基于二元多尺度熵的滚动轴承退化趋势预测

发布时间：2018-02-01 20:55

本文关键词： 滚动轴承多尺度熵参数优化退化趋势预测极限学习机　出处：《中国机械工程》2017年20期 　论文类型：期刊论文

【摘要】：针对轴承振动信号随机噪声干扰大、多尺度熵表征轴承退化趋势偏差大的问题,提出了一种基于二元多尺度熵的滚动轴承退化趋势预测方法。首先对滚动轴承振动信号进行局部特征尺度分解,采用多元多尺度熵理论对二阶信号进行计算,提取了二元多尺度熵特征。然后采用互信息法和假近邻法对算法中的嵌入维数和延迟向量等参数进行了优化。最后采用极限学习机预测模型对二元多尺度熵退化趋势曲线进行预测,并对比了不同激活函数的预测性能。结果表明,相对于传统多尺度熵,二元多尺度熵偏差较小;激活函数为sigmoid时极限学习机模型预测精确度较高。
[Abstract]:In order to solve the problem that the random noise disturbance of bearing vibration signal is large and the multi-scale entropy is used to characterize the large deviation of bearing degradation trend. A prediction method of rolling bearing degradation trend based on binary multi-scale entropy is proposed. Firstly, the vibration signal of rolling bearing is decomposed with local characteristic scale, and the second-order signal is calculated by multivariate multi-scale entropy theory. The binary multi-scale entropy feature is extracted, and the embedding dimension and delay vector in the algorithm are optimized by mutual information method and pseudo-nearest neighbor method. Finally, the prediction model of extreme learning machine is used to degenerate the binary multi-scale entropy. Forecast the trend curve. The prediction performance of different activation functions is compared. The results show that the deviation of binary multi-scale entropy is smaller than that of traditional multi-scale entropy. When the activation function is sigmoid, the prediction accuracy of LLM model is high.
【作者单位】：陆军工程大学石家庄校区导弹工程系;陆军工程大学科研学术处;
【基金】：国家自然科学基金资助项目(51541506)
【分类号】：TH133.33
【正文快照】： 0引言在旋转机械系统状态检修、故障预测与健康管理领域,滚动轴承性能退化评估和剩余使用寿命预测是研究的热点与难点,该研究内容的关键在于退化特征提取[1]。在故障模式识别研究中,故障特征仅需区分不同故障差异即可,而滚动轴承退化趋势预测和性能退化评估均要求退化特征对滚

【相似文献】