人工蜂蚁算法结合BP神经网络的PC刚构桥优化

Optimization of PC Continuous Rigid Frame Bridges Using Artificial Bee Colony Algorithm and BP Neural Network

摘要:

为了解决现有预应力混凝土（PC）连续刚构桥优化易陷入局部最优且难以系统地实现结构尺寸和钢束优化的问题，提出了一种人工蜂蚁（ABC）算法结合BP神经网络的方法，启发式算法避免了局部最优，目标函数兼顾结构造价和受力性能，以满足规范和构造要求为约束条件，系统地实现PC连续刚构桥结构尺寸和钢束的优化。依托一座跨径布置为（95+173+95）m的连续刚构桥，通过比较九种算法与神经网络结合的优化效果和效率，突显了ABC算法的优势。最优方案不仅满足规范要求，且相较于原桥，目标值降低了35.8%，钢束用量减少了46.3%，应力安全度方差降低了60.4%，目标预测值与实际值仅相差2.1%，优化和预测效果显著。此外，对参数进行了重要性和敏感性分析，探索了不同参数对于目标值的影响。

Abstract:

To address the challenges of optimization for prestressed concrete （PC） continuous rigid frame bridges， particularly the issues of local optima and the difficulty in simultaneously optimizing structural dimensions and prestressing tendons， this study proposes a method combining the artificial bee colony （ABC） algorithm with the back propagation （BP） neural network. The heuristic algorithm helps to avoid local optima. The objective function considers both structural cost and performance， with constraints to meet specifications and construction requirements， achieving systematic optimization of structural dimensions and tendons for PC continuous rigid frame bridges. Based on a continuous rigid frame bridge with a span arrangement of （95+173+95） m， a comparative analysis was conducted using nine different algorithms combined with neural networks. This comparison highlighted the superiority of the ABC algorithm in terms of optimization effect and efficiency. The optimal scheme not only meets the specifications but also reduces the objective value by 35.8%， tendon usage by 46.3%， and stress safety variance by 60.4% compared to the original bridge， showing significant optimization effects. The predicted value of the objective for the optimal scheme differed from the actual value by only 2.1%， demonstrating the effectiveness of the prediction. Furthermore， importance and sensitivity analyses were conducted to explore the impact of different parameters on the objective value.