Research On Optimization Algorithms Based On Subset Simulation For Management Objectives Of Construction Projects

Solving the optimization problems for management objectives of construction projects efficiently is the focus of construction project management.At present,the research on construction project optimization problems mainly focuses on single-objective optimization of time and cost,but with the development of the construction industry,management objectives such as construction quality and safety have also gradually become the mainstream of research.So there are more and more construction project many-objective optimization problems.Therefore,the performance of optimization algorithms is also higher required.Based on these reasons,this thesis proposes two new optimization algorithms to solve the single-objective and many-objective optimization problems respectively by introducing the subset simulation method.Firstly,this thesis chooses appropriate quantitative methods to quantify the construction time,cost,quality and safety from the perspective of execution modes.By this way,the inconsistent quantitative benchmark of sub-goal has also been solved.Based on this work,this thesis constructs discrete time-cost and time-cost-quality-safety trade-off model.Secondly,to solve the time-cost trade-off problem efficiently and stably,an optimization algorithm based on subset simulation is proposed.By introducing a group of thresholds with respect to the objective function,a sequence of shrunk regions can be defined in the proposed algorithm.By sequentially sampling in the shrunk regions,a small region around the optimal solution can be effectively explored in the proposed algorithm,in order to achieve the optimal solution.Besides,Markov chain Monte Carlo simulation using random walk with reflecting barriers is used in the algorithm,in order to address the problems of efficient stochastic sampling in the feasible region and the sequence of shrunk regions.Thirdly,based on the decomposition strategy,this paper proposes a many-objective subset simulation method.By combining with the characteristics of MOEA/D in weight vector and aggregation method,this method can solve the discrete time-cost-quality-safety trade-off problem efficiently and extend the subset simulation method to the field of many-objective optimization.Finally,in this thesis,the proposed optimization algorithm based on subset simulation is verified by example 1 and example 2,and compared with the widely used genetic algorithm.It is proved that the single-objective optimization algorithm proposed in this thesis has a great improvement in the stability of obtaining the optimal solution.Another proposed many-objective subset simulation method based on decomposition is verified by example 3.By comparison with MOEA/D,it is proved that the proposed optimization algorithm has more advantages in convergence,diversity and stability.