Data Management And Prediction Evaluation System Of Concrete Compressive Strength Deterioration In Marine Environment

Entering the 21 st century,the ocean has become the commanding height of political,economic and military competition among countries in the world.At present,China is also stepping up its efforts in the construction of Marine infrastructure.For example,the Hong Kong-Zhuhai-Macao Bridge was officially opened on October 24,2018,becoming the world's longest sea-crossing bridge,marking another important step for China to become a superpower by the construction of superengineering.However,the harsh service environment of the ocean leads to serious erosion and damage of concrete,which seriously affects the service life and safety of Marine infrastructure.Therefore,how to make concrete materials have a service life of 100 or even 120 years,that is,the durability of concrete in the Marine environment,has become a research hotspot for civil engineering materials researchers.This project is also based on the 973 national key research and development project(basic research on the long life of concrete materials and structures in harsh environment 2015CB655100)to carry out relevant research.For durability design and life prediction of concrete,the traditional method is to analyze the degradation mechanism and erosion process of concrete from the perspectives of physics,chemistry and biochemistry,and then build mathematical physical model and calculate it by mathematical analytical or numerical solution,and finally get the fitting prediction model.This method has strong theory and interpretability,but a lot of simplification has been done in the construction process,so the applicability and extrapolation ability of the model are limited.From the perspective of "big data analysis",this paper adopts the methods of "data mining" and "machine learning" to realize the prediction of concrete compressive strength in the Marine environment based on a large amount of real data.In addition,data management,data prediction and other functions are integrated to build an integrated big data analysis platform with functions of data collection,management,analysis and prediction,so as to provide reliable data reference and decision support for durability workers of Marine environment concrete.Before the construction of the concrete compressive strength degradation data management and prediction and evaluation system under the Marine environment,the 2018 data supplement of SQL database and the three-step optimization of SVM model were completed.Firstly,the data were collected by literature retrieval,and the extracted data were normalized by strength,type of cement and dry-wet cycle system.Finally,617 sets of authentic data were supplemented,and the total number of databases reached 10,184 sets.Then based on the best SVM model compared to artificial neural network and decision tree model in the application condition of this paper,three-step optimizations including the completion and main factors optimization of input parameters and concrete clustering optimization were carried out.Finally the average relative error of the original SVM has decreased to 14.4% from 33.3%,the median relative error is reduced to 10.8% from 23.4%,data points of relative error less than 30% are increased from 245 to 344 in 400 prediction points with the proportion of 86%.In addition,the strength of concrete with load parameters is realized through the transformation from dry and wet cycle to salt crystallization,which extends the application scope of the SVM prediction model of this system.The construction of the system adopts C# programming language with combining the SQL Server database and Matlab analysis application to realize the integration of data management,data prediction and other functions,and it finally builds a humancomputer interaction system that is convenient for the administrator to maintain and the user to use.Among them,"data management" is designed with the framework of "tree" to realize diversified data management functions,such as "data addition","data browsing management","data statistics" and "data option setting",etc.,which can meet the comprehensive management requirements of the administrator for data."Data prediction" adopts a more user-friendly and operational framework and a "linear" framework and operational process,which is carried out in a "step-by-step guidance" way to help users complete the process of "prediction parameter input","parameter confirmation","SVM strength prediction" and "result display".The final prediction results are presented visually with the combination of charts and graphs,and the reference deviations calculated by the system automatically,so as to realize the prediction and evaluation of users on the strength results.Finally,the system also provides the open and sustainable application program.The cloud computing mode is adopted to deploy the system on the cloud server with more hardware advantages,so that the computing speed and application experience of the system are improved and users can operate it through the web page.Secondly,based on the research results of the positive ratio between the prediction accuracy of SVM model and the data volume of training set,through collecting the data provided by the user for activating the system and using the system self-inspection to judge the validity of the newly added data,the health and sustainable development of the system is finally realized.