnace .
Vibrations 67 per sec. tion .
ress range\mdash ; Tons per .
in .
FIG. 4 .
Order of The amplitudes of the vibrations upon the ilnlS could be measured to 1/ 5 mm. , so that , at the maximum range of ) cm .
, the possible error did not exceed 1/ 250 in the amplitude or 1/ 125 in the energy .
This amounts to 1600 } per cubic centimetre .
The dissipation is calculated by obtaining the present every two complete cycles , taking the difference and dividing by two .
The difference at the higher ranges amounts to about 20,000 ergs , so that Mr. F. E. Rowett .
the possible error in the measurement of dissipated in any one cycle does not exceed 8 per cent. In the fair curve drawn through the plotted points for this quantity in terms of the number of cycles completed , the accidental errors of observation disappear in large measure and the error at the higher stresses probably does not exceed 4 per cent. Static Experim Static tests were performed upon the tubes used in the dynamical experiments , with the object of finding the actual shapes and areas of the hysteresis loops for various stress-ranges when these were performed slowly by hand .
The method employed was to submit the tube to cyclic stress-ranges performed in the direction of the arrows , fig. 5 .
After a number of cycles the process stopped at a point and the strain measured .
The cycle was then continued until the point was reached having exactly the same stress but being in the opposite portion of the ] The strain was again measured and the difference between the then value and the previous value for the same stress gave the length By varying the position of the points and and meanwhile keeping the stress-range constant the exact shape of the loop for this range was obtained .
The difficulty iu such experiments of eliminating completely the effects of friction and ysteresis in the apparatus which is used for measuring stress Elastic Hysteresis in Steel .
was overcome in this case by fixing the tube experiment at one end to a much longer and stronger hard drawn tube , which the torque applied , and by the twist of which measured .
The dimensions of the tube were such that the stress in it was about of that in the small tube .
Since the hysteresis stress difference measured varies approximately as the square of the , the value of the stress difference in the tube was about 2 ) cent. of its value in the small tube .
The hysteresis in the tube was renloved by small torque until the actual twist in the large tnbe was .
This small torque , however , removed some of the twist due to hysteleHis in the small tube , and the hysteresis observed in the small tube tYilb thus smaller than the actual value by 14 per cent. when both tubes were in the hard drawn condition .
When the small tube was annealed and the other hard drawn the error was about per cent. The tifl .
tube was fixed at one end and the torque was nitted to it through the experimental tnbc .
The of the tubes such that the ular twist in each was tely the same .
eneral view and details of the appal'atus .
compound tube GH was rigidly fixed at while the end was keyed to lever between adjustable stops and F. kSensitive micrometer levels , , were fixed to the ) at the positions , P ) , , and , and at and small plane mirrors , , were so that reflectio1ls from scales and were observed in telescopes and The weights exactly balance the weight of lever L. The mirrors and , in conjunction with the telescopes and scales , supply the means of measuring the extreme ranges of strain , and thus of Htress .
In making an ) el.iment the desired torque applied to and by means of the micronleter screws all four levels are adjusted so that the bubbles read zero .
The desired cyclc of stress is then applied by movement of lever , which is returned ) oximately to its position , and there adjusted by means of screw until the levels A read exactly zero .
If the hysteresis in the large tnbe is neglected the torque is now ) the same as before .
The levels and will now no read zero , and the diffel'ence between them represents the hysteresis corresponding to the cycle of stress which been erformed ( i.e. length in .
) . .
F. E. Rowett .
Fi VItb ) Fi FIG. 6 .
Shape of the Loop .
To obtain a comparison between the hysteresis at high frequency and that obtained by static experiments , it is necessary to know the area of the loop , that is to obtain the relation between the maximum and mean widths .
This has been done in the following manner : definite stress-range was first decided on , and the stops and adjusted as required , the range being calculated as before from the mirror readings at the extreme positions of lever L. After a few hundred cycles had been performed at this range , in order to Elastic IIysteresis ?
Steel .
get the material into a cyclic state , the process was stopped at a definite point in the cycle and , the stress rined by means of screw P. All the levels were now adjusted to read zero and the of the telescopes lloted .
From the latter and a knowledge of the for zero stress , the position in the cycle was ascertained .
The cycle was now continued to one extreme position and the leyer loack and screw adjusted until levels A and vere again .
This occurs hysteresis in the tube ) when the stress in the tube , and therefore in the small tube , is exactly the same as before .
The levels and vere no found to have other than zero , that the strain in the tube is htly different from what it was before .
Table II yives the results for a cycle a range of tons per square inch tons .
The last but one ives the ence of twist observed at various in the cycle , according as the stress is or decreasing .
The last ivesD the Corr S stress and is obtained by multiplying the strain in the colnmn the modulus C. These results are shown graphically , the figure representing the hysteresis loop being nified 4 times in the direction of the stress axis .
It should be noted that in setting out the it is assumed that the observed stress differe1lce , such as , fig. 7 , is bisected the line PQ vlJich corresponds to pel .
fact elasticity .
lmption is , of course , not necessarily true , and to this extent the real shape of the loop is uncertain , ) it does not affect the area of the loop , which it is ) nlain object of these experiments to determine .
The mean width of the loop is of the maximum width , and the area in of the cycle is to be obtained by the maximum width on the stress axis by the of strain and the constant kinson and Trevol-Williams assumed this constant to be 2/ 3 , and the dissipation for high-speed cycles then appeared to be times that observed in the statical ) eriments .
The use of the factor in reducing the results of the latter brings the two into close agreement .
Another experiment at a range of 6 tons per square inch gave the same ratio of maximum to mean width within 1 per cent. At lower the accuracy is not sufficient to determine the shape of the 1001 ) but it may be assumed that it is substantially independent of the stress-range , and that the area may be calculated from the width of the loop zero stress by the use of this same constant .
This width was determined a nnmbe of stress-ranges .
The procedure was exactly the same as described above , the only points determined in each case , however , being those correspond- ing to zero or approximately zero stress .
The results are plotted in fig. 8 .
The corrcsponding loop areas calculated 540 .
F. E. Rowett .
from the maximum widths by the use of the constant gave when plotted the curve shown .
On the same figure are shown the values of the energy dissipated per cycle at 67 cycles per second , as determined with the dynamic machine .
It will be seen that the two agree very closely .
Table II.\mdash ; Experiments on Annealed Tubes .
Shape of Hysteresis Loop .
tons per .
in .
Equivalent stress .
per sq .
in .
3.56 3.05 2.595 1.303 Nil 0.398 0.874 1.293 1.77 2.19 3.10 8.58 Table III.\mdash ; Experiments on Annealed Tubes .
Annealed at C. and Cooled in Settinmir.rorreadings Angle odianstwist , differ cyclstress .
Elastic Hysteresis in Steel .
5 41 An unaI)nealed tube was introduced into the static and reading taken for the width at the mid-point of hysteresis loop .
It was found as be expected from the -speed experiments thnt the hysteresis was very small except at high ranges of stress .
The value is given as illustration:\mdash ; Stress-range , tons per square inch dynes per centimetre ) .
Width of loop at centre , per square inch .
of loop , 1820 At the same stress-range the dissipation found in the high-speed machinle was 1650 ergs see .
difference between this result and obtained statically is within the of VOL. LXXXIX.\mdash ; A. Mr. F. E. Rowett .
FIG. 8 .
of Accuracy in the The levels used were placed 10 inches apart upon the experimental tube , and when in position the levels were calibrated against rotation of micrometer heads .
The divisions on the levels were about inch long and readings of these could be easily estimated to th of a division .
One diyision of the levels required , taking the mean value of levels and of a thousandth on the micrometer head .
Thus division in .
The micrometer head worked upon a radius of inches , so that the angle of twist could be accurately estimated to within radians .
The angle of shear could be thus estimated to radians .
Hysteresis Steel .
Assuming .
per square inch , the hysteresis stress difference could be calculated to withill .
per square inch .
For a stress range of tons per square inch the hysteresis stress difference is ton per square inch ( 244 lb. per square inch ) so that by means of the apparatus described above the width of the loop at the mid-point for this can be estimated correct to within per cent. of itself .
The above work was carried out at the ineering Laboratory , and I wish to express my thanks to Prof. B. Hopkinson for his kind help and inspiring interest .

