统计物理学
熵(时间箭头)
计算机科学
符号动力学
人工智能
数学
物理
热力学
纯数学
作者
Ao Shen,Yongbo Li,Khandaker Noman,Dong Wang,Zhike Peng,Ke Feng
标识
DOI:10.1177/14759217241237717
摘要
Health monitoring has garnered significant and increasing attention from the research community and industrial practices thanks to its critical role in ensuring the safe operation of machinery and maintenance schedule. With regard to this, this paper introduces a novel diagnostic approach called fluctuation-based symbolic dynamic entropy (FSDE), which can enhance noise immunity and computational efficiency by utilizing fluctuation-based entropy algorithm and symbolic dynamic filtering. Moreover, the noise immunity and computational efficiency of FSDE, permutation entropy, fuzzy entropy, and sample entropy are compared by simulation signals. The simulation results show that FSDE has a more robust anti-noise performance and higher computational efficiency than the other three entropy methods. In order to extract fault features more thoroughly, multiscale analysis is applied to the entropy method, and multiscale FSDE (MFSDE) is proposed. MFSDE divides the measurement data into several scale series by coarse-grained technology, and then, FSDE is used to process each scale series separately. A series of experiments verify the effectiveness of MFSDE in fault feature extraction of fault diagnosis. Furthermore, the experimental results substantiate that MFSDE outperforms the other three currently used entropy-based approaches in terms of accuracy in classifying different health conditions of transmission systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI