PHYSICS, CHEMISTRY AND ENGINEERING 183 



crystal. Figure \c shows the appearance of the same piece 

 of iron after several hundred more repetitions of stress. The 

 slip lines are more numerous than in Figure lb and more crys- 

 tals are "infected" by slip lines. Figure Id shows the ap- 

 pearance of the same piece of metal just before it broke under 

 repeated stress. The slip lines had become very numerous 

 and at aa had spread until a crack had formed between the 

 crystals. Shortly after the formation of this crack the piece 

 failed. 



The effect of this progressive failure of metal is to weaken 

 the section of a bar, just as a nick cut into it would do, and ex- 

 plains why under repeated stress failure takes place suddenly 

 just as it does when a nicked bar of metal is bent. 



The problem which faces the engineer is to design members 

 so that they will not fail under repeated stress. In general 

 the smaller the stresses the less the danger of failure by re- 

 peated stress. A common idea concerning metals is that there 

 exists an absolute elastic limit below which metal is absolutely 

 elastic and below which no amount of repeated stress can in- 

 jure material. It can not be stated positively whether such 

 an absolute elastic amount exists, but in tests under repeated 

 stress failure has occurred at stresses less than the elastic limit 

 has commonly determined by refined testing methods. 1 



The best method of determining safe stresses for metals 

 under repeated stress seems, to the writer, to be the direct ex- 

 perimental study of test specimens subjected to known stresses 

 of varying magnitude repeated many times until failure oc- 

 curs. Fig. 2 is plotted from the results of such a series of 

 tests, and its general form is typical. Two methods of inter- 

 preting the results of such a curve are in use; in one it is as- 

 sumed that the curve becomes horizontal, and from the test 

 data a horizontal asmyptote to the curve is drawn by estima- 

 tion, and the stress-ordinate of this asymptote taken as the en- 

 durance or "fatigue" limit for the metal. In the other method 

 the assumption of a horizontal asymptote is discarded, and an 

 attempt made to find some simple form of equation which 

 fits the test data. For a wide range of test results of fatigue 



'The method of determining the elastic limit of a material is to apply known 

 loads to a specimen of the material and then to release the loads. Measurements 

 of length of specimen are made before and after the application and release of each 

 load. When any change in length can be detected after release of load, the elastic 

 limit has been reached. 



