摘要
电机的可视化离不开各类传感器对电机运行信号的监测。为了进一步提高电机传感源数据网络检测对智能化影响的能力,在弥补支持向量机(SVM)训练时间过长问题的基础上,设计了一种深度信念网络(DBN)优化1/4超球面支持向量机QSSVM网络,实现电机传感源在线测试功能的异常检测。研究结果表明:当窗口扩大后,QSSVM准确度不断提高。随着抽样维数的增加,检测率发生一定程度的提高,QSSVM有良好的探测效果,实现94.16%的检测率,F保持较小召回比率。该研究有助于提高电机传感源数据异常检测能力,具有很好的实际推广价值。
The visualization of the electric motor is inseparable from the monitoring of the motor running signal by various sensors.In order to further improve the influence ability of motor sensing source data network detection on the intelligence,a Deep Belief Network(DBN)on the basis of making up for the long training time of Support Vector Machine(SVM)was designed to optimize 1/4 hyperspherical support vector machine QSSVM network,which has realized the abnormal detection of online testing function of motor sensing source.The research results show that the accuracy of QSSVM is improved continuously when the window is enlarged.With the increase of sampling dimension,the detection rate increases to a certain extent.QSSVM has a good detection effect,which achieves a detection rate of 94.16%and a small recall ratio of F.This research is helpful to improve the abnormal detection ability of motor sensing source data,and has a good practical promotional value.
作者
董军军
周军
Dong Junjun;Zhou Jun(College of Information Engineering,Jiaozuo University,Jiaozuo 454000,China;College of Artificial Intelligence,Jiaozuo University,Jiaozuo 454000,China)
出处
《防爆电机》
2025年第5期14-17,共4页
Explosion-proof Electric Machine
基金
河南省科技厅科技攻关项目(182102210508)。
关键词
异常检测
超球面支持向量机
深度信念网络
电机传感源
Abnormal detection
hyperspherical support vector machine
deep belief network
motor sensing source
