Processing Method of Missing Monitoring Data of Concrete Dam Deformation Based on GRU

Abstract

In view of the missing monitoring data due to instrument problems or human factors in the process of concrete dam deformation monitoring, this paper explores and proposes a processing method for the missing of concrete dam deformation data based on gated recurrent unit neural network (GRU). This processing method has the advantages of easy training and not easy to appear over fitting. It can eliminate the interference caused by the existence of missing data to the analysis of dam deformation behavior, and then improve the accuracy of objective evaluation or early warning of dam deformation behavior. Through an engineering example, the processing method of missing value of deformation data of concrete dam proposed in this paper is verified. The results show that the prediction residual value of this method is small, which is lower than the error limit specified in the specification. It can effectively process the missing data of deformation monitoring, and provide objective monitoring data for the subsequent analysis of dam deformation behavior.

Keywords

• concrete dam
• deformation monitoring
• missing data
• neural network of gated cycle unit