基于改进Elman神经网络的CFRP补强钢板界面脱粘预测研究  

Prediction of interfacial debonding of CFRP reinforced steel plate based onimproved Elman neural network

作　　者：王庆松 张玉[1] 张洪雨 陈柏桦 WANG Qingsong;ZHANG Yu;ZHANG Hongyu;CHEN Baihua(College of Safety and Ocean Engineering,China University of Petroleum,Beijing 102249,China;Sino-Pipeline International Company Limited,Beijing 102206,China)

机构地区：[1]中国石油大学(北京)安全与海洋工程学院,北京102249 [2]中油国际管道有限公司,北京102206

出　　处：《振动与冲击》2024年第3期120-127,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(52222111)。

摘　　要：针对碳纤维复合材料(carbon fiber reinforced polymer, CFRP)补强钢结构出现内部界面脱粘损伤后难以观测的问题,结合Lamb波检测方法和神经网络提出了一种界面脱粘预测方法。搭建了基于Lamb波的CFRP补强钢板信号分析试验平台,利用ABAQUS软件建立了CFRP补强钢板的机电耦合有限元模型,并通过试验验证了有限元模型的准确性。将长方形和圆形两种脱粘形状的信号在时域和频域内进行分析,基于自适应遗传算法改进的Elman神经网络建立了CFRP补强钢板脱粘预测模型,并将与脱粘面积相关性较高的信号特征数据作为预测模型的特征数据。对预测模型进行性能测试,脱粘形状为长方形和圆形预测值的平均绝对百分比误差分别为3.03%和8.06%,结果表明改进的Elman网络对于脱粘损伤具有较好的预测精度。Here,aiming at the problem of difficulty in observing internal interfacial debonding damages of steel structures reinforced by carbon fiber reinforced polymer(CFRP),a prediction method for interfacial debonding was proposed by combining Lamb wave detection method and neural network.A Lamb wave-based test platform for signal analysis of CFRP reinforced steel plate was established,and an electro-mechanical coupled finite element(FE)model for CFRP reinforced steel plate was established using the FE software ABAQUS.The correctness of the FE model was verified with tests.Signals of rectangular and circular debonding shapes were analyzed in time domain and frequency domain,and a prediction model for debonding of CFRP reinforced steel plate was established based on Elman neural network improved with adaptive genetic algorithm.Signal feature data with higher correlation with debonding area were taken as the feature data for the prediction model.Performance testing for the prediction model showed that average absolute percentage errors of predicted values for rectangular and circular debonding shapes are 3.03%and 8.06%,respectively;the improved Elman network has better prediction accuracy for debonding damages.

关 键 词：界面脱粘 LAMB波 碳纤维复合材料(CFRP) 脱粘预测 ELMAN神经网络 

分 类 号：TB553[理学—物理]