1991 年 40 巻 450 号 p. 276-282
The present paper deals with the reliability-based fatigue-proof design based upon a practical viewpoint through correct evaluation of engineering uncertainties caused by use of a limited number of experimental data. To this end is first discussed the distribution properties of the statistical failure probability constructed with the aid of parameter estimators in the Weibull distribution. At the same time, the effect of sample size on the estimation of failure probability is also investigated. Comparison is made between shapes of the distribution function of the statistical failure probabilities based upon a Weibull and a log-normal model in order to clarify which is more practical as the fatigue life distribution model of structural components. Furthermore, discussions are made on how to determine the optimum sample size of the fatigue experiment in order to meet the prescribed level of reliability imposed on structural components. Finally, a simple but important design principle is proposed for the reliability-based fatigue-proof design of structural components with Weibull fatigue life distribution.