摘要
针对往复泵泵阀冲击振动信号的非平稳特性,运用谐波小波包能量特征提取和最小二乘支持向量机(LSSVM)相结合的方法识别泵阀故障。通过对泵阀冲击振动信号进行谐波小波包分解,提取各频段谐波小波包系数的能量值,将各频段能量值组成的特征向量作为诊断模型的特征向量,输入到LS-SVM多类分类器中进行故障识别,并将谐波小波包与小波包在泵阀故障诊断中的诊断准确率进行了比较。试验结果表明将谐波小波包分解和LS-SVM相结合可以准确有效地识别泵阀故障类型。
Considering the non-stationary characteristics of the shock and vibration signal measured on reciprocating pump valve,the harmonic wavelet packet energy feature extraction and the least squares support vector machine（LS-SVM） are applied to identify the pump valve fault. A characteristic vector is constructed by the energy of different frequency segment,which is calculated by using the coefficient of harmonic wavelet package transform of the shock and vibration signals and input to the LS-SVM multi-classifier to recognize failures. Harmonic wavelet packet is compared with wavelet packet in fault diagnosis accuracy of pump valve. The experimental results show that this method can identify pump valve fault patterns accurately and effectively.
出处
《机械设计与制造》
北大核心
2015年第5期241-245,共5页
Machinery Design & Manufacture
基金
国家自然科学基金项目(51175051)
