USE OF CONSTRAINED LOGNORMAL-DISTRIBUTION IN RELIABILITY-ANALYSIS

This paper presents a method for adjusting the distribution of error probabilities in cases where one is faced with error bounds that are not mathematically meaningful. This approach is also amenable for fitting a distribution to positively skewed observational data, where the data represent probabilities of error. The technique is mathematically rigorous and is relatively straightforward to apply and can be automated easily. The technique was developed during a ground-up human reliability analysis in which uncertainties in basic human errors were propagated through the full human reliability event trees used to model operator errors.