| 输电铁塔紧固件松动声纹特征分析及智能识别算法研究 |
投稿时间：2026-05-06 修订日期：2026-05-25  点此下载全文 |
| 引用本文： |
| 摘要点击次数: 68 |
| 全文下载次数: 0 |
|
| 基金项目:输电铁塔螺栓松动声纹检测技术研究与智能检测装置研制 |
|
| 中文摘要:针对输电铁塔紧固件在风振、温差交变及长期服役条件下易出现预紧力衰减和连接松动的问题，提出一种面向现场巡检应用的多域声纹特征融合与轻量化识别方法。首先，建立锤击激励声信号采集流程，并以额定扭矩T0为基准构建正常、轻微松动、中度松动和严重松动4类状态样本；其次，按“带通滤波—谱减降噪—端点检测—分帧归一化”的顺序建立声信号预处理模型，提取时域、频域、倒谱域和小波包能量特征，形成松动声纹融合向量；最后，构建1D-CNN-BiGRU-Attention网络，实现紧固件松动等级判别。实验部分通过对比实验、消融实验、抗噪实验和部署效率测试验证所提方法的可行性与有效性，结果表明该方法能够有效区分不同紧固状态，并具备边缘端辅助巡检应用潜力。 |
| 中文关键词:输电铁塔 紧固件松动 声纹特征 1D-CNN-BiGRU-Attention 故障诊断 |
| |
| Soundprint Feature Analysis and Intelligent Recognition Algorithm for Fastener Looseness of Transmission Towers |
|
|
| Abstract::To address fastener looseness in transmission towers caused by wind-induced vibration, temperature variation and long-term service, a field-oriented recognition method based on multi-domain soundprint fusion and a lightweight neural network is proposed. An impact-excitation acoustic acquisition procedure is established, and four fastening states are constructed according to the rated torque T0. The acoustic signal is processed by band-pass filtering, spectral-subtraction denoising, endpoint detection, framing and normalization. Time-domain, frequency-domain, cepstral and wavelet-packet energy features are fused as a soundprint vector. A 1D-CNN-BiGRU-Attention model is then developed for looseness-level recognition. Comparative, ablation, anti-noise and deployment-efficiency experiments verify the feasibility and effectiveness of the proposed method, showing its potential for edge-side assisted inspection. |
| keywords:transmission tower fastener looseness soundprint feature 1D-CNN-BiGRU-Attention fault diagnosis |
| 查看全文 查看/发表评论 下载pdf阅读器 |
位访问者,今日访问
