基于最小二乘支持向量机的时间序列模型及油气管线点腐蚀损伤行为的预测

he pitting corrosion are usually existed in oil and gas pipelines, and the corrosion is easily to cause perforation or leak accident of the pipeline. Therefore, it is very significant to study the prediction of depth propagation behaviour. In this paper, based on the Least square support vector machine(LS-SVM), a time series analysis model is proposed to predict the pitting corrosion depth for oil and gas pipeline. As for the complex situation of parameter optimization process in LS-SVM model, a new Hybrid particle swarm optimization(HPSO) is presented by inducing of uniform design method, novel self-adaptive inertia weight and novel velocity updating strategy to improve the global optimization ability. The simulation experiment indicates that HPSO has more stronger search capability and more higher convergence precision. The results of a engineering case application show that LS-SVM model performs a better prediction performance compared with GM(1,1) model, auto-regrssion(AR) model and BP neural network model.