Attention-enhanced generative design of disc cutter layout for shield TBM considering multi-working conditions

The widespread application of Tunnel Boring Machines (TBMs) in composite strata and variable geological conditions poses significant challenges to the generalizability and robustness of cutter layout design. This paper introduces a robust generative design method for cutter layouts considering multi-working conditions, aiming to enhance excavation efficiency and safety in diverse geological settings. An attention-enhanced meta-model is trained using the dataset constructed via parametric dynamic analysis and subsequent post-processing. A multi-objective optimization (MOO) method, guided by a multi-condition evaluation, generates a Pareto optimal set, from which the optimal solution is selected using multi-attribute decision-making (MADM). A case study involving a TBM excavating through representative ground conditions validated the method’s feasibility, demonstrating 14.34 %, 33.71 %, and 2.63 % improvements across different working conditions while maintaining safety standards. This study contributes a generative design method integrating a graph attention meta-model and multi-condition evaluation strategy, and an efficient generative system based on a co-simulation platform.