Design And Application Of Concrete Intelligent Mix Proportion Based On Deep Learning

In the construction industry,mix proportion design often requires manual trial matching in the laboratory to obtain a formal construction mix ratio.This time consumption not only increases the waste of materials,but also increases the cost of concrete production.At the same time,the accurate concrete 28 d compressive strength prediction model also plays an important role in the design of intelligent mix proportions of concrete.Therefore,this thesis conducts the research on intelligent mix proportion design based on ensemble learning,deep learning and Monte Carlo method.Aiming at improving the prediction accuracy of 28 d compressive strength of concrete,a regression prediction model based on ensemble learning is proposed.Firstly,the data with missing,erroneous measured and unlabeled data are filled and deleted,and feature selection is performed by analyzing the Pearson correlation coefficient between attributes.Secondly,considering the accuracy of integrating multiple models is higher than single model,a regression prediction model based on ensemble learning is designed by integrating the base predictor Random Forest with Ada Boost framework.Finally,the proposed method is compared with the other seven regression prediction methods on two real concrete data sets.The results show that the proposed ensemble learning model is more effective for improving the prediction accuracy of concrete 28 d compressive strength.Aiming at solving the problem of intelligent design of concrete mix proportion,a concrete intelligent mix proportion design model based on deep learning is proposed.First of all,according to the user's requirements,fuzzy matching is performed in the database to output a series of matching ratios that meet the requirements of users.Secondly,the expert knowledge base is used to score the mix proportion.The expert knowledge base is composed of statistical methods to summarize the construction data set rules of a construction group in 2018,the national design standards of concrete mix proportion,and the expert scoring mechanism.Then,Monte Carlo method is used to fine tune the mix proportion,while the 28 d compressive strength prediction model based on deep learning and the national design standard are used to constrain the mix proportion design process.The cost of concrete mix proportion is taken as the optimization target to find the optimal mix proportion and feed back to user.Finally,the experimental results show that the designed mix proportion meets the national design standards and user's requirements,and also saves production cost.According to the concrete intelligent mix proportion design model based on deep learning,a concrete intelligent mix proportion system is designed and implemented.The system has two main functions,including the prediction of concrete compressive strength and the intelligent design of concrete mix proportion.The interface is simple,efficient and easy to operate.