A machine learning-based force-finding method for suspend dome structures and case study

Force-finding is a critical phase in the structural design of suspend dome, involving the determination of both prestress distribution and magnitude. This study introduces an innovative machine learning-based framework for force-finding in suspend domes, designed to mitigate the limitations posed by the iterative processes and parameter settings of metaheuristic algorithms. Three distinct machine learning algorithms, back propagation neural network (BPNN), radial basis function neural network (RBFNN), and deep belief network (DBN) were employed to predict the prestress states in three different cases, demonstrating the effectiveness and validity of the proposed framework. The comparison with the calculation results of the elastic support method shows that the three algorithms can accurately predict the prestress of the suspend dome structure and meet the requirements of engineering accuracy, with the RBFNN performing particularly well. The proposed framework excels in terms of robustness, achieving optimal results and reducing computational costs in force-finding problems for suspend domes.