基于加权最小二乘支持向量机改进算法的汽轮机通流部分故障诊断研究

汽轮机通流部分故障特征数据较多、故障类型复杂,很难建立精确的机理模型。提出一种基于加权最小二乘支持向量机(Weighted Least Squares Support Vector Machines,WLS-SVM)的改进算法,该算法用输出变量的留一交叉检验误差取代原有误差确定加权系数,解决了WLS-SVM由于加权系数与模型支持值相互影响,样本在剔除与不剔除之间反复变化而不收敛的问题。实验结果表明该方法能有效地剔除异常样本,减少故障特征量的数目,提高了校正模型的稳健性及WLS-SVM特征预测的速度和预测的精度。

Fault Diagnosis and Research About Flow Passage of Steam Turbine Based on an Improved Algorithm of Weighted Least Squares Support Vector Machines

Flow passage of steam turbine have lots of fault features,types of fault are complex,It’s difficult to establish the precise mechanism model.This paper provides an improved algorithm based on weighted least squares support vector machines(WLS-SVM),this improved algorithm use leave one out cross validation(LOOCV)deviation of output variable replace original deviation to determine weighted coefficient.resolved the problem that sample repeated changes and can not convergence because of the influence between weighted coefficient and support value of model.The experiment results show that this method can effectively remove abnormal samples,reduce the number of fault features.improve the robustness of calibration models and the speed and accuracy about features prediction.