结合多通道MTF和CNN的框架结构损伤识别方法

Damage Identification Method Using Multi-channel Markov Transition Field and Convolutional Neural Network for Frame Structures

为提高复杂框架结构损伤识别的准确率,提出了一种基于多通道马尔可夫变迁场(multi-channel Markov transition field,简称MCMTF)和卷积神经网络(convolutional neural network,简称CNN)的框架结构损伤识别方法。首先,采用MCMTF理论将原始一维振动信号转换为二维图像,实现数据升维和多通道数据融合;其次,以MCMTF转换后的图像数据集作为输入训练CNN模型;最后,经调参优化自动提取损伤敏感特征,并实现损伤识别。将该方法应用于IASC-ASCE Benchmark框架结构数值模型及3层钢框架结构模型试验,对比研究了多通道MTF、单通道MTF和原始数据矩阵3种数据输入方式,CNN、长短时记忆(long short term memory,简称LSTM)神经网络和深度神经网络(deep neural network,简称DNN)3种网络模型,以及噪声对框架结构损伤识别准确率的影响。结果表明:MCMTF与CNN结合方法的损伤识别准确率最优且具有良好的鲁棒性,其对Benchmark框架数值模型模拟损伤的识别准确率可达94.4%,对3层钢框架试验模型实际损伤的识别准确率可达98.4%。

To improve the accuracy of damage identification on complicated frame structures,a damage identification method based on multi-channel Markov transition field(MCMTF)and convolutional neural network(CNN)is proposed.First,MCMTF is adopted to transform the original one-dimensional vibration signals into two-dimensional images,which can realize the data dimension elevation and multi-channel data fusion.Then,the image datasets transformed by MCMTF are used as the input to train the CNN models.Finally,the sensitive damage features are automatically extracted after parameter tuning and optimization to identify the damage patterns.This method is applied to the IASC-ASCE Benchmark model and an experimental three-layer steel frame structure.The influence of three different data input modes including multi-channel MTF,single-channel MTF and original data matrix are investigated.Further,three different models including CNN,long short-term memory(LSTM)neural network and deep neural network(DNN)are compared,and the different noise levels on damage identification performance are obtained.The results show that the proposed method combing MCMTF with CNN has advantages in the accuracy of damage identification and good robustness.The damage identification accuracy of the method is 94.4%for the numerical Benchmark model and 98.4%for the laboratory three-layer steel frame structure,respectively.