Rotating machinery plays an increasingly crucial role in mechanical systems. For its normal operation, a novel fault diagnosis method is proposed in this paper, using composite multiscale fuzzy distribution entropy (CMFDE) and minimal error of convex hull approximation (MECHA). In this paper, CMFDE is utilized to extract essential information and measure time series complexity for vibration signals. Results indicate the CMFDE has less information loss and better stability. Then, to fulfill the classification tasks, the first several main features obtained by principal components analysis are fed into the proposed MECHA-based classifier. Results show MECHA has better classification performance. Using the laboratory data, we validate the feasibility and superiority of the proposed fault diagnosis method through two cases consisting of different fault types or fault severity degrees.

旋转机械在机械系统中起着越来越重要的作用。针对其正常运行，提出了一种新的故障诊断方法，该方法利用复合多尺度模糊分布熵（CMFDE）和凸包壳近似（MECHA）的最小误差。在本文中，CMFDE用于提取基本信息并测量振动信号的时间序列复杂度。结果表明，CMFDE具有较少的信息丢失和更好的稳定性。然后，为了完成分类任务，将通过主成分分析获得的前几个主要特征输入到提出的基于MECHA的分类器中。结果表明，MECHA具有更好的分类性能。利用实验室数据，