Fault Time Location And Data Reconstruction About Connected Pipe Optoelectronic Liquid Deflection Sensor

The security problems of bridge structures have been highly valued in academic and engineering circles, and to distinguish the system faults and sensor faults has become a new research focus in recent years. Traditional sensor fault diagnosis usually focuses on space location of sensors in the system. However as for large bridge monitoring system, when sensor fault occurs,it often requires a long period of time to resolve it. To ensure the normal operation of the system during this period of time, the sensor fault time location and fault data reconstruction are of far-reaching significance.This paper mainly studies the photoelectric liquid level deflection sensors of bridge health monitoring system, and Complete the following study:(1) Kalman filter is the most widely application of optimal estimated theory. This paper apply Kalman filter to data preprocess, to reduce the effective suppression of noise error. Experimental analysis shows that the data correlations enhance after filtering, and this provide better support for the follow-up study of a highly effective and accurate fault location method. At the same time, Kalman filtering is one-step recursive estimation, this makes it possible that to fault diagnosis high accuracy and real-time online.(2) This paper proposes a fault time locating method based on correlation analysis between the sliding time windows. The method is based on the strong correlation between the sensors in one group. It also uses an improved model of correlation to make analysis between the sliding time windows. Referred to the correlation value, it can determine faults, thus achieve fault time locating.This paper proposes a fault time locating method based on the standardized residuals. There is only a significant difference in amplitude but a high degree of consistency in variation trend between sensors in one group. It makes data analysis based on the standard residual. Referred to the residual value, it can determine faults, thus achieve fault time locating.Using the two kinds of time locating methods, as for the four types of common fault in the project, make simulation to analysis the validity and accuracy of two kinds of location methods. Experiments show, fault cash more obvious, positioning more accurate; correlation method to decrease in accuracy of fault shows obvious advantages, but the residual method positioning of the constant failure, fixed bias, drift faults performance than correlation method. Combining the two methods, the fault time locating can achieve a better result.(3)This paper proposes a new fault data reconstruction method named adaptive residual method. The method is based on the data of sensors between a group keep the same trend, so the standardized residuals have a trend of 0. With the target of minimum residual value, it can estimate fault sensor data for fault data reconstruction. Make a contrastive analysis of the reconstruction effect and the quantitative comparison of reconstruction residuals with the classical data reconstruction methods: RBF neural network, multiple regression analysis, the least square method. Experiments show, according to this study, the reconstruction effect of adaptive residual method mentioned in this paper is the best.