Particle Filter-based Identification Of Structural Damage And Sensor Fault

The identification of structural damage using parameter estimation has been extensively investigated in structural health monitoring(SHM),however,the data measured by sensor is always assumed reliable in the application of these methods,which means sensors are healthy in other words.Since the service life of sensor lasting for several or ten years is shorter than that of the civil structure,sensor fault will inevitably occur during the service period of civil structure.The structural damage identification assuming sensors are healthy and sensor fault diagnosis assuming structure is healthy may trigger a false alarm when the assumption is contradicted with reality.Hence,it is of great importance to evaluate the health condition of structure and sensors simultaneously for the reliability and precision of SHM system.In this thesis,two new parameter estimation models are proposed based on traditional structural parameter identification(SPI)model and the research on the identification of structural damage and sensor fault are carried out using parameter estimation technique.Particle filter serves as the estimation technique due to its good performance in nonlinear and non-Gaussian system.The main contributions of this thesis are as follows:(1)Since it is difficult to accomplish sensor fault diagnosis using the traditional structural parameter identification(SPI)model,the identification of structural damage and sensor fault is implemented with a novel model called structure and sensor identification(SSI)model,where structural parameters and sensor parameters are simultaneously augmented into the state vector.The estimated results obtained by particle filter are used for fault diagnosis with x-bar control chart.It is demonstrated in two-storeys frame simulation under ground motion and three-storeys frame experiment under impact force that this simple procedure is able to identify structural damage,sensor faults(gain,bias,drift and precision degradation)and simultaneous occurrence of structural damage and sensor fault.(2)Since the SSI model is highly nonlinear and it is difficult to accurately obtain structural parameters and sensor parameters without known external force or ground motion.A sensor fault identification(SFI)model is proposed with only sensor parameters added into the original state vector.The traditional SPI model is applied to detect abnormality,and then SFI model is used to distinguish between structural damage and sensor faults and to assess sensor health condition.It is proven in two-storeys frame simulation during free vibration and three-storeys frame experiment during free vibration that this method with two models is able to identify structural damage and sensor faults(gain,bias).