G. — ENGINEERING. 129 



investigation than they have so far received — e.g., the effect of speed 

 of .testing; the effect of rest and heat treatment in restoring fatigued 

 material; the effect of previous testing at higher or lower stresses on 

 the apparent fatigue limit of a test piece. Some observers have found 

 indications that the material may possibly be strengthened by subject- 

 ing it to an alternating stress below its fatigue limit, so that the results 

 of fatigue tests may depend on whether the limit is approached by 

 increasing the stress or by decreasing it. 



Improved methods of testing are also needed — particularly methods 

 which will give the results quickly. Stromeyer's method of measur- 

 ing the first rise of temperature, which indicates that the fatigue limit 

 is passed, as the alternating load is gradually increased, is most promis- 

 ing; it certainly will not give the true fatigue limit in all cases, for it 

 has been shown by Bairstow that with some ranges of stress a finite 

 extension occurs at the beginning of a test and then ceases, under 

 stresses lower than the fatigue limit. But the fatigue limit in that 

 case would not be a safe guide, for finite changes of shape are not 

 permissible in most machines, so that in that case also Stromeyer's 

 test may be exactly what is wanted. It can probably be simplified in 

 detail and made practicable for commercial use. Better methods of 

 testing in torsion are also urgently needed, none of those at present 

 used being free from serious defects. Finally, there is a fascinating 

 field for physical research in investigating the internal mechanism of 

 fatigue failure. Some most suggestive results have already been 

 obtained, which extend the results obtained by Ewing. 



For members of structures which are only subjected to steady loads 

 I suggest that the safe stress might be defined by limiting the corre- 

 sponding permanent set to a small amount — perhaps ^ per cent, or 

 J per cent. This principle has been tentatively adopted in some of the 

 aircraft material specifications by specifying a Proof Load which must 

 be sustained without a permanent extension of more than ^ per cent. 

 Whether this principle is suitable for all materials and how it will 

 answer in practice remains to be proved by experience. It is at any 

 rate a possible rational basis for determining the useful strength of a 

 material under steady loads. 



The relation between the proof stress and the shape oi the stress- 

 strain diagram is shown in the lantern sUde. The curve is the record 

 of an actual test on a certain copper alloy. If a length A B correspond- 

 ing to ^ per cent, elongation be set off along the base line and a line B P 

 be drawn through the point B parallel to the elastic line, to cut the curve 

 in P, then the stress at P is the stress which will give ^ per cent, 

 permanent set. Though I per cent, may appear rather a large 

 permanent set to allow it will be seen from the figure that it is less 

 than the elastic elongation would have been at the same stress, and we 

 do not usually find elastic elongations serious. 



As a commercial test the proof load is very easily applied. For this 

 alloy the specified proof load is shown by the horizontal line so labelled. 

 This load is to be applied and released, and the permanent ext>ensicn is 

 required by the specification to be Jess than ^ per cent. This sample 

 passes the test easily. On the figure the ■condition for complying with 

 1920 K 



