📄 中文摘要
低温碳钢管道的准确及时泄漏检测对于安全、环境保护和运营成本控制至关重要。现有的声学检测方法存在信噪比低、误报率高以及缺乏实时校准等问题。提出了一种多模态传感器融合框架，集成高频声学传感器、温度和振动传感器以及数字双胞胎模拟，以生成全面的训练数据集。采用混合卷积-长短期记忆网络（LSTM）架构进行端到端训练，利用大规模声学数据集的迁移学习和特定领域的微调。实验基于一个专有的10小时现场数据集和一个公开可用的数据集，验证了该方法的有效性。
深度学习增强的低温碳钢管道声学泄漏检测
📄 English Summary
**Deep Learning‑Enhanced Acoustic Leak Detection for Cryogenic Carbon Steel Pipelines**
Accurate and timely detection of leaks in cryogenic carbon steel pipelines is essential for safety, environmental protection, and operational cost management. Existing acoustic detection methods face challenges such as low signal-to-noise ratios, high false-positive rates, and a lack of real-time calibration. A multimodal sensor fusion framework is proposed, integrating high-frequency acoustic transducers, temperature and vibration gauges, and digital twin simulations to create a comprehensive training corpus. A hybrid convolutional-LSTM architecture is trained end-to-end, leveraging transfer learning from large acoustic datasets and domain-specific fine-tuning. Experiments conducted on a proprietary 10-hour field dataset and a publicly available dataset demonstrate the effectiveness of the proposed approach.