Duration-cost optimization in earthmoving operations using NSGA-II and simulation techniques

Innovative approaches for optimizing earthmoving fleets have been proposed in the field of construction management. Despite the emerging productivity analysis and optimization technologies, existing studies have witnessed the difficulty of real-life data collection on-site. Accordingly, this paper proposes a synthetic data generation method using the information extracted from the Korean Construction Standard Productivity Rate (CSPR) document. The extracted information was recalculated to activity times that were used as input in a Discrete Event Simulation (DES) model using the WebCyclone technique for producing synthetic data for conducting a productivity-prediction model-development practice using an Artificial Neural Network, XGBOOST, Random Forest and Duration-Cost optimization practice using non-dominated Sorting Genetic Algorithm II (NSGA-II). The comparison results showed that all three methods provide excellent goodness of fit and the NSGA-II can successfully deduce the Pareto front for Duration-Cost optimization.