Go to National Library of New Zealand Te Puna Mātauranga o Aotearoa
Volume 38, 1905
This text is also available in PDF
(841 KB) Opens in new window
– 515 –

Art. LX.—The Resistance of Steel to Mechanical Shock, and the Determination of Material suitable for Machinery.

[Read before the Philosophical Institute of Canterbury, 2nd August, 1905.]

From the earliest days the worker in steel has applied some test to insure that his labour might not be expended on worthless material. The blacksmith tests the quality of the iron he uses by nicking and breaking over the anvil, being guided to his conclusions by the force of the blows necessary to effect the fracture, and by the appearance of the broken surface. This test, crude as it may appear, is, in the hands of an experienced man, a more reliable indication of the suitability of the material for the manufacture of machine-parts than many of those tests which it has been customary for engineers to specify, and which entail the use of a large and expensive plant. In fact, this test, in a standardised form, is an accurate measure of the relative capacity of metals for resisting “shock.”

The introduction of the systematic testing of the materials of construction is of very recent date, and had its origin in the experiments on the strength of structures so often made by engineers in the early part of the last century, and rendered necessary by the development of important works at a more

– 516 –

rapid rate than the theory of applied statics and dynamics. As the theory developed, cumbersome full-sized experiments were replaced by those on models, when it became evident that if the physical properties of the material were accurately known experimental work would in ordinary cases be unnecessary.

Machines had been made for the proof testing by “pulling” of chain cables, suspension-links, &c., and only trivial modifications in design were required for the production of smaller appliances suitable for testing specimens of the material it was proposed to use. A period of refinement brought the sensitiveness of such machines up to 1 in 20,000, and instruments capable of measuring extensions as small as 1/200000 in. were constructed. With these it was ascertained that on being progressively loaded a bar of iron or steel is practically elastic up to a point of loading known as the “elastic limit,” and imperfectly elastic up to a further point known as the “yield-point”; at this point it suddenly breaks down, extending rapidly without further loading, then hardens, and the extension increases with—but more rapidly than—the load, being partially “elastic,” but for the most part “plastic,” until the point of maximum load is reached, when the material begins to flow, contracting rapidly until it fails at a lower total load than the maximum, but at a greater stress per unit, of the now reduced cross-section.

This behaviour is most readily shown by means of a diagram, similar to that drawn by the autographic recorder with which most testing-machines are equipped, in which the coordinates represent the extension of, and the corresponding load upon, the tested piece.

In such a test the ultimate and elastic strengths of the material are obtained, together with its extension and contraction of area; the latter are measures of its ductility. The test usually required by engineers is of this description, and must satisfy certain specified values.

In quite the early days of tensional testing it was noticed that metal which had given excellent results on test often failed under a comparatively low load in service, and that such failure occurred when the load, or some portion of it, was a live one. It might naturally have been

– 517 –

expected that if the stress engendered by a live load exceeded the elastic strength of the material, failure must eventually result from the accumulated plastic extensions; if, on the other hand, the elastic limit was not reached, that the piece would be safe. Many failures, however, occurred in which by no method of computation could the material be shown to be loaded above its elastic strength.

This apparent anomaly led Wöhler and subsequently Bauschinger and others to investigate the effect of repeated and alternating stress, or, in other words, of “fatigue,” on materials. The investigations have resulted in establishing that the “physical constants” of a bar are the cumulative result of its “life history,” and that its elastic limit may be artificially raised by overstrain or cold rolling, but that such an elastic limit is exceedingly unstable, and may be reduced to a very low value, or even zero, by heating, hammering, or alternation of loading. It can be seen, therefore, that failures by repeated loading within the nominal elastic limit may be explained by the assumption that the limit was an unstable one.

Further, Wöhler and Bauschinger found that with gradually applied repetitory or alternating loads the range of stress which the bar is worked over is the principal factor in its endurance— the smaller the range, the greater the load which can be carried— and that the relationship between the statical breaking-strength (t), the breaking-strength when the load is altogether removed and again reapplied an unlimited number of times (u), and the breaking-strength when the load is completely reversed (to the same magnitude but opposite sense) an unlimited number of times (s), is t:u:s:: 3:2:1.

Many formulæ have been devised to fit the results of these experiments. Gerber showed that if the minimum stresses were plotted as abscissæ, and the corresponding limiting ranges of stress as ordinates, the points fell upon the curve of a parabola.

On the basis that many of Wöhler's experiments gave the ratio of t:u:s:: 3:3/2:1, Launhardt and Weyrauch constructed equations, the former for the limiting ranges of repetitory, and the latter for the limiting ranges of alternating, loading. These formulæ have been very generally used, but on plotting the results given by them it will be found that the curve is non-continuous, there being a change of direction at the one stress zero-point. There are also other anomalies.

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Some years ago the writer discovered that if the more correct values of the relationship of t, u, & s of t: u: s:: 3:2:1 be adopted, the equations of Launhardt and Weyrauch reduce to the common form of a=⅔t(1+½L min.L max.;)

– 518 –

This formula may be always used when the statical breaking-strength (t) of the material is known, and gives, in terms of the ratio of the least (L min.) to the greatest (L max.) load on the piece, the limiting stress (a) per unit of cross-section.

In applying the formula to arrive at the safe working-stress (b), a factor of safety (say, 3) should be adopted. Further, this formula fits intermediate experimental results, and if the minimum stresses be plotted as absissæ and the corresponding ranges of stress as ordinates the points lie upon a parabola.

For some years after the publication of Wöhler's researches it was believed that the problem of the working-strength of iron and steel had been satisfactorily solved; but the frequent failure of material of presumably stable elastic limit within the

– 519 –

ranges of stress prescribed as safe soon raised a feeling amongst engineers that the theory was incomplete, the more so that test specimens cut from close to the plane of fracture usually gave normal elongation, contraction of area, and elastic limit. (The writer has himself tested with such results many samples from fractured gun-mountings, drawhooks, chains, crank-pins, piston-rods, &c.). To account for such failures, the theory was advanced that the fatigue had been localised on planes of such minute thickness that their influence on the physical constants of the specimen was so small as to escape detection in ordinary testing. It was pointed out that the presence on a stressed piece of an initial, scratch, flaw, or tool-nick must lead to a crowding of the stress-lines in its immediate neighbourhood which might easily result in overstrain there; with repeated loading the portion so overstrained must eventually fail, and a further crowding of the stress-lines and a further development of the flaw take place, and in this way progressive fracture would ensue. (See fig. 3 on next page.)

Now, although this may happen, this theory in no way accounts for the fact that material which has given most excellent results under tension test sometimes fails after having been in service for a period far too short for fatigue fractures to develop. In fact, the failure of a number of steel plates by falling a few feet from a crane-sling, after strips cut from them had given normal results in the testing-machine, led to the important discovery by Mr. Seaton that these otherwise unaccountable failures only occur when the piece is subjected to “shock,” and that the capacity of steel to resist shock—or, in other words, its anti-brittleness—appears to be in no way guaranteed by the passing of a satisfactory tensional test.

Pure carbon steel may be considered to be a solid solution of carbon in iron, and its principal properties can be illustrated in no better manner than by the commonly used analogy of the behaviour of solution of common salt in water. Guthrie has shown that by the addition of salt to water its freezing-point is lowered : the larger the percentage of salt up to about 23·5 per cent. the lower the freezing-point; for water containing 23·5 per cent. of salt it is 22° C. If there be less than 23·5 per cent. of salt in solution, on the temperature being lowered ice-crystals will first form; a portion of the water thus crystallizing out, the solution becomes stronger, and a further lowering of the temperature is required before further crystallization can take place. In this way the percentage of salt in the residual solution is increased as the temperature is lowered, until— 22° C. and 23·5 per cent. of salt is reached; at this point the water and salt crystallize out side by side, forming the cryo-

– 520 –

hydrate or eutectic of salt and water, and the result is evidently crystals of pure ice surrounded by the crystals of the cryohydrate.

Picture icon

Fig 3

On the other hand, if the percentage of salt originally in solution were greater than 23·5 per cent., on the temperature being lowered salt-crystals would first form, and the solution being thereby weakened, a further lowering of the temperature would again be required for a further deposit of salt-crystals to occur, and this action would continue until−22° C. and

– 521 –

23·5 per cent. of salt was reached, when, as before, the salt and ice crystals would separate out side by side, forming the intimate mixture, the cryohydrate or eutectic; but in this case we should have crystals of pure salt surrounded by crystals of the eutectic.

At each stage of the solidification there would be a retardation in the fall of the thermometer, which on reaching—22° C. would remain stationary till the whole mass solidified. It is evident that for all mixtures other than 23·5 per cent. there are at least two distinct retardations—the first, dependent on the strength of the solution, occurring when the crystals of ice or salt commence to separate out; the second, when the formation of the eutectic begins. Where the solution is originally of eutectic proportions (23·5 per cent. of salt) only the latter point of retardation occurs, there being no crystallization until—22° C. is reached. (The matter may be made clear by reference to fig. 4)

Picture icon

Fig 4

Now, the behaviour on cooling of a liquid solution of carbon in iron is very much that of the salt-in-water solution—if there is less than about 4·5 per cent. of carbon present we have first the separating-out of the iron, and then the solidification of the eutectic of iron and carbon. If there is more than 4·5 per cent. carbon present in the liquid solution the graphite first separates out, then when the temperature has fallen sufficiently the eutctic, the solution being by this time of euctetic proportion. If originally of eutectic proportion (4·5 per cent. carbon) there is only one point of retardation and solidification. But there is this great difference between the ice and the iron: the latter, on account of the high temperature at which it solidifies, retains a certain amount of carbon after it becomes solid. Hence there is a solid solution of carbon in iron which introduces fur-

– 522 –

ther changes after the iron has cooled to a temperature at which it is no longer capable of retaining dissolved carbon as such. This explains the existence of somewhat similar curves in the lower portion of fig. 5.

Picture icon

Fig 5

In this solid solution, with changes of temperature actual changes of structure take place, and on cooling crystallizing-out of its constituents occurs, accompanied with characteristic retardation of fall of temperature. The loci for varying proportions of carbon of such retardations, known as the Ar3 point, the Ar2 point, and the Ar1 point, are shown by the lower curves on the figure.

Taking a steel of, say, 0·15 per cent. carbon at a temperature of 1,000° C., such steel at that temperature will be in the form of a solid solution of carbon in iron; allowing it to slowly cool, on its temperature falling to about 850° C. pure iron will separate out in the form of ferrite-crystals—in much the same way that the ice-crystals formed in the water-and-salt solution. This action will go on until the temperature has dropped to about 760° C., when a second point of retardation marks a magnetic change in the iron.

Crystallization out of the ferrite still continues until at a temperature of about 680° eutectic proportions are reached. At this the critical point, Ar1, the carbon in the form of the

– 523 –

carbide Fe3C, separates out alongside the ferrite, forming the eutectic of iron and carbide of iron known as “pearlite.” We thus have as the final condition of our material ferrite-crystals imbedded in a matrix of pearlite which itself consists of an intimate mixture of iron and carbide of iron.

When the steel contains above 0·35 per cent. of carbon the Ar3 and Ar2 points are merged in one.

A steel containing 0·9 per cent. of carbon is of eutectic proportions: there is consequently only one point of retardation—the Ar or carbon point, at 780° C.—and no separating-out of ferrite crystals, the resulting steel being wholly pearlite. If the steel contains more than 9 per cent. of carbon, still following the analogy of the salt-and-water solution, the carbide Fe3C (cementite) first forms, then the pearlite, and the resulting steel consists of cementite in pearlite with no free ferritecrystals.

If a steel, instead of being slowly cooled, is quenched from a temperature the value of which is disputed, but which the writer considers it rational to suppose should be just above the Ar3 point, and hence dependent on the percentage of carbon, no time is given for the changes, indicated by the curves, to occur, and the material is retained in practically the same state as it was in at the quenching-temperature, the carbon remaining in its diffused or hardening condition. Steel so treated is fully hardened. To such quenched steel the name of “martensite” is given; if of eutectic proportions, “hardenite.”

Picture icon

Fig. 6.

The diffusion of carbon in steel has never been better illustrated than by an experiment performed last year by Mr. Stead. Six bars of varying carbon content were made coarsely crystalline by very slowly cooling down in the heart of a ladle of molten

– 524 –

blast-furnace slag. Then all of them were heated at one end to a little above 1,000°C. and kept comparatively cool at the other extremity. The temperatures at equal intervals along the bars were measured by a Chatalier pyrometer. After heating until the temperatures were constant each bar was quenched in cold water, ground bright, polished, and etched, and micro-photographs prepared. Fig. 6 shows the result.

In the case of No. 1 bar (pure iron) there is no apparent change of structure until a temperature of 870° C. is reached. Then reorganization is apparent, the crystals becoming smaller.

No. 2 Bar (0·2 per cent. carbon).—In proportion as the temperature has exceeded 750° C. so the carbon has diffused from the carbide areas (pearlite) into the surrounding ferrite, until at about 1,000° C. the diffusion is complete (martensite). It will be noticed also that the breaking-up of the ferrite-crystals occurs, as before, at the Ar3 point.

No. 3 Bar (0·4 per cent. carbon).—Diffusion is complete at about 830° C.

No. 4 Bar (0·6 per cent. carbon).—Diffusion is complete at 770° C.

No. 5 Bar (pure pearlite (saturated) steel which contains 0·9 per cent. carbon).—Diffusion or solid solution is complete at Ar1 point, 690° C.

No. 6 Bar (1·4 per cent. carbon).—The excess of cementite shows here. Diffusion is complete at 1000° C. (Ar3 point).

On the figure the writer has drawn lines, the loci of the Ar3 points, for the different percentages of carbon contained in the bars, and it is remarkable to note how nearly this comparatively rough experiment of Mr. Stead's agrees in each case with the theoretical point of diffusion.

From the foregoing it will be seen that steel, instead of being the homogeneous material it is popularly supposed to be, in reality partakes more of the nature of a crystalline rock; and this is especially true of the normal low and medium carbon steels so much used for constructive work and machine details. When the complex nature of its structure is taken into consideration it appears probable that such steel may break in more than one way; and apparently this is the case, for when broken by a gradually applied tensional or repeated or alternated load there are invariably present in the neighbourhood of the fracture the slip planes between the particles first noticed by Professor Ewen and Mr. Rosenhain, but slip planes have not been found in any case where the fracture has been due to shock.

That capacity to resist shock is a property distinct from that of resistance to progressive loading is shown by the following tables of the results of experiments conducted by Mr. Seaton,

– 525 –

Captain Sankey, and the writer. The impact test adopted by Mr. Seaton consists in allowing a tup of 6 lb. weight to fall through 2 ft. on a notched specimen, of square section, ½ in. side, placed on a span of 3 in., the specimen being turned over after each blow, its resistance to shock being measured by the number of blows required to produce fracture. The author has designed and had constructed a machine for carrying out the same test. This tester is illustrated by Fig. 7.

The machine used by Captain Sankey is on the Charpy principle as modified by Izod. A comparatively frictionless pendulum with light rod and heavy bob is allowed to fall freely

– 526 –

to the bottom of its swing, where it encounters a notched specimen, the work done breaking which is readily obtained, as the angular distance by which the pendulum fails to rise to the height which it attains when swinging freely is recorded by a pointer.

The following tables are extracted from results published by Messrs. Seaton and Jude and Captain Sankey. (The experiments at Canterbury College are not yet complete. They will appear as a supplement to this paper).

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Seaton And Jude (Table I).*
Tension 3 in Elongation per Cent on 2 in Impact Blows
Cold-drawn bar, less than 0·1 per a 31·63 31 1
cent. carbon steel b 31·27 20 180
Low-carbon steel, 0·15 per cent. a 28·3 37 175
carbon b 27 33 5
Medium-carbon steel, 0·25 per a 36·5 35 5
cent. carbon b 36·5 35 27
Mild cast steel, 0·35 per cent. car a 29·18 22·7 1
bon b 28·25 31 1

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Sankey (Table I)
Description Treatment Maximum Load Yield-point Elongation on 2 in. Reduction Area Impact
Per Cent Ft-Ib.
As received 34·7 21·6 21·5 52 1·3
Studs teel (very short).
Annealed 28·1 21·4 36·5 58 21·2
(very tough).

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Sankey (Table II).
Material Maximum Load Yield point Elongation on 2 in Impact
Steel for small forging 1 43·2 31·8 23·5 8·6
2 45·2 32·7 23·0 1·0
Steel for crank-shaft 1 30·8 14·8 30·0 0·9
2 28·7 13·2 32·0 4·8
High tensible steel 1 60·1 49·8 18·0 18·5
2 61·8 51·3 16·0 0·5
Steel for small crank-shaft 1 26·0 38·0 2·3
2 24·8 35·0 16·4
Steel for small forgi 1 39·4 28·0 3·0
2 40·3 32·0 13·4

[Footnote] * “Impact Tests on the Wrought Steels of Commerce” (Proceedings Mech Engrs, 1904).

– 527 –

That the impact test on notched bars is of marked value has been fully demonstrated by the clearing up by this method of testing of much of the mystery which has surrounded the results of the oil hardening and tempering of steel.

Oil tempering as generally practised for guns, axles, &c., consists in heating to about 850° C., then′ quenching in oil at 80° C., after which the material is reheated to between 550° and 650° C. By the first heating (to above the Ar3 point) fineness of structure is obtained, and by the quenching retained. By the second heating (to below the Ar1 point) any mechanical stresses due to the quenching are removed, the hardening carbon converted into cement carbon, but the fineness of structure is unaltered. Practice has proved that steel so treated is safe for gun-construction, and when not so treated unsafe. Yet the tension tests show only an increase of strength and ductility by oil tempering of something between 5 and 30 per cent.

The improvement in the material is out of all proportion greater than that indicated by these values; and Mr. Seaton, applying his test, finds that by so tempering, the impact strength of the steel, as measured by the number of blows required to produce fracture, is increased to between 500 and 600 percent. of that of the untempered steel in its best condition.

Messrs. Seaton and Jude also find that oil tempering has the property of levelling up the “shock” strength of steel to a fairly constant quantity. Two articles may have been made from the same grade of steel: owing to difference of heat treatment in manufacture one may be coarsely crystalline and dangerously brittle, the other fine-grained and tough. Oil quenching will bring the shock strength of these up to a common value—the increase in one case being measured by thousands and the other only by hundreds percent.

Since it would appear that the failure of steel members may be brought about by (1) statical overloading, (2)fatigue, (3) shock (the effect of which is apparently cumulative). it would seem rational that for all practical purposes material should be tested in the same manner as that in which it is to be loaded when in use. Thus, for structural work, where only a dead load is to be carried, a tensional test is sufficient, for from this in all but special cases the compressional, cross-bending, and shearing strengths can be computed. For structural work with gradually applied live load the tensional test is again required, as from this and the ratio of live to dead load the working-stress permissible, with security against fatigue, is obtained.

– 528 –

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Seaton And Jude—(Table II).*
Effects of Oil Quenching at about 800°C.
Normal Condition Oil quenched
Description Impact: Blows Max Load in Tons per sq in Elongation per cent on 2 in Condition Analysis Average per Cent Impact: Blows Max. Load in Tons per Sq. in Elongation per Cent on 2 in.
7 S=0·032 to 0·056 165
Low carbon acid 5 26·9 29·8 As received P=0·033 to 0·023 159 25·7 31
steel 165 Mn=0·63 to 0·75 177
163 Si=0·065 to 0·057 181
Basic steel- C=0·11, S=0·051 301 27·8 35
293 23·4 46 As received Mn=0·45, P=0·044
Low carbon C=0·17, S=0·050
217 26·3 40·5 As received Mn=0·65, P=0·052 280 31·5 38
C=0·73, S=0·056
High carbon 1 55·7 12 Mn=0·6, P=0·053 5 74·4 3
23 32·5 35 As received 105 37·8 31·4
35 Annealed in C=0·23
Admiralty steel lime
medium carbon 35 Annealed in S=0·036
5 As received P=0·058 lll
47 Annealed in Mn=0·76
Ordinary grade lime
5 Annealed in Si=0·05

[Footnote] * Proceedings Mech Engrs, 1904

– 529 –

Dangerous brittleness can be guarded against by the impact test. For rails, axles, ordinance, and all moving parts of machines and engines the impact test is of primary importance (it has been calculated that 85 per cent. of the parts of a modern high-speed engine are subject to shock); tensional tests must, however, also be made to obtain the data necessary for guarding against the effects of fatigue.

In conclusion it may be pointed out that no uncertainty need now exist as to the state of a fractured hook, chain, or machine part. A “shock” test will at once make clear whether the piece failed under overload, or was in a dangerously brittle condition. Further, the condition of existing parts and the rate at which the shock strength of axles, rails, tires, piston-rods, crank-pins, and chains is being used up can be readily ascertained, and the parts withdrawn from service before failure, possibly disastrous, ensues.

The object of this paper being to direct attention to and place in a condensed form some recent developments in the theory of steel, contemporary literature, such as “Harbord's Steel,” and the proceedings of various scientific bodies, has been liberally drawn upon.