摘要
为适应桁架结构健康监测技术的发展,进一步优化桁架结构损伤程度预测效果,提出一种基于随机森林的桁架结构损伤程度预测算法。算法建立并优化桁架结构损伤数据库,对多个损伤信号分量提取频域特征,通过孤立森林算法和缺失森林算法对数据库进行异常检测和数据补全,并采用主成分分析法对其进行特征降维。经实验验证,算法能够针对不同损伤程度的桁架结构敏感度做出有效判定,与极端随机树、AdaBoost、Bagging等回归算法相比,均方误差显著降低,为桁架结构健康监测提供了有效的预测手段。
In order to adapt to the development of truss structure health monitoring technology and further optimize the prediction effect of truss structure damage degree, a prediction algorithm of truss structure damage degree based on random forest is proposed. The algorithm establishes and optimizes the damage database of truss structure, extracts frequency domain features from multiple damage signal components, detects anomalies and completes data in the database by isolation forest algorithm and missing forest algorithm, and reduces the dimension of features by principal component analysis.Experiments show that the algorithm can effectively judge the sensitivity of truss structures with different damage degrees. Compared with regression algorithms such as extreme random tree, AdaBoost, Bagging,etc., the mean square error is significantly reduced, which provides an effective prediction method for the health monitoring of truss structures.
作者
姜璐
吕瑞宏
赵艺伟
JIANG Lu;LYU Ruihong;ZHAO Yiwei(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)
出处
《微处理机》
2022年第1期43-47,共5页
Microprocessors
关键词
随机森林
桁架损伤程度预测
孤立森林
缺失森林
主成分分析
Random forest
Prediction of truss damage degree
Isolation forest
MissForest
Principal component analysis