Estimating the mean completion time of PERT networks with exponentially distributed durations of activities (Q1319548)

The paper presents an efficient method for estimating the expected duration of a project with exponentially and independently distributed activity times. Since the hypoexponential distribution, being the convolution of exponential distributions, is not preserved under multipliation, the distribution of the maximum of two independent random variables with hypoexponential distributions is proposed to be approximated by a hypoexponential distribution having the same mean and variance. Though the method tends strongly to overestimate the exact value of the expected project duration, it does not have bounding properties. The accuracy of the method is examined on some simple examples, and the results are encouraging.