Efficiency, safety, and energy consumption during shield tunneling are difficult to optimize synergistically. To address this challenge, a multiobjective optimization decision-making method oriented toward real-time shield tunneling is proposed. This method aims to balance the trade-offs among efficiency, safety, and energy consumption, thereby enhancing overall tunneling performance. First, a dual-stage attention-based long short-term memory network is proposed to construct a prediction model linking tunneling parameters with performance indicators. Based on this model, objective functions are formulated, and both the mechanism and empirical constraints within the tunneling process are analyzed. Furthermore, to improve the efficiency of identifying global optimal solutions and enhance the diversity of the solution set under complex constraints, a nondominated sorting genetic algorithm II based on a dynamic operator update strategy is developed. The technique for order preference by similarity to the ideal solution method (TOPSIS) is then employed to conduct comprehensive decisionmaking on the Pareto front solution set, obtaining optimal tunneling parameter combinations under different preference strategies to guide parameter settings. Finally, using data from the shield tunneling ring database of the Zhanjiang Bay subsea tunnel project, experimental verification is conducted under two different tunneling conditions, along with section optimization tests. Under strategies prioritizing efficiency, safety, and energy consumption, performance improves by 7.65%, 10.26%, and 8.89%, respectively, compared to real-world tunneling performance, with other performance indicators also showing improvement.
