改进SSA-KELM模型在埋地腐蚀管道剩余寿命预测中的应用

Application of Improved SSA-KELM Model in Remaining Life Prediction of Buried Corrosion Pipeline

为提高埋地腐蚀管道剩余寿命预测精度,构建其剩余寿命预测模型。建立基于核主成分分析(KPCA)和改进麻雀搜索算法(ISSA)的核极限学习机(KELM)剩余寿命预测模型。首先采用KPCA预处理原始数据,提取埋地腐蚀管道主要特征向量并重构评价指标。其次针对SSA易陷入局部最优及迭代后期抗停滞性能降低等缺陷,提出SSA改进方案:利用Tent混沌提升其遍历性;引入自适应安全值调整麻雀搜索区域;使用高斯扰动重点搜索最优解附近区域,以提升SSA全局寻优能力。再次利用ISSA寻优KELM中核参数和惩罚系数,最终构建KPCA-ISSA-KELM埋地腐蚀管道剩余寿命预测模型。以某埋地管线为例进行仿真,结果表明:KPCA-ISSA-KELM模型预测结果均方误差、平均绝对误差值、决定系数为分别为0.249、0.096、0.998,均优于其他模型。证明KPCA-ISSA-KELM的埋地腐蚀管道剩余寿命预测模型具有较强的鲁棒性,为管道系统研究提供重要的参考依据。

In order to improve the residual life prediction accuracy of buried corrosion pipelines,a residual life prediction model was constructed.A residual life prediction model of Kernel Extreme Learning Machine(KELM)based on Kernel Principal Component Analysis(KPCA)and Improved Sparrow Search Algorithm(ISSA)was established.Firstly,KPCA was used to preprocess the original data,and the main feature vectors of buried corrosion pipelines were extracted and the evaluation indicators were reconstructed.Secondly,in view of the defects of SSA easily falling into local optimum and reducing the anti-stagnation performance in the later iteration,an improved SSA scheme was proposed:using Tent chaos to improve its ergodicity;introduce adaptive security value to adjust the sparrow search area;using Gaussian disturbance to focus on searching near the optimal solution region to improve the global optimization capability of SSA.The kernel parameters and penalty coefficients in KELM were optimized by ISSA again,and the KPCA-ISSA-KELM buried corrosion pipeline residual life prediction model was finally constructed.Taking a buried pipeline as an example,the simulation results show that the mean square error,mean absolute error value and coefficient of determination of the prediction results of the KPCA-ISSA-KELM model is 0.249,0.096,and 0.998,respectively,which are better than other models.It is proved that KPCA-ISSA-KELM's residual life prediction model of buried corrosion pipeline has strong robustness,which can provide an important reference for pipeline system research.