摘要
为研究典型船体结构的裂纹识别方法,通过模型试验和数值仿真相结合的方法获取了大量可靠的训练样本和验证测试样本。基于BP神经网络技术建立了裂纹损伤识别模型,分别以静态应力参数、动态振动参数、应力与振动参数结合三类信息作为特征参数,对裂纹位置坐标进行了识别,对识别精度进行了对比分析。结果表明利用神经网络模型进行加筋板裂纹损伤识别是可行的,其中以应力与振动参数结合作为裂纹识别的特征信息识别精度最高,绝大多数裂纹位置识别误差在5%之内。研究方法和研究成果可为船体结构裂纹损伤智能化识别提供参考。
In order to study the identification method for crack damage of typical hull structures,a large number of reliable training samples and validation test samples are obtained by combining model experiments with numerical simulations.Based on BP neural network technology,a crack damage identification model is established.The crack location coordinates are identified by using static stress parameters,dynamic vibration parameters,combined stress and vibration parameters as characteristic parameters.The recognition accuracy is compared and analyzed.The results show that the neural network model is feasible for crack damage identification of stiffened plates.Among them,the combination of stress and vibration parameters as the feature information of crack identification has the highest accuracy.Most of the crack location identification errors are within 5%.The research methods and results can provide reference for intelligent identification of crack damage in hull structures.
作者
牟金磊
张仲良
王健
彭飞
闵少松
MU Jinlei;ZHANG Zhongliang;WANG Jian;PENG Fei;MIN Shaosong(College of Naval Architecture and Ocean Engineering,Naval University of Engineering,Wuhan 430033,China)
出处
《船舶工程》
CSCD
北大核心
2020年第5期46-50,共5页
Ship Engineering
基金
国家自然科学基金(51779261)
海军工程大学自然科学基金(20161595)。
关键词
船体结构
加筋板裂纹识别
神经网络
hull structure
crack identification of stiffened plates
neural network
