Due to the complexity of the environment, sensors are vulnerable to external disturbance in structure monitoring, which will reduce the quality of monitoring information, and then affect the fusion results. To assess the structural health state reliably, a fusion method considering sensor disturbance is proposed under the framework of evidential reasoning (ER) rule. The reliability of the information provided by the disturbed sensor is measured according to the information entropy method. Based on sensor perception probability, the relative importance of sensors is calculated by sensor monitoring contribution. In addition, a new weighting strategy is proposed to modify the weight of the sensor and fuse the disturbance information of the sensor. Finally, the applicability of the proposed method is illustrated by an example of wharf structure health state assessment.