Scholarlink

<p>Salt crystallization cycles in museum mural pigments can generate weak and intermittent acoustic emission (AE) signals that are easily masked by noise, making early damage detection difficult. This study presents an AE-based ultra-early warning framework for pigment-layer degradation under dynamic temperature–humidity loading. Under controlled cycling conditions, AE and environmental data were synchronously collected and aligned into overlapping time windows. Multi-scale features combining environmental trends, AE statistics, burst characteristics, and parameter histograms were extracted to capture faint precursors. The warning task was formulated as a sequence-based prediction problem, using 20 min of historical data to forecast high-risk AE activity within the next 2 min. Thresholds were optimized using segment-level F<sub>1</sub>-score under a false-positive constraint, and temporal post-processing was applied to generate coherent alarm episodes and a two-level alert scheme. A gated recurrent unit model achieved the best balance of accuracy, recall, and alarm continuity, supporting preventive conservation of mural heritage.</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

Acoustic emission monitoring for early warning of pigment-layer degradation in mural heritage under dynamic environmental loading

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月7日

查看全文
  • Zhiyong Lu,
  • Yuanzhi Zhu,
  • Huiyang Xiao,
  • Jia Wang,
  • Zhang He,
  • Yijian Cao,
  • Zhengzheng Wu,
  • Xiaonan Liu

摘要

Salt crystallization cycles in museum mural pigments can generate weak and intermittent acoustic emission (AE) signals that are easily masked by noise, making early damage detection difficult. This study presents an AE-based ultra-early warning framework for pigment-layer degradation under dynamic temperature–humidity loading. Under controlled cycling conditions, AE and environmental data were synchronously collected and aligned into overlapping time windows. Multi-scale features combining environmental trends, AE statistics, burst characteristics, and parameter histograms were extracted to capture faint precursors. The warning task was formulated as a sequence-based prediction problem, using 20 min of historical data to forecast high-risk AE activity within the next 2 min. Thresholds were optimized using segment-level F1-score under a false-positive constraint, and temporal post-processing was applied to generate coherent alarm episodes and a two-level alert scheme. A gated recurrent unit model achieved the best balance of accuracy, recall, and alarm continuity, supporting preventive conservation of mural heritage.