Since the existing mathematical model of single-point monitoring is defective in reflecting the structural behavior of the whole dam and diagnosing dam's abnormal behavior, it's necessary to establish model by relating the data of multiple monitoring points. Based on Bayes Theory of data fusion and taking variance as characteristic parameter, we established a fusion model of diagnosing the abnormal behavior of dam using multi monitoring points, presented the criteria for the model, and provided a project case. As the research shows, the fusion model serves as a new and effective approach for the quantitative description of overall dam behavior and for  the diagnosis of abnormal monitoring points. 