Sequential experimentation for estimating the optimal delay of activities in PERT networks

Applies the response surface methodology (RSM) utilizing simple fractional designs to optimize the expected present value (EPV) of a stochastic project network with respect to the delay of activities. Such a delay could increase the EPV by postponing negative cash flows, possibly at the expense of also delaying the final payment for the project. The problem was challenging, due to a combination of high variability with a relatively flat objective function near the optimum. Comparisons with the true optimal solutions, where available, indicated the robustness of the approach.