| 声磁注意力融合的地下电缆故障定位方法 |
投稿时间：2026-05-15 修订日期：2026-06-16  点此下载全文 |
| 引用本文： |
| 摘要点击次数: 49 |
| 全文下载次数: 0 |
|
| 基金项目:基于递阶协同的电缆故障可靠定位关键技术 |
|
| 中文摘要:针对地下电缆故障声学与电磁瞬态信号非平稳性强、强噪声与多径反射条件下模态可靠性动态变化导致定位性能不稳定的问题，提出一种声–电磁双模态注意力融合定位方法。首先采用短时傅里叶变换（STFT）将两通道信号转换为时频表征，并利用轻量级卷积神经网络提取故障瞬态特征；随后构建样本级注意力模块，自适应分配两模态权重，抑制受噪声污染通道对融合表示的负面影响；最后通过“区段分类+区段内回归”的联合学习实现粗到细距离估计。实验基于包含多径反射、参数扰动及多种噪声干扰的可控仿真数据集开展。结果表明，在SNR=10 dB条件下，所提方法区段分类准确率达到90.1%，平均绝对误差为3.12 m，优于单模态方法和固定权重融合方法，且在低信噪比场景下表现出更平缓的误差增长趋势。该方法为复杂噪声环境下地下电缆故障的智能定位提供了一种轻量、鲁棒的计算方法 |
| 中文关键词:地下电缆故障定位 声–电磁双模态 时频特征 注意力融合 多任务学习 |
| |
| Underground Cable Fault Location Method Based on Acoustic–Electromagnetic Attention Fusion |
|
|
| Abstract:Underground cable fault localization becomes unreliable when acoustic and electromagnetic transients are both non-stationary and their modality reliability changes under strong noise and multipath reflections. This paper proposes an acoustic-electromagnetic dual-modal attention-fusion method. First, the two synchronized channels are transformed into time-frequency representations by short-time Fourier transform (STFT), and lightweight convolutional neural networks are used to extract transient fault features. Then, a sample-wise attention module adaptively assigns fusion weights to the two modalities, suppressing the negative influence of corrupted observations. Finally, a coarse-to-fine strategy that combines section classification and intra-section regression is adopted for distance estimation. A controllable simulation dataset containing multipath reflections, parameter perturbations, and multiple noise disturbances is constructed for reproducible evaluation. Results show that at SNR = 10 dB, the proposed method achieves a section classification accuracy of 90.1% and a mean absolute error of 3.12 m, outperforming single-modality baselines and fixed-weight fusion methods while exhibiting a slower error increase under low-SNR conditions. The proposed method provides a lightweight and robust computational solution for intelligent underground cable fault localization in complex noisy environments. |
| keywords:underground cable fault location acoustic-electromagnetic dual modality time-frequency features attention fusion multi-task learning |
| 查看全文 查看/发表评论 下载pdf阅读器 |
位访问者,今日访问
