摘要
多尺度形态滤波器中如何准确确定分析尺度一直是一个没有得到很好解决的问题,但这一问题又非常重要,因为尺度选取的差异会直接影响滤波结果。为此,提出一种变尺度形态滤波器。在滤波过程中,结构元素的长度和高度不再是固定的,它可以根据待分析信号的几何形貌特征作自适应的改变,使得结构元素与原始信号更为匹配,从而避免多尺度滤波中尺度选取这一难题。为了进一步提升滤波精度,构造一种新的结构元素,并与传统的扁平结构元素进行比较。将所提出的方法应用于在万向轴不平衡试验台上采集到的试验数据分析中,研究结果表明,变尺度形态滤波方法能有效辨识万向轴动不平衡引起的基频和倍频故障特征,与多尺度形态滤波方法相比,变尺度形态滤波更能突出故障特征。
How to accurately determine the filter scale is still an unsolved problem for multi-scale morphology filter. However, this problem is of great importance that the filtering accuracy is directly influenced by the filtering scale. A time-varying scale morphology filter is proposed. The length and height of a structure element are on longer fixed in the filtering process. The structure element can adjust adaptively according to the geometric shape of a signal being analyzed, which makes it could match the raw signal more closely. Therefore, the scale selection problem in multi-scale morphology filter is avoided. To improve the filtering accuracy further, a new type of structure element is constructed and compared with the traditional flat structure element. The proposed method is then applied to the experimental data collected from an unbalance test rig of cardan shaft. The research results show that the proposed time-varying scale morphology filter is effective to identify the fundamental frequency and multiple frequencies caused by the unbalance of cardan shaft. Comparing with the multi-scale morphological filter, time-varying scale morphology filter is more capable to manifest fault features.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2018年第4期278-284,共7页
Journal of Mechanical Engineering
基金
国家自然科学基金(51705431,51375403)
国家重点研发计划(2016YFB1200602-17)
中国博士后科学基金(2017M612935)资助项目
关键词
变尺度
多尺度
形态滤波
结构元素
time-varying scale
multi-scale
morphology filter
structure element
