Research And Development Of Bridge Health Monitoring And Damage Identification Cloud System

As an important infrastructure of our country,bridges play a pivotal role in improving transportation efficiency and boosting the development of the national economy.According to statistics,there are more than 850,000 large and small bridges in China,and the number of bridges is continuously increasing.While enjoying the convenience brought by bridges,we should also pay attention to the daily operation and maintenance of bridges for the long-term economic development and the protection of people’s lives and property.As time goes on,it is inevitable that bridges will be exposed to wind and sun,vehicle loads,and so on which makes them dangerous and affects normal operations.In this paper,a remote health monitoring system is designed for the daily operation and maintenance of the bridge,and the damage identification of the bridge is studied through certain data analysis methods.The main work is summarized as follows:(1)This paper designs a bridge health monitoring cloud platform based on cloud technology.Due to achieve the purpose of remote monitoring,we collect bridge data through sensors such as vibration,acceleration,cracks,displacement,temperature and humidity,and wirelessly upload the data to the cloud platform for storage and analysis through 4G,Lo Ra and other gateways.In order to improve the stability of the cloud platform,a database system that integrates caching,distributed,and security is designed.Besides,the cloud platform includes 7main modules.They are monitoring data upload and hierarchical early warning,historical data query,historical data analysis,expert system,user management,project management,and digital management of manual inspections respectively.(2)Aiming at realizing early warning of bridge damage,a simple beam simulation model in the scene of different depth cracks is designed to acquire the vibration data set.Based on the dataset,we utilize time domain,frequency domain,and time-frequency domain methods to deeply analyze the characteristic performance of the corresponding vibration signals when the cracks depths are different.The experiments have proved the effectiveness of these methods so it could provide a theoretical basis for early damage detection of bridges.(3)The bridge’s vibration characteristics will change accordingly when different crack lengths are happening.Hence depending on this feature we propose a deep neural network framework to do some crack recognition and classification tasks.The crack damage was simulated with a finite element model of a simply supported beam monitored by a group of acceleration sensors,which is the same as(2).Multiple vibration data was fed into designed framework for training and testing.The results are compared with fully convolutional networks(FCN),residual network(Res Net),long short-term memory network(LSTM)and bidirectional LSTM(Bi-LSTM)that are well-performed in tackling time series problems.The experiments show that crack damage is correctly detected and classified using the monitoring data with 3%accuracy improving at least compared with those four methods.Finally,we apply the proposed framework to an SHM and visualize the results on PC via the Internet cloud techniques.