准确的预测海底腐蚀管道极限承载力对于评估海底管道寿命、保证海底油气管道安全运行具有重要意义。针对BP神经网络（BPNN）模型学习效率低、对初始权重敏感且容易陷入局部最优状态等缺点，采用人工蜂群（ABC）算法来优化BPNN的初始权值和阈值，建立了ABC-BPNN海底腐蚀管道极限承载力预测模型。使用MATLAB软件搭建模型并进行预测，同时与BPNN模型、遗传算法（GA）优化的BPNN模型和粒子群优化算法（PSO）优化的BPNN模型进行对比分析。结果表明：ABC-BPNN模型预测海底腐蚀管道极限承载力的平均相对误差为1.975 0%，优于BPNN、GA-BPNN和PSO-BPNN模型的预测结果；ABC-BPNN模型的预测结果所拟合出来的直线与Y=X直线最为贴近，证明了ABC-BPNN模型作为预测海底腐蚀管道极限承载力工具的准确性和稳定性。

Accurate prediction of the ultimate bearing capacity of submarine corroded pipelines is of great significance for assessing the life of submarine pipelines and ensuring the safe operation of submarine oil and gas pipelines. Aiming at the disadvantages of low learning efficiency of BP neural network (BPNN) model, sensitive to initial weights and easy to fall into local optimal state, artificial bee colony (ABC) algorithm was used to optimize the initial weights and thresholds of BPNN, and ABC-BPNN prediction model for ultimate bearing capacity of submarine corroded pipeline was established. Using MATLAB software to build the model and to make predictions, and compare and analyze with the BPNN model, the BPNN model optimized by the genetic algorithm (GA-BPNN) and the BPNN model optimized by the particle swarm optimization algorithm (PSO-BPNN). The results showed that the average relative error of the ABC-BPNN model to predict the ultimate bearing capacity of submarine corrosion pipelines was 1. 975 0%, which was better than the prediction results of the BPNN, GA-BPNN and PSO-BPNN models. The line fitted by the prediction results of the ABC-BPNN model was the closest to the Y=X line, which proved the accuracy and robustness of the ABC-BPNN model as a tool for predicting the ultimate bearing capacity of submarine corrosion pipelines.