Application of Deep Learning Algorithm in Abnormal Diagnosis of Bridge Health Monitoring Data

With the aging of urban infrastructure, bridge health monitoring has become an important task to ensure public safety. Bridge monitoring usually involves the processing of a large amount of sensor data and image data. However, traditional threshold-based diagnosis methods have problems such as sensitivity to data noise, lack of automation and low accuracy. To this end, this paper proposes a method that combines Stacked Autoencoder (SAE) and Convolutional Neural Network (CNN) for abnormal diagnosis of bridge health monitoring data. First, CNN is used to process image data and automatically extract key image features through convolutional layers. Then, SAE is used to analyze data from sensors and construct a low-dimensional representation of the data through layer-by-layer encoding and decoding. Then, the feature outputs of these two networks are fused to construct a comprehensive model that can process image and sensor data simultaneously. In the experimental conclusion, the research method shows excellent performance in the abnormal diagnosis task in bridge health monitoring. Specifically, in multiple experimental scenarios, the deep learning model based on CNN and SAE outperforms other methods in terms of accuracy, precision, recall and F1 value, with an accuracy of 98.3%. In addition, the model still maintains high robustness in the presence of noise interference and data missing, proving its feasibility and stability in practical applications.