Due to natural disasters and man-made reasons, bridges are prone to structural damage during long-term usage, which reduces the associated carrying capacity, increases natural aging, and reduces safety. It is urgent to monitor the health status of bridge structure via intelligent technology. This paper proposes a bridge fault recognition structure. First, the signals of bridge parameter are collected by using distributed sensors. Then, the collected signals are processed by signal processing to extract the features in time and frequency domain. Lastly, the extracted features are used to learn an intelligent classifier. The large margin distribution machine is adopted as a classification model. The experimental results have proven the feasibility of the proposed bridge fault recognition structure.