Structure damage diagnosis of bleacher based on DSKNet model

Bleacher usually carries a large number of people, and their safety and stability are critical. Structure damage diagnosis of bleacher can find problems and repair them in time to ensure the safety of personnel. Based on Densely Connected Convolutional Networks (DenseNet) and Selective Kernel Networks (SKNet) models with excellent performance in image recognition, this paper proposes a new DSKNet model for structure damage diagnosis of bleacher. Using the bleacher simulator of Qatar University as the experimental object, the proposed DSKNet model is used to study the damage location and type diagnosis. In addition, the diagnosis results are compared with DenseNet, SKNet, 1DCNN (One Dimensional Convolution Neural Networks), and SVM (support vector machine) models under the same experimental conditions. In order to verify the anti - noise ability of the proposed model in this article, experiments are carried out between the DSKNet model and the above four models under different signal-to-noise ratios. The experimental results show that the DSKNet model can accurately judge the location of the damage. In the damage type experiment, the accuracy of the testing dataset can reach 100% when the model training epoch reaches 30. Under a normal and strong noise environment, the diagnosis performance of the DSKNet model is superior to DenseNet, SKNet, 1DCNN and SVM, which can accurately diagnose the structure damage of bleacher.

Automated summary

This paper proposes a new DSKNet model for structure damage diagnosis of bleacher, which demonstrates superior diagnostic performance and noise resistance in experiments. It can accurately identify the location and type of damage, ensuring the safety of bleacher personnel.