首页|基于集成算法的腐蚀管道失效压力预测研究

基于集成算法的腐蚀管道失效压力预测研究

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
为了提高腐蚀管道剩余强度的预测精度、解决单一预测模型易受训练数据的质量影响而发生运行及预测输出不稳定的问题,引入两种集成模型方法.首先对于串行结构集成方法,以支持向量回归(SVR)融合正余弦策略改进的黑猩猩优化算法(IChOA)为基础建立AdaBoost-IChOA-SVR模型;其次对于双层并行结构方法,根据预测问题筛选出相关性低且学习效果良好的预测算法作为第一层基预测器,调节新数据集形成方式及相关参数设置,建立Stacking堆叠集成模型.以含腐蚀缺陷管道失效压力爆破数据为例,利用MATLAB分别进行仿真模拟,与基础SVR和PSO-ELM模型的预测结果及评价指标进行对比分析.研究结果表明:集成预测模型具有更好的预测输出性能,且串行结构的AdaBoost集成学习模型的构造流程较为简洁,运行速度及精度更高;该模型对腐蚀缺陷管道失效压力预测问题的拟合度可达0.996,相对误差均值可达3.69%,可为后续腐蚀管道相关预测模型建立和防护维修策略制定提供参考.
Research on Failure Pressure Prediction of Corroded Pipeline Based on Integrated Algorithm
To improve the prediction accuracy of the residual strength of corroded pipelines and solve the problem that the relevant single prediction model is susceptible to the quality of training data and its operation and prediction output are unstable,two integrated model methods are introduced.Firstly,for the serial structure integration method,the AdaBoost-IChOA-SVR model is established based on the support vector regression(SVR)and sine cosine strategy improved chimpanzee optimization algorithm(IChOA).Secondly,for the two-layer parallel structure method,the prediction algorithm with low correlation and good learning effect is selected as the first layer base predictor according to the prediction problem,and the new data set formation method and related parameter settings are adjusted to establish the Stacking stacked ensemble model.Taking the failure pressure blasting data of corrosion defect pipeline as an example,the simulation is carried out by MATLAB,and the prediction results and evaluation indexes of the basic SVR and PSO-ELM models are compared and analyzed.The results show that the integrated prediction model has better prediction output performance,and the AdaBoost integrated learning model with serial structure has simpler construction process,higher running speed and accuracy.The model has a fitting degree of 0.996 and a mean relative error of 3.69%for the failure pressure prediction of corroded pipeline,which can provide ref-erence for the subsequent establishment of related prediction models and protection and maintenance strategies for corroded pipelines.

safety engineering science and technologyintegrated modelcorrosion pipeline failure pressureAdaBoost ensemble learningStackingchimpanzee optimization algorithm

骆正山、张佳琦、骆济豪

展开 >

西安建筑科技大学 管理学院,陕西 西安 710055

北京理工大学睿信学院,北京 102488

安全工程科学技术 集成模型 腐蚀管道失效压力 AdaBoost集成学习 Stacking 黑猩猩优化算法

国家自然科学基金项目

41877527

2024

10.20165/j.cnki.ISSN1673-629X.2024.0044

计算机技术与发展
陕西省计算机学会

计算机技术与发展

CSTPCD
影响因子：0.621
ISSN：1673-629X
年,卷(期)：2024.34(5)
  • 23