290 REPORTS OK THE STATE OP SCIENCE.— 1916, 



over two sections symmetrically placed about the axis of rotation. In 

 the tests the distance of the hanging weight from the axis of the specimen 

 was 8| inches, so that the ratio of the twisting moment to the bending 

 moment was about 20. 



For reversals of simple bending, a test of a specimen in an ordinary 

 fatigue-testing machine of the Wohler type would have been sufficient 

 for the prediction of the fatigue limit. It was considered, however, of 

 fundamental importance to determine if the effect of the reversals of 

 bending produced in this machine were of the same amount as those 

 produced by the continuous rotation of a loaded bar as in the ordinary 

 Wohler test, and for this purpose a special device was employed, which 

 is illustrated in fig. 2c, which is an elevation of the testing head with the 

 specimen in position. It will be seen that the axis of the load is made to 

 intersect the axis of the specimen, i.e. the torsional moment is made zero, 

 by extending the hanger so as to envelop the head when rotating, and 

 the load is transmitted to the specimen through the ball-bearing in the 

 specimen itself. In this way reversals of simple bending are produced 

 in the specimen, the essential difference between this case and the Wohler 

 test being that in the former the maximum stress is confined to the axial 

 plane in the specimen perpendicular to the axis of rotation. 



The Method of Carrying Out the Tests. 



In the ordinary system of testing for the prediction of the limiting 

 fatigue range of stress it is customary to have a fairly large number of 

 specimens, and to commence by imposing a range of stress which will 

 probably cause fracture after a few thousand reversals. The next specimen 

 is then tested under a smaller stress range, and so on until a range is 

 found which the specimen will bear indefinitely. In the present case the 

 cost of each specimen was so considerable that the reverse method to the 

 above was adopted, i.e., a comparatively small range was first imposed, 

 and if after three million reversals fracture had not occurred, the load 

 was increased by about five per cent., and the test carried on. Finally 

 a stage was reached when fracture took place with less than three million 

 reversals. A new specimen was then fitted to the machine and tested 

 at what was considered to be the limiting range. In this method the 

 time taken in the series of tests required for the prediction of the fatigue 

 limit is longer than in the former case, but considerable economy in the 

 cost of preparation of specimens is eflected. 



