基于KPCA-BAS-GRNN的埋地管道外腐蚀速率预测

目的 提高埋地管道外腐蚀速率的预测精度。方法 建立基于核主成分分析法（KPCA）和天牛须搜索（BAS）算法优化的广义回归神经网络（GRNN）腐蚀速率预测模型，通过KPCA对原始数据进行预处理，提取影响管道外腐蚀的主要因素，应用GRNN建立埋地管道外腐蚀速率预测的数学模型，并采用BAS算法对模型进行优化，减小了人为设置参数的影响。以川气东送埋地管段为例，分析选取出12种关键影响因素，建立了埋地管道外腐蚀指标体系，借助MATLAB-R2014a编写程序进行仿真，并与实际值进行对比。结果 模型的预测结果与实际值基本一致，KPCA可有效降低指标体系的维度，提取出包含原始信息97.9%的3个主因素—土壤电阻率、氧化还原电位、氯离子含量，简化了运算过程。采用的BAS-GRNN模型将预测精度提高到7.83%以内，平均相对误差5.21%，决定系数取值0.93。与其他模型相比，该模型性能较好，预测精度更高。结论 采用KPCA提取的主要影响因素符合工程实际，建立的BAS-GRNN模型预测精度高，有较好的适应性，为埋地管道外腐蚀速率预测提供了新思路，对管道的维护更新工作提供了参考依据。

Prediction of External Corrosion Rate of Buried Pipeline Based on KPCA-BAS-GRNN

The work aims to improve the prediction accuracy of the external corrosion rate of buried pipeline. The corrosion rate prediction model of buried pipeline was established based on kernel principal component analysis (KPCA) and the general regression neural network (GRNN) optimized by Beetle antennae search (BAS) algorithm. The main factors affecting external corrosion of buried pipeline were extracted by preprocessing the original data through KPCA. GRNN was used to build a mathematical model to predict the external corrosion rate of buried pipeline and BAS algorithm was adopted to optimize the model to reduce the effects of artificially set parameters. In addition, the pipelines buried in natural gas transmission project from Sichuan to East were utilized as an example to analyze 12 key influencing factors and establish the external corrosion index system of buried pipeline. MATLAB-R2014a software was used for simulation processing, and compared with the actual values. The predicted results of the model were basically consistent with the actual values. KPCA could effectively reduce the dimensions of the indicator system and extract three main factors with 97.9% original information, including soil resistivity, oxidation-reduction potential and Cl? content. Thus, the calculation process was simplified. BAS-GRNN model was adopted to improve the prediction accuracy to 7.83%. The average relative error was 5.21%, and the determination coefficient was 0.93. Compared with other models, this model had better performance and higher prediction accuracy. Thus, the main influencing factors extracted by KPC Accord with engineering practice. BAS-GRNN model provides a new idea for the prediction of external corrosion rate of buried pipeline and a reference basis for the maintenance and updating of buried pipeline by higher precision and better adaptability.