Research On Intelligent Analysis Method For Deformation Monitoring Of Gravity Dam

The deformation monitoring data of gravity dam contain important information of dam deformation.Information mining,analysis and prediction,safety evaluation for the data are important methods to master the safety state of dams.With the development of dam safety monitoring,the collection of monitoring data is becoming more and more compreh,ensive,more and more intelligent,and the amount of data is also increasing,which puts forward higher requirements for the ability of data analysis and processing.How to excavate more useful information from massive data is the basis of understanding the operation behavior of dams.How to preprocess mass data is the premise of improving data quality.It is an important content of dam safety management to keep long-term accurate and scientific and reasonable prediction of effect variable.Therefore,in the context of the rapid development of artificial intelligence,this paper takes the deformation monitoring data of gravity dam as the research object,and introduces data mining,intelligent algorithm,machine learning and other methods to study the analysis methods of monitoring data,such as outlier detection,forecast and safety evaluation.The main research contents and achievements are as follows:(1)Through theoretical analysis and measured data verification,the general law of gravity dam deformation is summarized.Firstly,the correlation analysis of the spatial dimension panel data shows that there is a high linear correlation between different dam sections.On this basis,the shape similarity coefficient is put forward to describe the relationship between the deformation amplitude of different measuring points.The calculation results show that the dam body is symmetrically separated by the middle dam section.Generally,the deformation similarity of the dam section in the symmetrical position is higher,while the deformation similarity of the adjacent dam section is higher.Then,the autocorrelation and partial autocorrelation analysis are carried out on the single measurement point data of the time dimension,so as to know that a certain data is significantly correlated with the previous 1,2,or 3 moments.Finally,the variation characteristics of gravity dam deformation monitoring data are analyzed and summarized from three aspects:global,local and spatial,in order to study the outlier detection algorithm suitable for this kind of data characteristics.(2)The abnormal values in the deformation monitoring data of gravity dam are defined,and the classification and characteristics of the abnormal values are summarized.Then analyzed the outlier detection algorithm based on distance on the applicability of the gravity dam deformation monitoring data,and draw lessons from the thought of this algorithm.Multiple local anomalies coefficient method was proposed,the method by extracting with the readings before k a local window of data,and based on the criterion of setting,can be simple,fast real-time outlier detection.This algorithm is mainly aimed at the pre-processing detection of outliers in the monitoring data,which is an important content for obtaining high-quality data and facilitating the subsequent modeling and analysis.(3)The selection method of the initial solution of the Self-adaptive Differential Evolution Algorithm was improved to improve the global optimization ability of the algorithm,that is the Improved self-adaptive Differential Evolution Algorithm(ISADE)and it was applied to the optimization of the Online Sequential Extreme Learning Machine.A deformation prediction model of gravity dam based on ISADE-OSELM was proposed and established.The model can update the parameters of the existing model only by training the latest data and realize the self-updating of the model,which improves the updating way of the traditional model.At the same time,combining with the optimization algorithm,the precision and generalization of the model are improved.The results show that the comprehensive performance of ISADE-OSELM model is better than that of stepwise regression model,ELM model and OSELM model.(4)On the basis of the traditional confidence interval method,the randomness and fuzziness of the monitoring data are considered,and the cloud model is integrated into the confidence interval method,so that the confidence interval with a clear boundary is extended to a confidence interval with an interval as the boundary.The monitoring index of gravity dam deformation based on cloud confidence interval method is proposed and established.The method takes the error of ISADE-OSELM prediction model as the research object,so it can be updated with the update of ISADE-OSELM model,which makes the method of monitoring index more efficient.The calculation results of the example show that the cloud confidence interval not only has the function of the traditional confidence interval,but also can evaluate the security of the data near the boundary of the confidence interval according to certain membership degree.This evaluation method is more in line with the actual situation,and the deformation monitoring of the dam is more reasonable.