Scholarlink

Purpose <p>In response to challenges in accurately predicting the operational states of rotating machinery, an innovative anomaly detection model is presented in this paper.</p> Methods <p>The proposed sequence holospectrum conditional probability density autoencoder (SHCPDAE) model fetch spatiotemporal information in vibration signals by utilizing the data reconstruction capabilities of long short-term memory (LSTM) autoencoder in latent space. Simultaneously, the probability density distribution of sequence holospectrum parameters is calculated by constructing parameterized estimator combined with autoregressive algorithm.</p> Conclusions <p>The experimental results demonstrate that the proposed SHCPDA method outperforms traditional approaches in detecting abnormal states of rotating machinery, showcasing improved accuracy and robust generalization capabilities. Furthermore, the integration of holospectrum technology in this method proves to be an effective strategy for enhancing the interpretability of equipment health monitoring models, which holds promise for advancing the field of rotating machinery condition monitoring.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sequence Holospectrum Conditional Probability Density Autoencoder for Machine Anomaly Detection

  • 期刊论文

  • 发表日期：

    2025年4月10日

查看全文
  • Xin Huang,
  • Xiaojin Liu,
  • Wenwu Chen,
  • Dingrong Qu,
  • Huajin Shao,
  • Weiya Zhang,
  • Long Pan

摘要

Purpose

In response to challenges in accurately predicting the operational states of rotating machinery, an innovative anomaly detection model is presented in this paper.

Methods

The proposed sequence holospectrum conditional probability density autoencoder (SHCPDAE) model fetch spatiotemporal information in vibration signals by utilizing the data reconstruction capabilities of long short-term memory (LSTM) autoencoder in latent space. Simultaneously, the probability density distribution of sequence holospectrum parameters is calculated by constructing parameterized estimator combined with autoregressive algorithm.

Conclusions

The experimental results demonstrate that the proposed SHCPDA method outperforms traditional approaches in detecting abnormal states of rotating machinery, showcasing improved accuracy and robust generalization capabilities. Furthermore, the integration of holospectrum technology in this method proves to be an effective strategy for enhancing the interpretability of equipment health monitoring models, which holds promise for advancing the field of rotating machinery condition monitoring.