Intelligent Seepage Analysis For Large-scale Sonar Data

Seepage theory is widely used in many fields such as construction,water conservation,geology,environmental protection,chemical industry,and biology,etc.It can provide a basis for developing water resources and preventing infiltration deformation of building foundations.In the construction industry,for example,almost all construction activities are accompanied by the participation of seepage,and the annual personal and economic losses caused by seepage accidents are incalculable.At present,the industry mostly uses sonar technology to detect seepage defects in underground concealed works under natural flow fields,and in a production environment,the technology requires the classification of different types of sonar seepage data.Moreover,due to the variable conditions and large volume of seepage data,the existing expert manual classification methods can no longer meet the needs of real-time classification,so it is necessary to analyze the seepage and noise waveforms with the help of computer technology.Based on this,this paper discusses the intelligent seepage analysis technology for large-scale sonar data to analyze the large-scale sonar data and explore the potential laws after different sonar seepage manifestations.The main work of this paper is shown below:1.A sonar data classification model called SWCD(Sonar Waveform Classification based on Deep GBM)is proposed.The model can flexibly select different feature extraction methods according to the size of the sonar seepage sample set and the actual requirements for training time,so as to train a sonar classifier with good results.The model first extracts36-dimensional data from sonar waveforms,and then uses different dimensionality reduction methods to form three sub-models,SWCD-R,SWCD-P and SWCD-G.Finally,the sonar data are classified by Deep GBM model to distinguish different sonar waveforms from noise in order to achieve the purpose of screening sonar data.2.A model called MSDCL(Mass Sonar Data Classification based on Light GBM)is proposed.The model is based on Spark framework and serves to perform classification and data mining operations on seepage waveform files in a distributed environment at a faster rate.The model first continues to reduce the dimensionality based on SWCD feature extraction by algorithms such as Relief F,and then uses the Spark-based distributed Light GBM model for data classification,and finally employs decision trees for data mining of the massive data stored in HDFS.3.An intelligent seepage analysis platform for large-scale sonar data is implemented,which includes five modules including user management module,sonar data management module,sonar waveform analysis module,sonar denoising module and sonar data mining module.The platform integrates the above SWCD classification model and MSDCL distributed analysis model,which reduces the cost of seepage analysis,improves the speed of seepage analysis,and standardizes the seepage analysis process,and is a one-stop seepage analysis platform that can meet the daily use of seepage analysis engineers.