Abstract：Sutong Bridge located in the tidal reach adopts the pile-group foundation with large-diameter and over-length bored piles working with the steel casing together. Therefore，extensive angle monitoring is adopted to observe the structure response of steel casing，reinforcing steel bar and concrete. Because the sensors of different types and various locations have different working principals，precisions，installing and embedding manners and influence factors，the types and characteristics of measured data have great differences. So，multi-sensor information fusion technology is used to filter abnormal values，extracting characteristic information and constitute integrative decision．However，the observed data of prototype contain the strong noises caused by environmental factors such as the tide，which seriously interferes the analysis and forecasting of data. Furthermore，this type of noise would make the measured data bring about anomalous process and anomalies group，which is difficult to be detected through the redundant information and spatiotemporal contrast of the measured data. On the basis of data detection according to fusion algorithm and gross errors rejection，with the help of wavelet threshold denoising method，the noises caused by a single sensor are further eliminated gradually. According to the credibility and importance for the observed results from different types of sensors，the measured data from multi-sensor after denoising are fused based on the data fusion algorithm of optimal weight distribution principle. The practice has proven that the wavelet layered denoising technology can be used to effectively separate different noises；and the observed data offered by various sensors can be combined in phase by making use of multi-sensor data fusion technology，so that the stress response of pile-group foundation of Sutong Bridge and the surrounding environment can get a more reliable judgment.
