基于SSA-BPNN的海底腐蚀管道极限承载力预测  

Prediction of ultimate bearing capacity of submarine corroded pipeline based on SSA‑BPNN model

作　　者：刘博 周卫军[1] 马荣彬 LIU Bo;ZHOU Weijun;MA Rongbin(Oil and Gas Transportation and Sales Department,PetroChina Tarim Oilfield Company,Korla 841000,China)

机构地区：[1]中国石油塔里木油田公司油气运销事业部,新疆库尔勒841000

出　　处：《精细石油化工进展》2025年第1期48-54,共7页Advances in Fine Petrochemicals

摘　　要：全面掌握海底腐蚀管道极限承载力的情况有利于指导该管道的安全运行。由于单一BP神经网络(BPNN)模型存在学习效率低、对初始权重敏感且容易陷入局部最优状态等缺点,故采用麻雀搜索算法(SSA)来优化BPNN的初始权值和阈值,建立SSA-BPNN组合模型预测极限承载力,并与BPNN模型、遗传算法优化的BPNN(GA-BPNN)模型和粒子群算法优化的BPNN(PSO-BPNN)模型进行对比。结果显示:SSA-BPNN模型的平均相对误差为1.2693%,远远好于其他模型;SSA-BPNN模型的预测结果与有限元法得到的结果进行线性拟合后与直线Y=X最为贴近,其决定系数为0.99948,说明SSA-BPNN模型是一种准确性高且稳定性良好的海底腐蚀管道极限承载力预测工具。A comprehensive understanding of the ultimate bearing capacity of submarine corroded pipelines is beneficial for guiding the safe operation of the pipeline.Due to the drawbacks of low learning efficiency,sensi-tivity to initial weights,and susceptibility to local optima in a single BP neural network(BPNN)model,the Sparrow Search Algorithm(SSA)was adopted to optimize the initial weights and thresholds of BPNN.A combi-nation model of SSA-BPNN was established for predicting ultimate bearing capacity.Then,the SSA-BPNN model was compared with BPNN model,genetic algorithm optimized BPNN(GA-BPNN)model and particle swarm optimization optimized BPNN(PSO-BPNN)model.The results showed that the average relative error of the SSA-BPNN model was 1.2693%,which was much better than other models.The prediction results of the SSA-BPNN model linearly fitted with the results obtained from the finite element method and found to be clos-est to the line Y=X,with a determination coefficient of 0.99948.This indicated that the SSA-BPNN model was a highly accurate and stable tool to predict the ultimate bearing capacity of submarine corroded pipelines.

关 键 词：海底腐蚀管道 极限承载力 有限元法 麻雀搜索算法(SSA) BP神经网络(BPNN) 

分 类 号：TE988.2[石油与天然气工程—石油机械设备]