极端波浪作用下深水浮桥结构失效概率主动学习方法

Recent national strategies for developing comprehensive multidimensional transportation systems in China have prompted bridge construction projects to connect islands and crossing straits along the coast. Deepwater floating bridges have attracted considerable attention as potential technical solutions. Analyzing the structural failure probability of deep-water floating bridges under extreme wave loads is important for conceptual design. Accordingly, a computational framework was established in this study for estimating the extreme response and structural failure probabilities of floating structures under narrow-banded random waves using an active learning approach. A deep-water floating bridge segment finite element model was developed for the dynamic analysis. Active learning was employed to examine failure probability learning patterns under various failure modes based on displacement and tension leg forces with different thresholds, and finally obtain failure probability predictions. The active learning method requires only limited samples to achieve convergent failure probability predictions, requiring a low computational cost. The active learning framework was applied to both the single-and multi-objective failure criteria. For the floating bridge studied, under random waves with a return period of 100 years, the failure probability based on the midspan displacement was 2. 5% at the threshold ratios of W_S./W_B= 2. 2 and 1. 3% based on the tension leg force at the threshold ratio of F_s/F_B= 2. 5. These rapid computational failure probability estimates can provide information for feasibility assessments, designs, disaster prevention, and protection technologies for future deepwater floating bridge projects.