Intelligent generation method for innovative structures of the main truss in a steel bridge

The main truss is the core load-bearing component of a steel truss bridge, but most of the current steel trusses have the problems of a single form and insufficient optimization. Aiming at these issues, this paper proposes an intelligent generation method based on topology optimization and deep learning, which can automatically generate innovative structures with novel shapes and better mechanical performance. Firstly, the approach uses the topology optimization solver to optimize the initial main truss model under various working conditions, and optimization results are collected to establish a deep learning dataset. Then, the topologically optimized dataset of the main truss is fed into the least squares generative adversarial networks (LSGANs) algorithm for deep learning. Cloud computing technology is configured to generate a variety of new models intelligently. Finally, the developed schemes of innovative structures are evaluated from the aspects of novelty, diversity, mechanical performance, and mass. The research results show that the method can not only efficiently generate multiple innovative main truss structures in batches but also further optimize the mechanical performance and material consumption, which can provide a reference for the conceptual design of the main truss in a steel bridge, while the program for this method has not been developed to make it more suitable for practical engineering.