摘要
为保障水利枢纽的泄洪安全,以Inception模块为主体结构,结合门控循环单元(GRU)和高效通道注意力(ECA)机制,提出了Inception-GRU深度神经网络模型,通过采集的多测点泄洪振动数据,智能识别泄洪建筑物的结构安全状态。用工程数据进行测试,结果表明,该模型能以97.15%的准确率完成结构安全智能识别任务,准确率较Inception模型、CNN-GRU模型、Inception-LSTM模型分别提高6.90、5.69、2.03个百分点。对泄洪振动数据进行数据预处理后,以三维矩阵的形式输入网络,可以有效降低模型参数量,提高模型效率。
In order to ensure the flood discharge safety of the water conservancy project,in consideration of inception,gate recurrent unit and effi⁃cient channel attention,a hybrid neural network was proposed to identify intelligently the structural safety of flood discharge buildings based on multi⁃channel vibration data.Tested by experimental data,Inception⁃GRU neural network can identify the structural safety of flood discharge build⁃ings with 97.15%accuracy,with 6.90%,5.69%and 2.03%accuracy better than that of Inception model,CNN⁃GRU model and Inception⁃LSTM model.Data should be pre⁃processed in matrix form before sent into model,so it can effectively reduce parameters and run model faster.
作者
刘昉
陈浩东
梁超
庞博慧
LIU Fang;CHEN Haodong;LIANG Chao;PANG Bohui(State Key Laboratory of Hydraulic Engineering Simulation and Safety,Tianjin University,Tianjin 300354,China;Huaneng Lancang River Hydropower Inc,Kunming 650214,China)
出处
《人民黄河》
CAS
北大核心
2022年第12期101-105,111,共6页
Yellow River
基金
国家自然科学基金资助项目(51909185)。