<p>Structural Health Monitoring (SHM) plays a crucial role in extending the service life of buildings and infrastructure by enabling early damage detection and timely maintenance. This study proposes a deep learning-based approach for diagnosing damage in a 3D space frame structure and a real-world bridge. To improve model generalization and efficiency, data augmentation techniques are first applied to diversify the dataset, followed by Principal Component Analysis (PCA) to reduce feature dimensionality while retaining key structural information. A hybrid deep learning architecture combining a One-Dimensional Convolutional Neural Network (1DCNN) and a Bidirectional Gated Recurrent Unit (BiGRU) is then employed, where the 1DCNN extracts salient features from the vibration signals and the BiGRU captures bidirectional temporal dependencies. The proposed PCA-1DCNN-BiGRU model is benchmarked against baseline models including RNN, GRU, LSTM, BiLSTM, BiGRU, and 1DCNN-BiGRU. Experimental results show that the PCA-1DCNN-BiGRU model achieved 93.6% validation accuracy and 92% test accuracy on the space frame dataset, significantly outperforming baseline models. Furthermore, on the Z24 bridge benchmark, the proposed model demonstrated superior performance with average validation and test accuracies of 95.2 and 92.6%, respectively. These findings highlight the effectiveness of integrating PCA, data augmentation, and hybrid deep learning architectures for accurate structural damage detection.</p>