一种基于轻量化神经网络面向富水岩体破裂前兆识别方法
By constructing a multi-scale fusion module and a depthwise separable convolution module using a lightweight neural network, combined with a channel attention mechanism and a global context classifier, the problem of identifying weak fracture signals in water-rich rock masses is solved, achieving high-precision, real-time identification of fracture precursors, and is suitable for resource-constrained edge devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SANYA SCI & EDUCATION INNOVATION PARK WUHAN UNIV OF TECH
- Filing Date
- 2026-04-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to accurately capture weak fracture signals in water-rich rock masses, suffer from severe identification biases due to fluid noise interference, have insufficient model generalization capabilities, and are computationally complex enough to hinder real-time deployment, thus failing to meet the high robustness and real-time requirements of water-rich rock mass engineering.
A lightweight neural network is used, which combines a multi-scale fusion module, a depthwise separable convolution module, a channel attention weighting module, and a global context classifier to build a lightweight model. The model training process is optimized through multi-scale parallel convolution, depthwise convolution, and channel attention mechanisms. Focal Loss and physical mechanism data augmentation strategies are introduced to achieve high-precision identification of rupture precursors.
It achieves high-precision identification of weak fracture signals in water-rich environments, reduces the impact of fluid noise interference, improves the model's generalization ability and real-time performance, is suitable for resource-constrained edge devices, and meets the real-time monitoring needs of water-rich rock mass engineering.
Smart Images
Figure CN122087731B_ABST