基于特征选择的半潜式平台故障信号探究

Research on Fault Signal of Semi-submersible Platform Based on Feature Selection

“蓝鲸2号”第七代深水半潜式钻井平台,其工作环境恶劣且远离港岸,故保障平台的平稳运行和安全是重中之重。该平台电力系统警报信号特征种类众多、特征重要程度模糊,仅利用单分类器方法无法准确划分故障警报信号,因此,引入集成学习算法并结合特征选择技术,提出基于支持向量机递归特征消除(SVM-RFE)的Bagging-AdaBoost分类模型(SRBA),用于解决多特征分类问题。结果显示提出的SRBA集成学习算法综合分类正确率达96%,在分类精度上优于Bag⁃ging、AdaBoost、Bagging-AdaBoost分类器对比模型,该方法具有较高的稳定性和分类准确度,是一种更为有效的分类手段。

The seventh-generation deep-water semi-submersible drilling platform of"Blue Whale 2"has a bad working envi⁃ronment and is far away from the port shore,so ensuring the smooth operation and safety of the platform is the top priority.The plat⁃form has many kinds of features and fuzzy importance of power system alarm signals.Only using single classifier method can not ac⁃curately classify fault alarm signals.Therefore,an integrated learning algorithm and feature selection technology are introduced to propose a Bagging-AdaBoost classification model based on Support Vector Machine Recursive Feature Elimination(SVM-RFE)to solve the classification problem with multiple features(SRBA).The results show that the comprehensive classification accuracy of the proposed SRBA ensemble learning algorithm reaches 96%,which outperforms the Bagging,AdaBoost,Bagging-AdaBoost clas⁃sifier comparison models in classification accuracy.It shows that this method has high stability and classification accuracy,and is a more effective classification method.