摘要
为了提高滚动轴承故障类型诊断准确度,提出了磷虾算法优化多分类支持向量机的轴承故障诊断方法。对于时频域特征参数的提取,将CEEMD算法与小波包优势结合,提出了CEEMD与小波包半软阈值去噪相结合的提取方法;对于特征参数降维,针对轴承振动信号的非线性特点,使用局部线性嵌入算法降维,对降维后特征参数使用模糊C均值聚类进行验证,可以看出LLE降维不仅降低了计算量而且有利于模式识别;将二叉树法与投票法支持向量机结合,给出了混合多分类支持向量机,使用磷虾算法对其进行参数优化。实验验证可知,磷虾算法优化的多分类支持向量机具有很高的输出精度,轴承状态识别准确率为100%,使用粒子群算法优化的支持向量机输出精度低,轴承状态识别准确率为79%。
To improve fault diagnosis accuracy degree of rolling bearing, fault diagnosis method based on multi- classification support vector machine optimized by krill algorithm is proposed. For extracting time and frequency domain parameters, combing advantages of CEEMD and wavelet packet, parameters extracting method integrating CEEMD and wavelet packet is put forward. For characteristic parameters dimension, considering nonlinearity of bearing vibration signal, LLE algorithm is used to reduce the dimension, clarified by fuzzy C-means clustering, dimensionality reduction not only can reduce computation, but can also benefit pattern recognition. Binary tree and voting method are integrated, so that a new mixed multi-classification SVM is given, and its parameters are optimized by krill algorithm. It can be seen through experiment, output accuracy of SVM optimized by krill algorithm is very high, and bearing state recognition accuracy degree is 100%. Output accuracy of SVM optimized by PSO algorithm is relatively low, and bearing state recognition accuracy degree is 79%.
作者
吕震宇
LV Zhenyu(Shandong Polytechnic, Jinan 250104, CHN)
出处
《制造技术与机床》
北大核心
2019年第5期130-136,共7页
Manufacturing Technology & Machine Tool
基金
省教育厅科研课题(KJ2018ZBB022)
