改进离散熵在列车轴承损伤检测中的应用

Application of Improved Dispersion Entropy to Fault Detection of Axle‑Box Bearing in Train

针对轴箱轴承早期损伤的检测问题,提出一种基于改进多尺度离散熵算法(improved multiscale dispersion entropy,简称IMDE)和支持向量机的诊断模型,通过提取振动信号中的关键信息有效识别轴承的健康状态。首先,考虑传统多尺度离散熵(multiscale dispersion entropy,简称MDE)因数据点重合和粗粒化尺度不断增大而引起的熵值误差增加、分布混乱及波动明显等缺陷,通过对粗粒化过程和离散熵的优化改进算法;其次,结合实际算例,针对高速列车轴箱轴承在不同运行状态下的振动数据进行试验验证。结果表明,相较于MDE,IMDE计算熵值的误差更小,鲁棒性更好,且支持向量机分类结果显示IMDE取得了更高的诊断精度。

Aiming at the problem of early damage detection of axle box bearings,a diagnosis model based on improved multiscale dispersion entropy(IMDE)and support vector machine is proposed to effectively identify the health state of bearings by extracting key information from vibration signals.This method mainly considers the shortcomings of traditional multi-scale dispersion entropy(MDE),such as the increase of entropy standard deviation,the confusion of distribution and the obvious fluctuation caused by the problem of data point coincidence and the increase of coarse-graining scale.The algorithm is improved by optimizing the coarse-graining procedure and the dispersion entropy.Then,a practical example is given to verify the vibration data of high-speed train axle box bearing under different operating conditions by experiments.The results show that the standard deviation of IMDE is smaller and the robustness is better than MDE,and the classification results of support vector machine show that IMDE has higher diagnosis accuracy.