Abstract
Crushing tests of ceramic bearing balls are usually carried out by compressing two balls of the same size against each other. In the conventional data analysis, it is considered that n pairs of balls yield the crushing load data of sample size n. In the present paper, a new method for data analysis was examined. In this method, it is considered that n pairs of balls yield the crushing load data of the sample size 2n of which about half are censored values. The new method is based on the experimental observation that one ball usually remains intact even if the other ball fails. Hence, the method seems to be more reasonable than the conventional method. Applying the new method to our experimental data reported in the previous paper, the followings were found: (1) a two-parameter Weibull distribution fitted well to the experimental data when analyzed by the new method as well as when analyzed by the conventional method, (2) the shape parameter by the new method was almost the same as that by the conventional method, and (3) the scale parameter by the new method was slightly larger than that by the conventional method, and the difference was only 5%. These results justify the use of the conventional method from a practical point of view. Furthermore, a theoretical analysis was performed, and the above results (1), (2) and (3) were found to agree with the theoretical predictions.
