基于KPCA-CSO-RVM模型的工艺管道腐蚀速率预测

Prediction of Corrosion Rate of Process Pipelines Based on KPCA-CSO-RVM Model

针对石油化工企业内工艺管道的腐蚀问题,提出一种基于核主成分分析(KPCA)算法、鸡群优化(CSO)算法、相关向量机(RVM)算法组合模型的工艺管道腐蚀速率预测方法,即KPCA-CSO-RVM组合模型。该方法首先对KPCA算法、CSO算法、RVM算法进行理论介绍,提出算法的组合方法;然后使用KPCA算法对工艺管道腐蚀速率的影响因素进行优选,使用CSO算法对RVM算法中的参数进行优选;最后采用某石油化工企业工艺管道腐蚀速率及其相关影响因素数据对KPCA-CSO-RVM组合模型进行训练,对部分工艺管道腐蚀速率进行预测,并通过计算预测误差来验证KPCA-CSO-RVM组合模型的预测效果。结果表明:基于KPCA-CSO-RVM组合模型的工艺管道腐蚀速率预测结果的最大预测误差不超过4%,平均相对误差仅为2.03%,均方根误差为0.0035,其预测效果优于其他预测模型,证明KPCA-CSO-RVM组合模型在工艺管道腐蚀速率预测方面具有可行性和先进性。

Aiming at the corrosion problem of process pipelines in petrochemical enterprises,this paper proposes a corrosion rate prediction method based on a combination model of Kernal Principal Component Analysis(KPCA),Chicken Swarm Optimization(CSO),and Relevance Vector Machine(RVM).Firstly,a theoretical introduction to the kernel principal component analysis algorithm,chicken flock optimization algorithm,and correlation vector machine algorithm is given,and a combination method of the algorithm is proposed.The nuclear principal component analysis algorithm is used to optimize the factors affecting the corrosion rate of process pipelines,and the chicken flock optimization algorithm is used to optimize the parameters of the correlation vector machine.Using a petrochemical enterprise’s process pipeline corrosion rate and related influencing factors data to train the combined model,predict the corrosion rate of some process pipelines,and calculate the prediction error to verify the prediction effect of the combined model proposed in this study.The research shows that the maximum prediction error of the corrosion rate prediction model of process pipelines based on the KPCA-CSO-RVM combined model does not exceed 4%,the average relative error is only 2.03%,and the root mean square error is 0.0035.The prediction results are better than those of other prediction models,which proves feasibility and advancement of the combined KPCA-CSO-RVM model in predicting the corrosion rate of process pipelines.