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

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. © 2024 Nanjing University of Aeronautics an Astronautics. All rights reserved.