Critique of test methods for lifetime predictions.

Failure predictions for ceramics depend on the experimental parameters that measure the strength distribution (m and sigma 0) and the time-dependency of strength (n). These parameters can be determined by measuring strength as a function of stressing rate in a test environment that simulates the service environment. To minimize the uncertainty in these experimental parameters, at least 30 samples per stressing rate should be tested over a stressing rate range of at least 3 orders of magnitude. The uncertainty in these experimental parameters can be taken into account in design calculations by the use of appropriate safety factors. Thus, through well designed experiments coupled with a reliability analysis, rational design decisions can be made that ensure the successful use of ceramics in demanding structural applications.