Application Of Deep Learning In Bridge Health Diagnosis

In the long-term operation of the bridge structure,the hazardous events,such as structural overrun,overdue service,fatigue,corrosion,etc.,will lead to serious structural harbor security risks,which may cause significant loss of life and property.Therefore,to monitor the status of the bridge timely are demanded in the bridge health monitoring system.As the core of the monitoring system,bridge health diagnosis still use the diagnostic method based on the precise structural model.For the structural parameters of the bridge in the long-term service,the establishment of a precise structural model of health diagnosis method has been slightly behind.The bridge health monitoring system collects the massive data,and the evolution rule of the structural system performance can be excavated from the information of the structural damage characteristics.Based on the great advantage of deep learning in large data processing,this paper studies the application of deep learning in bridge health diagnosis.The main research contents are summarized as following:(1)The importance of bridge health diagnosis is discussed,the existing bridge health diagnosis methods are summarized,and their shortcomings are analyzed.Compare with the advantages of application of deep learning in bridge health diagnosis,which highlights the importance of bridge health diagnosis based on deep learning.(2)Bridge health diagnosis methods based on the deep belief network model and the deep auto-encoder network model are proposed.And two kinds of diagnosis methods are trained and tested by real bridge data.(3)Taking the monitoring data of Masangxi Yangtze River Bridge in Chongqing as an example,this paper puts forward the practical application of bridge health diagnosis methods based on deep belief network and deep auto-encoder network.The results indicate that the bridge health diagnosis methods based on the deep belief network and the deep auto-encoder network can effectively diagnose the health status of the bridge.And the tests indicate that the diagnostic accuracy of the diagnostic method based on deep auto-encoder network is slightly higher than the bridge health diagnosis method based on the deep belief network.