REASONABLE RECONSTRUCTION OF ONE-DIMENSIONAL MECHANICAL STRENGTH PARAMETERS

Monte-Carlo simulation method for the reconstruction of one-dimensional probabilistic mechanical strength parameters is developed to realize reliability analysis at arbitrary given probability levels. This method is valid to the case that the probabilistic parameters are given only in the form of special survival probabilities (Ps) or combined survival probability-confidences (P-Cs) and the reliability analysis can not be made except for the given Ps or P-Cs. To avoid the common simulations with numerous sampling sizes over possible practice to make a non-conservative evaluation, a simulation policy is newly suggested with from 7 to 20 sampling size for material small specimens and, at most, 10 for structural specimens, and 5% as the acceptable error for restoring original parameters. The details of the method are studied for the six possible distributions, i.e. normal, lognormal, three-parameter Weibull, two-parameter Weibull, extreme maximum value, and extreme minimum value ones. The reconstruction practice for the fatigue limits of 10 engineering materials has illustrated the availability and feasibility of the present method.