基于数据融合及残差卷积自编码器的结构损伤识别方法

Data fusion and residual convolutional auto-encoder based structural damage identification

深海混输立管作业期间持续遭受外部风浪流耦合荷载及内部矿液两相流体磨蚀作用,长期作用下结构损伤逐渐积累。立管结构呈高长细比、柔性状态,传统结构损伤识别过程中存在模态参数识别困难、单测点响应损伤敏感度低等诸多问题。针对上述问题,该研究提出了基于数据融合及一维残差卷积自编码器(one dimension residual convolution autoencoder, 1D-RCAE)的深海混输立管结构损伤识别方法,以结构损伤敏感的应变动态响应为输入,使用主成分分析(principal component analysis, PCA)进行多测点应变响应特征融合,进一步利用1D-RCAE自动提取损伤敏感特征,以结构损伤前后敏感特征间的马氏距离构建结构损伤判定指标,实现混输立管结构健康状态监测。通过500 m深海混输立管结构数值模拟和立管物理缩尺模型试验对该研究提出的方法进行验证,结果表明,有限测量信息下能够有效实现深海混输立管结构损伤识别,其中数值模拟验证损伤识别准确率高于99%,物理模型验证损伤识别准确率高于98%。同时探究了噪声污染、海洋环境因素变化对该研究提出方法损伤识别性能的影响规律。

Deep-sea mining riser continuously suffers from internal solid-liquid two-phase fluid abrasion and external wind-wave-current coupling load during operation, and structure damage gradually accumulates. Because of high slenderness ratio and flexibility, it is difficult to identify modal parameters, and the damage sensitivity of single measuring point response is low when applying traditional damage detection methods to deep-sea mining riser. To solve the problems, a data fusion and one-dimension residual convolutional auto-encoder(1D-RCAE) based method was proposed for deep-sea mining riser damage identification. Firstly, principal component analysis(PCA) was used to fuse bending strain responses from multiple measuring points into one variable. Then, the 1D-RCAE was used to extract the damage sensitive feature(DSF) from the fused variable. Lastly, the Mahalanobis distance between the extracted DSF under the currently testing and the baseline conditions was selected as the damage index. The damage detection effectiveness was verified on a 500 m numerical model and a laboratory model of deep-sea mining riser, and the result shows that the accuracy of damage identification is higher than 98%. At the same time, the effects of noise pollution and changing marine environment was explored.