摘要
传统的单通道信号分析容易造成信息缺失和诊断结论不一致等问题,这些问题可由全矢谱分析技术来解决。动态支持向量数据描述算法是对传统支持向量数据描述的改进算法,它的分类边界随着被测样本数的不断增加而不断更新,具有自学习能力。将全矢谱分析技术与动态支持向量数据描述算法相结合而提出全矢谱动态支持向量数据描述(vector spectrum dynamic support vector data description,VSDSVDD)的故障诊断新方法。运用全矢谱技术对数据进行处理,并提取特征矢量,作为VSDSVDD的输入参数,建立起分类模型即可以对机器运行状态进行分类。实验表明,该方法具有很好的分类准确性。
Traditional signal analysis methods based on single channel can result some mistakes, for example some of the information is lost or the diagnosis results are conflicting with the same signal, and these problems can be solved by means of full vector spectrum signal analysis method. Dynamic support vector data description algorithm is the amelioration algorithm of the traditional support vector data description and its classification boundary is updated constantly with the increasing number of tested samples, and it has the ability of self-learning. A new fault diagnosis method named vector spectrum dynamic support vector data description（VSDSVDD） is put forward, in this method the vector spectrum analysis and the dynamic support vector data description are combined. By using the vector spectrum method to process the signals and extract the characteristic vector to be used as the input parameters of VSDSVDD, the classification model is set up and therefore the running state of the machines can also be classified. Experiments show that this method has more class veracity.
出处
《机械强度》
CAS
CSCD
北大核心
2013年第2期152-155,共4页
Journal of Mechanical Strength
基金
国家自然科学基金(50675209)
河南省自然科学基金(0611022400)
河南省高等学校精密制造技术与工程重点学科开放实验室开放基金资助项目
关键词
全矢谱
特征提取
故障诊断
动态支持向量数据描述
Full vector spectrum
Feature extraction
Fault diagnosis
Dynamic support vector data description （ DSVDD ）
