380 REPORTS ON THE STATE OF SCIENCE, ETC. 
From the general point of view, therefore, with material having a low yield- 
ultimate ratio as in mild steel, in the case of simple direct stresses the concentrations 
caused by rivet-holes do not greatly affect the endurance of the metal, so that 
under these conditions, if the real maximum stress could be accurately anticipated, 
then a standard stress a little below the yield-point could be adopted.1 
Nevertheless, as is shown by the above experiments, fatigue can occur in per- 
forated mild-steel pieces subject to unidirectional stresses: and it is desirable and 
important to avoid arrangements of rivets that produce undue concentrations of 
stress. This is particularly true for metals having comparatively high yield-points, 
for if higher stresses were allowed in such metals, fatigue would become imminent as 
indicated in Fig. 10. But even in very mild steel it is desirable that the range of 
alternating stress (A) should not exceed some definite fraction of the basis fatigue 
range (A,) no matter what value the steady stress may have. The values of this 
fraction, itis hoped, may be determined—for different perforations and joints—in 
the course of further experiments. 
If a high stress approaching the elastic limit be fixed as the standard for design, 
the stresses must be exactly calculated, allowing for secondary stresses in order to 
arrive at the actual stresses in the members, and the impact allowance must also be 
known with certainty, for the margin of strength, or, more exactly, margin of 
endurance, will be in the difference between the yield stress and the standard stress 
and in the impact allowance. 
The conception of ‘factor of safety’ has to be considerably modified when 
strength is related to endurance. The record of the test D shows that a maximum 
stress of 114 tons per sq. in. could be endured millions of times, but that anything 
above 124 would have produced fracture in a few thousand repetitions. . Assuming 
the load and impact damaging effects to be at their maximum and repeated uniformly, 
a member designed on a standard stress of, say, 114 tons per sq. in., to use the above 
figures, would do its work for an indefinite period, whereas, if designed on, say, 12? 
tons per sq. in., it would not immediately fail, but would deteriorate rapidly. The 
factor of safety cannot be related to these two stresses or to the number of repetitions 
necessary to produce failure in each case. 
Considering two such comparative designs with respect to normal traffic condi- 
tions, the occurrence of the maximum impact effect, depending on the worst engine 
and train combination travelling at a critical speed, might be so infrequent 
that the 123 stress design would show no more signs of deterioration than the other, 
and both would have a substantial margin. 
Evidence of the magnitude of impact factors and the frequency with which maxi- 
mum impact effects occur is afforded by the bridges which have withstood such effects 
for many years. For instance, if in a girder the measured maximum stress under a 
stationary locomotive or train were 9 tons per sq. in., it is certain that the impact 
factor cannot have been 100 per cent.; it is equally certain that the margin in the 
strength of the girder has been sufficient to withstand all the impact effects. It is for 
reasons similar to that suggested above that the measurement of actual static stresses 
in proved bridges promises a solution of the impact question. Measured stresses 
could be related to stresses as calculated in the usual way, and nominal working 
stresses adapted to different conditions could be fixed. 
The Relation of Measured Strains to Actual Stresses. 
By means of extensometers and recording strain-meters, strains in bridge members 
caused by stationary or slowly moving loads can be accurately measured. With 
rapidly moving loads the vibration of the bridge member, combined with that of the 
moving parts of the instrument, render the strain measurements less reliable, although 
the strain records obtained with Fereday-Palmer photographic instruments are 
considered very satisfactory. The conversion of the measured strains to the actual 
stresses introduces some difficulties. The rivet-holes in bridge members introduce 
1 Tt is interesting to note that where the permissible stress has to be raised to 
its highest limit to avoid rebuilding old bridges, in one instance the elastic limit has 
been fixed as the standard stress by one of the American Railways.—Indian Railway 
Bridge Committee, Fourth Report, vol. ii. 
