82 
then case-hardened and ground. No further grinding or other treatment to 
produce a greater degree of accuracy was given to any of the bearings used. 
Four tests were made at each of four speeds for every load, so that sixteen 
tests were made for each load. With some of the roller-bearing tests, 
where the results varied markedly, a greater number were made. In the 
curve of tests showing the comparative friction, each point is the average 
of four tests, so the position of the line is the result of sixteen tests. The 
loads were varied from fifty pounds to five hundred pounds. It was found 
that while the friction increased with the load in a nearly constant ratio 
for the balls, for the rolls there was a great variation at and after three 
hundred pounds. This is shown by the points on sheet 2. An examination 
of the shaft showed that it was being tor away in small flakes under the 
300-pound load, and this tearing increased as the load was made greater. 
At 500 pounds the shaft was torn away quite rapidly, especially at the 
higher speeds, and after a few minutes’ operation, a ridge was formed on 
the outside edges of the path of the rollers. This ridge had to be filed 
down on one side before the cage and rollers could be removed from the 
shaft. Neither the rolls nor their hardened steel race were affected, 
though, as already mentioned, the sides of the cage were cut by the ends 
of some of the rolls. Of the fourteen rolls in one cage, this wearing oc- 
curred at both ends of four of them. In making measurements on the 
dynamometer, a scale reading to fractions of ounces was used, so in plot- 
ting curves the unit used was the ounce. 
The diagram of the friction curve for the roller bearings shows the 
points for the measurements taken at each load to be within spaces that 
increase slightly until the 250-pound load is passed, when the spaces be- 
tween points increase in such manner as to show that the pull on the seale 
was due to more than the friction. It will be seen, however, that most of 
them seem to fall below the line made by the curve up to that point, if 
the line were produced. The points were so distributed that a curve 
drawn through the average position would not mean much. Why they 
fell so low in some cases I was unable to determine. 
The diagram for the friction of the ball bearings shows the points 
within small spaces up to the 500-pounds load. How much farther this 
would continue with the kind of bearing used I did not determine, though 
I found on another test made on smaller balls and bearings, that both 
balls and bearings began to pit soon after the load exceeded 500 pounds, 
