Intelligent design of key joints in aerial building machine using topology optimization and generative adversarial network

Abstract

Joints are crucial connections in an aerial building machine (ABM), yet they often undergo experience-based local optimization design. This paper presents an intelligent design method for key joints in the ABM using a generative adversarial network (GAN), aiming to achieve new and superior global optimization schemes. A database of topology-optimized structures is fed into the boundary equilibrium GAN (BEGAN) for training, which in turn generates innovative and diverse design schemes. The optimal scheme selection under multi-working conditions is then realized by the multiple-attribute decision-making (MADM) method. A case study of an ABM joist confirms the effectiveness of this method, showing it meets safety requirements under various conditions and achieves significant improvements (43.45 % for construction, 43.67 % for jacking, and 42.89 % for shutdown). Additionally, the BEGAN model surpasses existing generative models for ABM joint design. To determine evaluation schemes and optimal designs, this paper provides a method for global optimization of joints that considers the integrated effects of multiple conditions, constructing a rapid and comprehensive solution for designing and evaluating key joints in the ABM.