摘要
针对陀螺仪输出样本不平衡和噪声干扰大的特点,提出一种使用DSP实现非平衡FSVM陀螺仪故障诊断方法。系统核心算法FSVM的隶属度函数设计由非平衡特征因子和去噪模糊因子两部分组成,用于解决样本不平衡和噪声干扰大导致SVM分类精度降低的问题。首先,FSVM的训练阶段在上位机上实现,采集陀螺仪的无故障和故障信号,经过小波包特征提取后得到训练样本,然后由FSVM训练得到分类识别函数,最后将分类识别函数移植到DSP硬件平台上实现FSVM的测试应用。速率陀螺故障诊断的实验结果表明,该实现方法不仅提高了故障样本的诊断精度,而且满足系统实时性要求,具有一定的实用价值。
According to the features of unbalanced samples and high noise from gyro outputsignals,an unbalanced FSVM fault diagnosis method of gyro is presented based on DSP.The membership function of FSVM algorithm of system consists of unbalanced characteristicfactor and de-noising fuzzy factor, which is used to solve the classification accuracy reduc-ing problem of SVM resulting from unbalanced samples and high noise. Firstly, the trainingphase of FSVM is implemented in host computer. The normal and fault signals of gyro arecollected, and then the samples are obtained through the method of wavelet package featureextraction. Then, the classification function is transplanted, which is obtained by FSVMtraining model for DSP to achieve the test and recognition phase. The fault diagnosis experi-ment results of rate gyro shows that the implementation means not only improves the diag-nostic accuracy of fault samples, but also meets the real-time requirements. So it has a cer-tain practical value.
出处
《沈阳理工大学学报》
CAS
2015年第1期46-51,共6页
Journal of Shenyang Ligong University