A computational study on bounding the makespan distribution in stochastic project networks

Due to the practical importance of stochastic project networks (PERT-networks), many methods have been developed over the past decades in order to obtain information about the random project completion time. Of particular interest are methods that provide (lower and upper) bounds for its distribution, since these aim at balancing efficiency of calculation with accuracy of the obtained information. We provide a thorough computational evaluation of the most promising of these bounding algorithms with the aim to test their suitability for practical applications both in terms of efficiency and quality. To this end, we have implemented these algorithms and compare their behavior on a basis of nearly 2000 instances with up to 1200 activities of different test-sets. These implementations are based on a suitable numerical representation of distributions which is the basis for excellent computational results. Particularly a distribution-free heuristic based on the Central Limit Theorem provides an excellent tool to evaluate stochastic project networks