| Dam safety monitoring is an important means to grasp the operational status of dams,and the monitoring data provides the basis for dam condition assessment and structural safety evaluation.The analysis and processing of dam safety monitoring data can identify potential dam safety problems always and provide guidance for later maintenance and operation.The purpose of this paper is to analyze the massive monitoring data of spatial multi-effects of an arch dam.This paper based on an arch dam,aims to analyze its massive monitoring data of spatial multi-effects and effectively solve the problems of mining massive date information and extract key information by principal component analysis,random forest algorithm,Spearman rank correlation analysis,recurrent neural network algorithm,etc.The main results are as follows.(1)Research on the method of compression and reconstruction of dam multisource monitoring data based on the principal component analysis method.Firstly,use spatial field multi-effect monitoring data to constitute the multi-source monitoring data set,Secondly,study and apply multiple interpolation and random forest algorithms to interpolate and complement the missing values by integrated learning.Finally,recombine multi-source data sequences by principal component analysis method to form a new set of comprehensive indicators,and realize mass compression and extraction of comprehensive information,to achieve the data dimension reduction and mining of efficient data.(2)Spearman rank correlation analysis based on the dam multi-source multieffect correlation analysis method research.Considering the characteristics of strong structural integrity of arch dams and strong data correlation,the Spearman rank correlation coefficient analysis method suitable for dam monitoring data is studied;aiming at the dimensionality reduction data extracted by principal component analysis method,the correlation of multi-physical quantity monitoring data of different parts is analyzed according to the spatial distribution of monitorinsg points,and the connotation relationship between multi-effect monitoring data of dams is mined and extracted.(3)Research on recurrent nural network prediction model of dam monitoring volume based on multi-effect correlation.First,Analyze the recurrent neural network model which suitable for time series data,and study the method of modified GRU as the model processing unit.Second,select the monitoring points with strong rank correlation as the main influencing factors,establish the prediction model of multieffect correlation GRU cyclic neural network.Then,create the model and analyze the fracture data with strong nonlinearity by using the network model.And after comparing the results with the prediction results of linear regression model,ridge regression model and other statistical methods,it shows that the model is with higher computational accuracy and better prediction performance.Aiming at multi-source data analysis of dam spatial field,this paper studied the method of analysis and processing of massive linked data and its learning algorithms based on machine learning method.The results can provide new research ideas for big data analysis,data mining and intelligent analysis of dam monitoring information,and provide reference for dam intelligent monitoring and intelligent management. |
