基于改进残差网络的文化遗产区地震监测响应判别方法研究OA

To evaluate seismic responses in cultural heritage building complexes,a three-channel residual net-work(TF-ResNet)with an attention mechanism is proposed.The framework employs a residual neural net-work(ResNet)as its backbone,integrating an efficient channel attention module and a joint spatial-channel convolutional attention module.A three-channel parallel input architecture is designed,and multi-source fea-ture fusion is realized through fully connected layers,thereby enhancing the discrimination accuracy of seismic responses.Using the seismic motion data from a group measurement point in a cultural heritage area,the TF-ResNet validation experiments are conducted,and the ablation and comparison tests are carried out with the existing deep learning models.Five evaluation metrics,including the accuracy,precision,recall,F1 score,and confusion matrix,are employed to evaluate the model performance.The results demonstrate that the classification accuracy of the TF-ResNet achieves 90.67%,and the engineering validation accuracy is 92.93%.The TF-ResNet can significantly improve the discrimination accuracy and stability of seismic re-sponses for building complexes,providing a practical technical solution for earthquake monitoring systems.