THE MEASUREMENT OF STRESS BY THERMAL METHODS. . 243 
approximately square in section, with 31 couples, and had a resistance of 5°55 ohms. 
The short connecting wires or leads had a resistance of 0°52 ohms. In a few experi- 
ments, which are specially noted, a linear pile was used of 10 couples and of 0°18 ohms 
resistance, and also long connecting wires or leads of 1°81 ohms resistance were used in 
some cases. A plot of these readings is shown on fig. 6, in which curve I shows the 
relation of the stress to the galvanometer readings. In order to obtain the true reading, 
correction must be made for the losses due to radiation and change of resistance. ‘The 
first is the only important one. The readings corrected by the formula D=D,(1 +5) 
are shown in the table above, and the plot of these, with the load as abscissee, is shown by 
eurve II, giving almost exactly a straight line to near the yield point (fig. 6). ‘The stress- 
strain relation obtained from the extensometer readings is plotted for comparison upon 
the same diagram, the unit of extension being 0:00001 inch, and this also exhibits a 
nearly linear relation up to the yield point. The result of the experiment appears to 
show that the thermal changes do not indicate a range of imperfect elasticity within the 
apparent limits of elasticity of shape. 
A second experiment upon a wrought-iron specimen having a section 2 inches by 
0°25 inches was next subjected to stress in the testing machine in a similar manner, and 
the observations are recorded in Table VI., and a plot of the readings is shown in fig. 7. 
The observations made to determine the radiation loss are omitted, as they are of a 
similar character to the example quoted above. ‘The value of k obtained was 0°0031, 
the time being measured in seconds. 
The general character of the diagram is the same as in the last case; there is a 
gradual bending over of the galvanometer readings towards the time axis, the deviation 
from a straight line being nearly in a geometrical progression with regard to time. The 
apparent coincidence of the lower readings with the dotted straight line is probably not 
exact. It should be noted that the stress-strain curve would practically coincide with 
the corrected thermal stress curve if sheared over, except near the upper end, where the 
heating effect begins to play a part. In both cases the thermal readings begin to show 
deviations from a linear relation to the stress at about the same value of the stress. In 
other experiments upon different bars of iron and steel, results were obtained confirming 
those quoted above. It therefore appears probable that the thermal change is very 
nearly proportional to the stress, in the same manner as the strain; and that, for the 
material experimented upon, there appears to be no range of imperfect elasticity as 
measured by thermal change, coinciding with a part of the range of perfect elasticity as 
determined by the strain. 
6. Toe RELATION OF STRESS TO STRAIN AND THERMAL CHANGE IN 
SHORT CoMPRESSION MEMBERS. 
[t is well known that the relation of stress to strain in short compression members 
of wrought-iron and steel follows a linear relation for a considerable range of stress, and 
