基于多尺度空间的直升机滚动轴承故障诊断

Fault Diagnosis of Helicopter Rolling Bearing Based on Multi-Scale Space

针对直升机系统与传递路径复杂,采集信号中成分多样,传统方式提取的特征难以有效反映信号健康状态,影响滚动轴承诊断精度等问题,在传统时域指标的基础上,结合多尺度空间对特征空间重叠和信号跨尺度复杂性问题上的优势,构建多尺度指标作为故障分类的依据。根据ReliefF算法对原始高维多尺度特征迭代计算得到权重,利用权重值进行特征选择,同时减轻计算成本。权重最大的一部分特征将作为随机森林模型的输入,利用其多分类器集成学习的优势,进行滚动轴承故障分类诊断。通过滚动轴承公开数据集来说明所提方法的优势和可行性。数据处理结果表明,多尺度特征较原始时域特征具有更好的分类性能,并且随机森林在该算法中较其他分类模型分类效果更好。

In order to solve the problems that the helicopter system and the transmission path are complex,the components in the collected signal are diverse,and the features extracted by traditional methods are difficult to reflect the health condition effectively,so that to affect the diagnosis accuracy,the traditional time domain indicators are combined with the multi-scale spatial pairing of features.Based on the advantages of spatial overlap and signal cross-scale complexity issues,multi-scale indicators are constructed as the basis for fault classification.According to the ReliefF algorithm,the original high-dimensional multi-scale features are calculated iteratively to obtain the weights,and the weight values are used for feature selection,and at the same time,the calculation cost is reduced.Part of the features with the largest weight will be used as the input of the random forest model,whose superiority of multi-classifier ensemble learning will be used to diagnose rolling bearing faults.The rolling bearing open data set is used to illustrate the superiority and feasibility of this proposed method.The data processing results prove that compared to the original time domain characteristics,the multi-scale feature has better classification ability,and random forest has better performance than other classification models in this algorithm.