Using Gaussian copula to simulate repetitive projects

An important requirement for simulating repetitive projects is to treat correlations inherent in the repetition of same crews working at various locations. To attain the requirement, this study develops a new Monte Carlo simulation model implementing a Gaussian copula in conjunction with the inverse-transform method to generate correlated duration samples in repetitive projects that have pre-specified marginal distributions and pairwise rank correlations. The proposed model is equipped with an automatic approximation procedure to adjust an infeasible correlation matrix, if necessary. The proposed model is statistically verified through a real-life residential apartment project. The simulation results are compared to two conventional analyses (PERT and simulation without correlation) to show the aggregated impact of correlations. The proposed model contributes to the state-of-the-art in handling non-linear dependencies among activity durations that may have non-normal distributions. Moreover, it is flexible in the ways of correlation assessments (qualitative or quantitative), the magnitudes of correlations (weak to strong), and the types of marginal distributions (symmetrical or skewed).