Glushakova, Iuliia1; Zhang, Yu2 and Zhou, Guangchun3
1 Ph.D. Candidate, Key Lab of Structures Dynamic Behavior and Control of Ministry of Education, Key Lab of Smart Prevention and Mitigation of Civil Engineering Disasters of the Ministry of Industry and Information Technology, School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China, iulinchina@yahoo.com
2 Doctor, Key Lab of Structures Dynamic Behavior and Control of Ministry of Education, Key Lab of Smart Prevention and Mitigation of Civil Engineering Disasters of the Ministry of Industry and Information Technology, School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China, zhangyuhit@hit.edu.cn (corresponding author)
3 Professor, Key Lab of Structures Dynamic Behavior and Control of Ministry of Education, Key Lab of Smart Prevention and Mitigation of Civil Engineering Disasters of the Ministry of Industry and Information Technology, School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China, gzhou@hit.edu.cn
KEYWORDS
This paper describes the use of a deep learning neural network (NN) in combination with cellular automata (CA) method to direct prediction of the cracking patterns of masonry panels with openings subjected to lateral loading. The Moore CA neighbourhood model was used to establish the CA numerical model for masonry panels. One masonry panel with an opening was chosen as a base panel to predict the cracking pattern of a new (hypothetical unseen) panel based on establishing similar zones. Also different sizes of base and unseen panels can be used to predict crack formation of one another. The deep learning NN, in this study, was represented by backpropagation NN with two hidden layers. CA numerical model along with dimensions of a panel composed the training data for the NN. Using the cracking pattern of the base panel as the output data, the NN modeled the cracking pattern for an unseen panel. The predictions from the NN have been validated using number of experimental data of masonry wall panels with different sizes.
102
