Research On Fault Diagnosis Of Bridge Monitoring Sensor Based On Data Driven

In recent years,with the rapid development of sensor technology and computer science and the need for bridge safety assessment,bridge health monitoring systems have been widely used in existing bridges.The bridge health monitoring system mainly relies on the sensor system to collect parameters related to the safety of the bridge,and then the professionals conduct real-time analysis to evaluate the safety of the bridge structure at any time.During the operation of the bridge health monitoring system,some sensors may fail,making the measured data insufficiently reliable.Therefore,it is necessary to diagnose the sensor failures on a regular basis.Now the bridge health monitoring for sensor fault diagnosis is less research in this sector,the paper combines the measured data archway Wanzhou Yangtze River Bridge health monitoring system,were studied bridge monitoring sensor fault diagnosis based on data-driven,the following key elements:By reviewing data-driven sensor fault diagnosis methods in other fields and combining the characteristics of the bridge health monitoring sensor system,a new datadriven bridge health monitoring sensor fault diagnosis method is designed.The method is divided into two steps.First,we should detect the sensor data anomaly,and then determine whether the data anomaly is caused by sensor failure.In the sensor data anomaly detection stage,two algorithms,principal component analysis in multivariate statistical analysis and long short-term memory neural network in the field of deep learning,are selected,two data anomaly detection methods are designed,and an example is established to compare the two methods.The results of the calculation examples show that the detection method based on the principal component analysis theory is relatively simple,but its detection effectiveness is poor when faced with non-stationary time series and lesser data anomalies.When the long short-term memory neural network is used for data anomaly detection,it has good applicability when facing the above two problems.In the stage of judging the cause of data abnormality,the relevant knowledge of sensor information fusion is used for research.In the feature-level fusion stage,the gray correlation degree,an index that characterizes the correlation of data between various sensors,is selected.In the decision-level fusion,through the analysis of the gray correlation degree in the feature-level fusion,the criterion for judging that the reason for the abnormal data is the sensor failure is summarized.Finally,the data-driven bridge health monitoring sensor fault diagnosis method designed in the thesis is applied to the health monitoring sensor system of Wanzhou Pailou Yangtze River Bridge.The fault diagnosis of the deflection sensor,acceleration sensor and temperature sensor is carried out respectively.It is compared with the manual inspection results of the operating unit of the health monitoring system to prove the practicability of the thesis design method.