摘要
柱塞泵是工程机械的关键部件,其性能好坏将直接影响整个设备的正常工作。针对柱塞泵提出基于特征选择支持向量机的智能诊断方法。对采集的振动信号基于小波包分解提取能量特征,然后利用Fisher准则函数选择对智能诊断最有利的特征,利用支持向量机进行训练,并将每个二类支持向量机按二叉树的组织形式构成系统的诊断模型。以汽车起重机柱塞泵为研究对象,其6种故障形式,包括正常、轴承内圈故障、滚动体故障、柱塞故障、配流盘故障、斜盘故障,用于检验所提算法的诊断能力,并与传统的BP神经网络和最近的蚁群神经网络方法进行对比。诊断结果表明:所提出的算法优于另外两种方法,具有较好的诊断效果。
In truck crane, the plunger pump is the key equipment, and the quality of the pump affects directly the performance of whole mechanical system. A novel intelligent diagnosis method based on features selection and support vector machine （SVM） was proposed for plunger pump in truck crane. Based on the wavelet packet decompose, the wavelet packet energy was extracted from the original vibration signal to represent the condition of equipment. Then, the Fisher criterion was utilized to select the most suitable fea- tures for diagnosis. Finally, each two-class SVM with binary tree architecture was trained to recognize the condition of mechanism. The proposed method was employed in the diagnosis of plunger pump in truck crane. The six states, including normal state, bearing inner race fault, bearing roller fault, plunger fault, thrust plate wear fault, and swash plate wear fault, were used to test the classification performance of the proposed Fisher-SVMs model, which was compared with the classical and the latest models, such as BP ANN, ANT ANN, respectively. The experimental results show that the Fisher-SVMs is superior to the other two models, and gets a promising resuit.
出处
《机床与液压》
北大核心
2013年第19期164-168,147,共6页
Machine Tool & Hydraulics
基金
国家自然科学基金资助项目(51205371)
上海科技创新行动计划资助项目(11dz1121500
11JC1405800)
机械系统与振动国家重点实验室资助项目(MSV-2012-06)
