COMPLEX STRESS DISTRIBUTIONS IN ENGINEERING MATERIALS. 191 
tion in any kind of test ; or to stress accumulation (No. 58). From the 
known difficulty of getting an axial pull or push in a direct static test, it is 
to be expected that there will be some bending in alternate stress tests 
with directly applied tension and compression. In No. 43 it would appear 
that bending oscillations would surely have been detected by the apparatus 
used to measure the lengthwise strain. The records of the only other 
experiments in which straims corresponding to directly applied stresses were 
measured, viz. Nos. 4 and 74, do not state that any bending effect was 
observed. It is noteworthy that the experiments in No. 82 with varying 
ratios between the maximum tensile and maximum compressive stresses 
gave little variation in the values of the limiting range; showing that 
bending, if any, had little effect ; or that the bending was caused equally 
during tension and compression. Though it is difficult to draw conclusions, 
it seems likely that the line of resultant force in the specimen does not 
appreciably alter its position after the first few alternations of the approxi- 
mate limiting stresses. This early change of position in this line would be 
one tending to equalise the distribution of stress in the specimen.* 
Alternate bending tests upon solid rotating bars give an apparently 
greater value for the Wohler limiting range. The number of revolutions 
required for fracture is considerably greater than the number of reversals 
in tests with stresses directly applied. The‘ /, »’ curves for the former 
are not, generally speaking, even approximately parallel to the axis of 
‘n’ after 10® revolutions ; while in the latter there is indication that the 
curve is asymptotic to a line, parallel to the axis ‘ n,’ and not far removed 
from the curve, at this number of cycles. It is to be expected that the 
calculated maximum stresses in a bending test will be somewhat higher 
than the actual, even when the bendings give the Wéhler limiting stresses, 
because of stress equalisation near the ‘ skin’ of a specimen (see remarks 
on No. 23). Unwin remarks f that ‘ Bending experiments are not less 
trustworthy than tension experiments, and for stresses considerably less 
than the statical breaking weight probably the error in the calculated stress 
is not a large one.’ Hollow test bars are found to give ‘ f, n’ curves more 
nearly approaching the curves for directly applied tension and compression. 
Carbon-Content of Steels and Resistance to Alternating Stress. 
Speaking of steels which consist partly of pearlite and partly of ferrite, 
Rosenhain (No. 66) remarks: ‘ From the point of view of the resistance te 
comparatively steadily applied alternating stresses, the higher the carbon- 
content up to 0-9 per cent. of carbon, the better the resisting power of the 
metal.’ Nos. 23, 47, 62, 82, 90, 93, and especially No. 48, contain evidence 
in accordance with this statement. Heat treatment of steels may have, of 
course, an enormous influence on their resistance. 
Effects of Annealing and Quenching upon Resistance to Alternating Stress. 
The effect of the ‘annealing’ which has been done t¢ upon the speci- 
mens of published alternate stress tests has been, in general, to diminish 
the resistance as compared with that of the material in untreated com- 
mercial condition ; and the effect of the quenching done has been to 
* See No. 82, and Proc. I.C.E., elxvi. p. 100. 
_ > The Testing of Materials of Construction, 1910 ed., p. 377. 
{ See Appendix II. 
