基于EMI技术与PSO-BP神经网络铝梁损伤定位研究  

Research of Damage Localization in Aluminum Beam Based on EMI Technique and PSO-BP Neural Networks

作　　者：张军[1] 陈纯洁 ZHANG Jun;CHEN Chunjie(School of Artificial Intelligence,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学人工智能学院,安徽淮南232001

出　　处：《安徽理工大学学报（自然科学版）》2023年第5期1-7,共7页Journal of Anhui University of Science and Technology:Natural Science

基　　金：安徽省重点研究和开发计划基金资助项目(201904a05020004);教育部产学合作协同育人基金资助项目(201901059052);安徽理工大学研究生核心课程项目(2021HX013)。

摘　　要：为了定位铝梁等一维结构的损伤位置,采用压电阻抗(EMI)技术与粒子群(PSO)-BP神经网络进行研究。首先,搭建损伤检测试验平台,使用阻抗分析仪测出健康和损伤状态下的压电导纳曲线,分析不同位置压电陶瓷传感器(PZT)测量的结构谐振峰特征,并通过Pearson相关系数对导纳数据进行处理;然后,构建PSO-BP神经网络,以不同位置上的PZT传感器测得的导纳值作为网络的模式样本进行训练。结果表明,压电阻抗技术能有效识别铝梁健康、损伤工况;Pearson相关系数与PZT传感器和损伤之间的距离呈线性关系,与损伤位置间距越小,PZT测得的导纳曲线的Pearson值越大;选取PZT导纳值变化明显的频率点作为神经网络的输入向量,经过训练后的PSO-BP神经网络能够快速准确地识别铝梁损伤位置。To determine the location of damage in one-dimensional structures such as aluminum beams,the Electro-Mechanical Impedance technique and Particle Swarm(PSO)-BP neural network were employed.Firstly,a damage detection testing platform was constructed,and the admittance curve was measured under both healthy and damaged conditions using an Impedance analyzer.The structural resonance characteristics measured by piezoceramic transducer at different positions were analyzed,and the admittance data were processed by the Pearson correlation coefficient.Subsequently,an PSO-BP neural network was constructed and trained using the admittance values measured by Piezoelectric Ceramic Sensors at different positions as input vectors.Experimental results showed that piezoresistive reactance technology effectively identify the health and damage conditions of aluminum beams The Pearson correlation coefficient has a linear relationship with the distance between the PZT sensor and the damage.Furthermore,the Pearson value of the admittance curve measured by the PZT that was closer to the damage location was found to be larger.The frequency point with obvious change in PZT admittance value selected as the input vector of the neural network,and the trained PSO-BP neural network quickly and accurately the damage location of the aluminum beam.

关 键 词：压电阻抗技术 结构健康监测 PSO-BP神经网络 相关系数 损伤定位 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]