Scholarlink

<p>As intelligent manufacturing scales with growth in monitored machine tools and sensor deployments, vibration data volumes have surged dramatically. Cloud-based centralized condition monitoring faces challenges such as delayed maintenance decision-making and risk of data privacy leakage during high data throughput. Edge computing offers an effective solution to address these challenges. This paper proposes an adaptive perceptual vibration hashing method for edge-based condition monitoring, aiming to reduce data dimensionality and preserve machine state information while ensuring data privacy and security. First, wavelet packet decomposition and two-dimensional discrete cosine transform extract high-dimensional sub-band features representing machine conditions. Then, a one-way irreversible adaptive hashing coding strategy integrating dictionary learning and symbolic aggregate approximation is proposed, further compressing the sub-band features into a fixed-length machine condition hash with minimal symbols for online monitoring. It transforms kilobyte-long vibration signals into few-byte symbol sequence, significantly improving storage and transmission efficiency at edge terminals. The effectiveness of the proposed method is validated with tool wear and bearing fault datasets. Results demonstrate that the proposed method achieves a good performance in both degradation trend evaluation and fault identification with data privacy preserving.</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

Adaptive perceptual vibration hashing for real-time and privacy-preserving condition monitoring

  • 期刊论文

  • 发表日期：

    2026年7月6日

查看全文
  • Shiwei Yan,
  • Fengli Zhang,
  • Tiantian Liu,
  • Jinjiang Wang

摘要

As intelligent manufacturing scales with growth in monitored machine tools and sensor deployments, vibration data volumes have surged dramatically. Cloud-based centralized condition monitoring faces challenges such as delayed maintenance decision-making and risk of data privacy leakage during high data throughput. Edge computing offers an effective solution to address these challenges. This paper proposes an adaptive perceptual vibration hashing method for edge-based condition monitoring, aiming to reduce data dimensionality and preserve machine state information while ensuring data privacy and security. First, wavelet packet decomposition and two-dimensional discrete cosine transform extract high-dimensional sub-band features representing machine conditions. Then, a one-way irreversible adaptive hashing coding strategy integrating dictionary learning and symbolic aggregate approximation is proposed, further compressing the sub-band features into a fixed-length machine condition hash with minimal symbols for online monitoring. It transforms kilobyte-long vibration signals into few-byte symbol sequence, significantly improving storage and transmission efficiency at edge terminals. The effectiveness of the proposed method is validated with tool wear and bearing fault datasets. Results demonstrate that the proposed method achieves a good performance in both degradation trend evaluation and fault identification with data privacy preserving.