Surrogate model assisted solution of time cost trade-off problem in the construcion projects

This thesis introduces an approach to solving the discrete time-cost trade-off problem (DTCTP) in construction projects by combining the Arithmetic optimization algorithm (AOA) with a Multilayer perceptron (MLP) surrogate model. The proposed method alleviates the iterative computational effort of the Critical path method (CPM) by using the MLP surrogate model to predict CPM calculation results. The hyperparameters of the surrogate models were optimized using grid search. The MLP surrogate model demonstrated superior prediction accuracy and computational efficiency compared to other surrogate models, such as Support vector regression (SVR) and Random forest (RF). The performance of the surrogate model-assisted AOA was tested on DTCTP solutions for construction projects comprising 81, 146, 208, and 291 activities. The Pareto front solutions obtained using the surrogate model-assisted AOA reduced project duration by 4.61%, costs by 6.71%, and computational time by 71.24% compared to solutions obtained using AOA without surrogate model. These results indicate that the surrogate model-assisted AOA outperforms the conventional AOA, particularly in terms of computational efficiency, for solving large-scale DTCTPs in construction projects. This model offers a computationally robust tool for solving discrete DTCTPs, enabling project managers and decision-makers to generate faster solutions to this problem. By integrating a machine learning-based surrogate model into the optimization algorithm, the study contributes to advancing research in the field.